CROSS-REFERENCE TO RELATED APPLICATIONS

The present application claims the benefit of priority based on:

- U.S. provisional patent application no. 61/450,800 filed on March 9, 2011; and

- U.S. provisional patent application no. 61/504,479 filed on July 5, 2011;

The contents of the aforementioned documents are incorporated herein by reference.

The present application is also related to international PCT patent application no. PCT/CA2011/000024 filed January 7, 2011 and entitled "Method and apparatus for performing a security scan", which claims the benefit of priority based on U.S. provisional patent application no. 61/293,132 filed on January 7, 2010. The contents of the aforementioned document are incorporated herein by reference.

FIELD OF THE INVENTION

The invention relates to the Advanced Imaging Technology (AIT) field and more specifically to techniques for the use of such technology in determining if a clothed individual presents a security threat. Such techniques can be used for determining topographic information about the person's body, in particular it can be used as part of a security checkpoint application to render an image of the person's body on a display monitor and to determine whether the person carries concealed weapons, explosives or contraband material. Typical uses for this technology include detection of items for commercial loss prevention, smuggling and screening at government buildings and airports security checkpoints. BACKGROUND

Body scanners are increasingly being used at airports security checkpoints or other security sensitive locations to determine if clothed individuals present a security threat. Advanced Imaging Technology (AIT) typically makes use of data generated by a whole-body imaging device, also referred to as a body scanner. A body scanner operates by interrogating the subject with electromagnetic radiation, such as millimeter waves, that easily penetrates clothing but is reflected by the skin. In this fashion, it is possible to create an image of the subject in a virtually "unclothed" state likely to reveal potential threats, such as weapons, explosives or contraband material.

Body scanners are very useful because they can perform a security scan relatively rapidly, in a folly automated fashion. Most importantly, they replace the more traditional "pat-down" search that requires a security agent to physically touch the person to check for threats concealed under the clothing. In addition to its lack of effectiveness, the "pat-down" search is also objectionable since it can be very uncomfortable for the subject.

Conventionally, the body scanner gathers body image data and renders an image of the person being screened on a display device. The rendered image is visually examined by a security agent to determine if the subject presents a security threat.

A deficiency with conventional systems is that they are reliant on the security agent (a human operator) to identify potentially tlireatening objects on the subject. However, the performance of the security agent greatly varies according to such factors as poor training and fatigue. As such, the process of detection and identification of threatening objects is highly susceptible to human error.

Another deficiency is that the images displayed provide little, if any, guidance as to what is being observed. It will be appreciated that failure to identify a threatening object, such as a weapon for example, may have serious consequences, such as property damage, injuries and human deaths. Moreover, if the subject has a container of liquid on his/her person, determining whether the container of liquid constitutes a threat cannot be ascertained based on a visual examination of the image alone. This has important consequence since in certain situations the container of liquid may not be easily discarded by the subject under inspection. Some examples of containers of liquid that may not easily be discarded by a subject include implants (e.g. breast implants), portable dialysis devices and a urostomy bag amongst others. Consequently, there is a need in the industry for providing a method and system for use in performing a security scan on a person that alleviates at least in part the deficiencies of the prior art.

SUMMARY

In accordance with a first aspect, the present invention provides processes, systems and devices for at least partially automating the processing of body image data generated by a full body scanning devices in order to identify the presence of a potential anomaly. In a first specific example, gender-based contextual information is used to influence what is "expected" to be seen in a body image and to detect an anomaly when an observed feature in a body image is inconsistent with what is "expected".

In a second specific example, body-part-based contextual information is used to influence what is "expected" to be seen in a body image for a given location on the body of the individual being screened and to detect an anomaly when an observed feature on that portion of the body image is inconsistent with what is "expected". In accordance with another aspect, the invention provides a system for performing security screening at security checkpoint on an individual. The system comprises:

- a full body scanning device for subjecting the individual to a scan from a source of radiation to generate body image data;

- a processor programmed for processing the body image data to determine if the body image data conveys an anomaly.

In a specific example of implementation, the full body scanning device directs millimeter wave energy at the individual under inspection to generate the body image data, wherein the body image data conveys energy reflected by the individual under inspection.

In a specific implementation, the processor is programmed for processing of the body image data generated by the full body scanning device in order to identify the presence of an anomaly. In a specific implementation, the processing of the body image data comprises:

- processing the ("raw") body image data generated by AIT scanners to perform image reconstruction and information extraction; and

- processing the reconstructed image using an automated threat detection process to detect the presence of potential threats and/or other target obj ects/substances .

In accordance with another aspect, the invention provides a method for performing security screening at a security checkpoint on an individual. The method comprises receiving body image data derived by performing an inspection of the individual using a body scanning device. The method also comprises processing the body image data with a computing device programmed with software to identify potential anomalies at least in part based on gender information associated with the individual and releasing data conveying at least some of the identified potential anomalies. Advantageously, by making use of gender information associated with the individual when detecting anomalies, the method provided allows taking into account gender-based contextual information when detecting anomalies in a body image. In accordance with a specific example of implementation, the method comprises identifying a potential anomaly at least in part by using the computing device to perform a comparison between observed information derived by processing the body image data and expected information derived at least in part based on the gender information associated with the individual.

In accordance with the specific example of implementation, if the gender information conveys that the individual is more likely to be a male, the image data is processed to identify anomalies at least in part based on male contextual information. Conversely if the gender information conveys that the individual is more likely to be a female, the image data is processed to identify anomalies at least in part based on female contextual information.

In a first non-limiting example, the computing device is used to perform a comparison between observed body proportions derived by processing the body image data and expected body proportions derived at least in part based on the gender information associated with the individual. A potential anomaly is identified when the observed body proportions differ from the expected body proportions by an amount exceeding a threshold amount. In a second non-limiting example, the computing device is used to perform a comparison between an observed body geometry derived by processing the body image data and an expected body geometry derived at least in part based on the gender information associated with the individual. A potential anomaly is identified when the observed body geometry differs from the expected body geometry by an amount exceeding a threshold amount. For example, if the gender information conveys that the individual is more likely to be a female, if the observed body geometry of the crotch area is not consistent (or differs the expected body geometry by an amount exceeding a threshold amount) with the expected body geometry of the crotch area for a female, then an anomaly can be identified.

In a third non-limiting example, the computing device is used to perform a comparison between observed items worn derived by processing the body image data and expected items worn derived at least in part based on the gender information associated with the individual. Examples of expected items worn may include, without being limited to, shoe types and undergarment types. A potential anomaly is identified when the observed items worn are inconsistent with the expected items worn. For example, if the individual is a male, it would not be expected that we would find a brassiere on his chest area. Separately or in combination with the above features, in specific examples implementation, the method comprises receiving data conveying the gender information associated with the individual being screened.

Separately or in combination with the above features, in alternative examples of implementation, the body scanner includes an interface allowing a human operator to specify gender information conveying whether the individual being screened is a male or a female, and the method comprises receiving data from the body scanner conveying the specified gender information. In accordance with such alternative examples of implementation, the interface of the body scanner may include an input device such as a keypad, a touch sensitive screen, a pointing device and/or a voice recognition module for enabling a human operator to specify gender information for the individual being screened. Separately or in combination with the above features, in yet other alternative implementations, the method comprises deriving the gender information associated with the individual at least in part by using the computing device to process the body image data.

In accordance with such other alternative examples of implementation, the method comprises processing the body image data to derive geometric information associated with the individual under inspection and processing the geometric information to derive the gender information associated with the individual.

The geometric information may convey, for example, upper body shape information associated with the individual. The gender information may be derived at least in part by considering a chest-to-waist ratio and/or a hip-to-waist ratio derived at least in part based on the upper body shape information.

Alternatively or in addition to the above, the geometric information may be processed to derive information associated with a groin area of the individual under inspection. The gender may be derived at least in part by considering the information associated with the groin area. The information associated with the groin area may convey, for example, geometric information associated with the groin area and/or reflectivity information associated with the groin area.

Alternatively or in addition to the above, the geometric information may be processed to detect a presence of a gender specific item and the gender information may be derived at least in part based on whether the presence of the gender specific item was detected. Non-limiting examples of gender specific item include, without being limited to, gender specific articles of clothing (e.g. undergarment, brassiere, types of shoes, jewelry, skirt, dress, necktie, etc..) and gender specific hygiene products (sanitary napkins etc...). In specific examples implementation, the body scanner used to derive the body image data subjects the individual to millimeter waves when performing the inspection and the body image data conveys reflected radiation information. The body image data may convey, amongst others:

- radiation reflection intensity information, which provides information dependent on the type of surface reflecting incident millimeter wavelength radiation; and

- reflection depth information, which provides information on the depth at which the reflection occurred. Reflection depth information may be useful, for example, in providing information on the geometry of the body of the individual being screened.

In specific implementations, the body scanner used to derive the body image data includes an enclosure, in which the individual to be screened is positioned during the inspection, and a source of millimeter wave radiation configured for subjecting the individual to radiation to generate the body image data. The individual being screened is instructed to enter the enclosure and to assume a predetermined body position while the inspection is being performed. In a non-limiting example, the predetermined body position involves the individual standing with his/her legs slightly parted and both arms being raised over the head on either side of the body.

In a specific example implementation, used separately or in combination with the above features, the step of processing the body image data with the computing device to identify potential anomalies includes:

a) processing the body image data to identify a feature of interest on the individual, the feature of interest being a candidate for conveying an anomaly; b) associating the identified feature of interest with a body region; c) determining if the identified feature of interest conveys a potential anomaly at least in part based on the associated body region and on the gender information associated with the individual.

The body region may be selected from a set of possible regions of the body and may convey, for example, a groin region, a head region, limb regions (arms/legs), back and stomach. The body region may also convey an extremity, such as a hand or a foot, for example.

In accordance with this specific example of implementation, the method may identify a potential anomaly in the identified feature of interest at least in part using the computing device to perform a comparison between:

- observed information derived by processing at least a portion the body image data corresponding to the body region associated with the identified feature of interest; and

expected information derived at least in part based on body region associated with the identified feature of interest and the gender information associated with the individual. In a first non-limiting example, the computing device is used to perform a comparison between:

- observed proportions of the body region associated with the identified feature of interest, the observed proportions being derived by processing the body image data; and

- expected proportions of the body region associated with the identified feature of interest, the expected proportions being derived at least in part based on the gender information associated with the individual. A potential anomaly is identified when the observed body proportions differ from the expected body proportions by an amount exceeding a threshold amount. In a second non-limiting example, the computing device is used to perform a comparison between:

an observed geometry of the body region associated with the identified feature of interest, the observed geometry being derived by processing the body image data; and

an expected geometry of the body region associated with the identified feature of interest, the expected geometry being derived at least in part based on the gender information associated with the individual. A potential anomaly is identified when the observed body geometry differs from the expected body geometry by an amount exceeding a threshold amount.

In a third non-limiting example, the computing device is used to perform a comparison between:

observed items worn on the body region associated with the identified feature of interest, the observed items worn being derived by processing the body image data; and

expected items worn on the body region associated with the identified feature of interest, the expected items worm being derived at least in part based on the gender information associated with the individual. Examples of expected items worn may include, without being limited to, shoe types and undergarment types. A potential anomaly is identified when the observed items worn are inconsistent with the expected items worn. In specific examples implementation, the data released is configured for causing an image of the individual being screened to be rendered on a display device, the rendered image conveying at least some of the identified potential anomalies. For example, the rendered image may convey location information associated with the identified potential anomalies by highlighting or otherwise drawing attention to portions of the rendered image where the potential anomalies exist.

In accordance with another aspect, the invention provides an apparatus for performing security screening at a security checkpoint on an individual. The apparatus comprises a processing element having an input in communication with a body scanning device for receiving body image data derived by performing an inspection of the individual. The processing element is programmed for identifying potential anomalies at least in part based on gender information associated with the individual in accordance with the above described method. The apparatus also comprises an output for releasing data conveying results obtained by the processing element.

In accordance with another aspect, the invention provides a computer readable storage medium storing a program element for execution by a computing device. The program element, when executed by the computing device, causes the execution by the computing device of a method of the type described above for performing security screening at a security checkpoint on an individual, the method identifying potential anomalies at least in part based on gender information associated with the individual. In accordance with yet another aspect, the invention provides a system for performing security screening at a security checkpoint on an individual. The system comprises a body scanning device for performing an inspection of the individual to generate body image data. The system also comprises a processing element having an input in communication with the body scanning device and being programmed for identifying potential anomalies at least in part based on gender information associated with the individual being screened in accordance with the above described method. The system also comprises a display device in communication with the processing element for visually conveying results obtained by the processing element. In accordance with yet another aspect, the invention provides a method for performing security screening at a security checkpoint on an individual. The method comprises receiving body image data derived by performing an inspection of the individual using a body scanning device. The method also comprises processing the body image data with a computing device programmed with software to identify potential anomalies at least in part by processing the body image data to identify a feature of interest on the individual and determining if the identified feature of interest conveys a potential anomaly at least in part based on a body region associated with the identified feature of interest. The method also comprises releasing data conveying at least some of the identified potential anomalies.

Advantageously, the method provided allows taking into account a location on the body of the individual being screened when attempting to detect an anomaly in a body image. In accordance with a specific implementation, the body image data is processed to associate portions of the body image data with respective body regions. The body region may be selected from a set of possible regions of the body and may convey, for example, a groin region, a head region, limb regions (arms/legs), back and stomach. The body region may also convey an extremity, such as a hand or a foot, for example.

Any suitable manner of associating body regions to portions of the body image data may be used, m a non-limiting example, the respective body regions are associated with portions of the body image data at least in part by processing the body image data to identify one or more anchor points. Examples of anchor points on a human body include, without being limited to, the head and joint areas (e.g. elbows, waist, hips, knees, ankles, wrists, shoulders and the likes). In accordance with this specific example of implementation, the method may identify a potential anomaly in the identified feature of interest at least in part using the computing device to perform a comparison between:

- observed information derived by processing at least a portion the body image data corresponding to the body region associated with the identified feature of interest; and

- expected information derived at least in part based on body region associated with the identified feature of interest. In a first non-limiting example, the computing device is used to perform a comparison between:

observed proportions of the body region associated with the identified feature of interest, the observed proportions being derived by processing the body image data; and

- expected proportions of the body region associated with the identified feature of interest. A potential anomaly is identified when the observed body proportions differ from the expected body proportions by an amount exceeding a threshold amount. In a second non-limiting example, the computing device is used to perform a comparison between:

an observed geometry of the body region associated with the identified feature of interest, the observed geometry being derived by processing the body image data; and

- an expected geometry of the body region associated with the identified feature of interest. A potential anomaly is identified when the observed body geometry differs from the expected body geometry by an amount exceeding a threshold amount. In a third non-limiting example, the computing device is used to perform a comparison between:

observed items worn on the body region associated with the identified feature of interest, the observed items worn being derived by processing the body image data; and

expected items worn on the body region associated with the identified feature of interest. A potential anomaly is identified when the observed items worn are inconsistent with the expected items worn. For example, a belt around the waist may be considered "normal" since it is expected in the "waist" body region while a belt around one thigh or on an ankle underneath pants may raise some questions and be considered to be an anomaly since it is inconsistent with what would be "expected" in the thigh or ankle body regions. In specific examples implementation, the data released is configured for causing an image of the individual being screened to be rendered on a display device, the rendered image conveying detected anomalies in relation to the respective body regions.

In specific examples implementation, the body scanner used to derive the body image data subjects the individual to millimeter waves when performing the inspection and the body image data conveys reflected radiation information. The body image data may convey, amongst others:

- radiation reflection intensity information, which provides information dependent on the type of surface reflecting incident millimeter wavelength radiation; and - reflection depth information, which provides information on the depth at which the reflection occurred. Reflection depth information may be useful, for example, in providing information on the geometry of the body of the individual being screened. In specific implementations, the body scanner includes an enclosure, in which the individual to be screened is positioned during the inspection, and a source of millimeter wave radiation configured for subjecting the individual to radiation to generate the body image data. The individual being screened is instructed to enter the enclosure and to assume a predetermined body position while the inspection is being performed. In a non- limiting example, the predetermined body position involved the individual to stand with his/her legs slightly parted and both arms being raised over the head on either side of the body. In specific examples implementation, the data released is configured for causing an image of the individual being screened to be rendered on a display device, the rendered image conveying at least some of the identified potential anomalies. For example, the rendered image may convey location information associated with the identified potential anomalies by highlighting or otherwise drawing attention to portions of the rendered image where the potential anomalies exist.

In accordance with another aspect, the invention provides an apparatus for performing security screening at a security checkpoint on an individual. The apparatus comprises a processing element having an input in communication with a body scanning device for receiving body image data derived by performing an inspection of the individual. The processing element is programmed to identify potential anomalies in accordance with the above described method. The apparatus also comprises an output for releasing data conveying results obtained by the processing element. In accordance with another aspect, the invention provides a computer readable storage medium storing a program element for execution by a computing device. The program element, when executed by the computing device, causes the execution by the computing device of a method of the type described above for performing security screening at a security checkpoint on an individual. In accordance with yet another aspect, the invention provides a system for performing security screening at a security checkpoint on an individual. The system comprises a body scanning device for performing an inspection of the individual to generate body image data. The system also comprises a processing element having an input in communication with the body scanning device and being programmed for identifying potential anomalies in the body image data in accordance with the above described method. The system also comprises a display device in communication with the processing element for visually conveying results obtained by the processing element.

Other aspects and features of the present invention will become apparent to those ordinarily skilled in the art upon review of the following description of specific embodiments of the invention in conjunction with the accompanying Figures. BRIEF DESCRIPTION OF THE DRAWINGS

A detailed description of examples of implementation of the present invention is provided hereinbelow with reference to the following drawings, in which: Figure 1 shows a system for performing security screening on an individual at a security checkpoint in accordance with a non-limiting example of implementation of the invention.

Figure 2 shows a process suitable to be implemented by the system shown in Figure 1 for performing security screening on an individual according to a non-limiting example of implementation of the present invention;

Figure 3 is of a block diagram of a processing engine for use in the system depicted in Figure 1 in accordance with a non-limiting example of implementation of the invention; Figure 4 is a perspective view of a body scanner using millimeter wave electromagnetic radiation; Figure 5 is an example of a keypad suitable for use in connection with a body scanning device, such as the body scanner shows in Figure 4, for allowing a user to specify gender information associated with an individual being screened in accordance with a non- limiting example of the invention; Figure 6 shows a block diagram of an anomaly detection module suitable for use in the processing engine depicted in Figure 3 according to a non-limiting example of implementation of the present invention;

Figure 7 is a process for deriving gender information based in body image data in accordance with a non- limiting example of implementation of the present invention;

Figure 8 illustrates five (5) examples of body types;

Figure 9 shows renderings of a crotch area of an individual derived at least in part using data generated by a body scanner using millimeter-wave radiation;

Figure 10 shows frontal and back views of an individual derived at least in part using data generated by a body scanner using millimeter- wave radiation; Figure 11 is a frontal view an individual showing several anomalies appearing as bulges at several locations along the legs and thighs of the individual, the frontal view being derived at least in part using data generated by a body scanner using millimeter-wave radiation; Figure 12 shows the frontal view the individual depicted in Figure 11 in which the anomalies have been highlighted;

Figure 13 shows an example of a body segmentation reference template using a head height as a basis of measurement;

Figure 14 depicts a hierarchical representation of contextual information for the human body in accordance with a non-limiting example of implementation of a context database suitable for use in connection with the processing engine depicted in Figure 3;

Figure 15 depicts an overview of contextual information for a crotch area in accordance with a non-limiting example of a context database suitable for use in connection with the processing engine depicted in Figure 3;; Figure 16 is a block diagram of a computing apparatus suitable for use in connection with the processing engine illustrated in Figure 3 in accordance with a specific example of implementation of the invention; and

Figure 17 shows a functional block diagram of a distributed system suitable for performing security screening on an individual at a security checkpoint in accordance with an alternative example of implementation of the present invention.

In the drawings, embodiments of the invention are illustrated by way of example. It is to be expressly understood that the description and drawings are only for purposes of illustration and as an aid to understanding, and are not intended to be a definition of the limits of the invention. DETAILED DESCRIPTION

Figure 1 shows a system 100 for performing security screening at a security checkpoint on an individual in accordance with a specific example of implementation of the invention. The security checkpoint may be located at an access point to any secure location, for example at an airport, at a secure building facility or other. The system 100 depicted includes a body scanning device 102 for generating body image data 103 and a processing engine 104 for processing the body image data 103 to identify potential anomalies. In identifying potential anomalies, the processing engine 104 takes into account contextual information including, for example, gender information associated with the individual being screened and/or particular regions of the body of the individual where artifacts have been identified. The system 100 also includes a display device 106 for conveying results obtained by the processing engine 104.

Figure 2 depicts an example of a process for scanning an individual at a security checkpoint using the system 100 depicted in Figure 1.

At step 200, a body image of the individual screened is obtained at a security checkpoint. In the system depicted in Figure 1, this step is performed by using the body scanning device 102 (shown in Figure 1) to generate body image data associated with the individual. In a specific implementation, the screening is performed on a clothed individual and aims to determine if the individual is hiding under his or her clothes any potential threats such as concealed weapons, explosives or contraband material. Different image sources for generating body image data can be used. One example is to use image data generated by a full body scanner that uses electromagnetic radiation to irradiate the subject in the frequency range of about 200 MHz to about 1 THz. Another example of image data source is an X-ray machine that uses more penetrating radiation. Yet another example of image data source is a visible light camera which uses visible light with little or no clothes penetrating capabilities to produce an image of the person being scanned. Yet another example of image data source is an infrared camera which senses temperature patterns over the body of the subject. As a result of step 200, digital information, in the form of body image data, is produced by one or more image sources. Preferably, the digital information is stored on a storage device such that the data can subsequently processed with software.

At step 202, surface features of the subject are extracted from the body image data. The extraction process is software implemented and includes processing the image data to produce a virtual representation of the human body or portions thereof. When the image data source used to perform the scan relies on millimeter wave electromagnetic radiation the source of the topology information that is conveyed by the image data is obtained from the reflections of the electromagnetic radiation. More specifically, the interrogation source produces electromagnetic radiation that easily penetrates the clothing but is reflected back by the skin. By successively sending bursts of the electromagnetic radiation and measuring the time of flight of the reflections, it is possible to generate a three dimensional map of the body surface. As a result of step 202 a reconstructed image of the body is generated and released for further processing. At step 204, the reconstructed image of the body is processes by software to automatically determine the presence of anomalies, which may be indications of security threats, such as concealed weapons, explosives or contraband material. The detection of anomalies made at step 204 relies in part on contextual information 206, such as gender information and/or body region information. The processing performed at step 204 will be described in greater detail later.

At step 208, a signal conveying anomalies detected at step 204 is released. The signal may convey whether or not anomalies were detected and, optionally, when anomalies have been detected, the location on the body of the individual where such anomalies were detected. The system then proceeds to step 210.

At step 210, if no anomalies were detected, the system proceeds to step 214. If anomalies were detected, the system proceeds to step 212.

At step 212, which is initiated when potential anomalies were detected, the potential anomalies are conveyed to a security operator. The manner in which such anomalies are conveyed may vary between implementations. For example, the potential anomalies may be conveyed by generating a signal, that conveys visually, audibly or both that an anomaly has been identified. In a specific example, the identified anomalies are conveyed by rendering an image of the individual being screened on a display device, such as display device 106 (shown in Figure 1), where the rendered image indicates the existence of potential anomalies and the location of those anomalies on the human body. An example of image that would be rendered on the monitor is shown in Figure 12. The image provides a rendering of the body features assembled in a complete image and also shows locations of anomalies indicative of possible security threats. The anomalies shown in the image are highlighted as at 22. Any suitable highlighting method can be used such as adding color to the anomalies, overlaying highly visible geometric shapes (such as the circles shown in the drawing) or any other suitable image treatment technique.

At step 214, a security operator determines, at least in part based on information provided by the processing engine 104, whether the individual under inspection can be released or whether the individual should be subjected to further inspection (secondary screening) to more clearly establish his/her security status. The secondary screening may involve another automated screening by a different method or a manual screening such as by the "pat-down" method. The different components of the system 100 (shown in Figure 1) for performing security screening at a security checkpoint on an individual will now be described in greater detail below. Body scanning device 102

The body scanning device 102 performs an inspection of the individual to generate body image data. The inspection is performed on a clothed individual and the body image data may convey information on the whole body of the individual or on a portion of the body of the individual.

Different types of body scanning devices can be used to generate body image data in different implementations of the invention. One example is to use image data generated by a full body scanner of the type shown in Figure 4 that uses electromagnetic radiation to irradiate the individual in the frequency range of about 200 MHz to about 1 THz. At this frequency range, the electromagnetic radiation tends to easily penetrate clothing but is reflected by the skin. An example of the operation of a full body scanner of this type can be found in U.S. Patent 6,507,309, entitled "Interrogation of an object for dimensional and topographical information", issued on January 14, 2003. The contents of the aforementioned document are incorporated herein by reference. Typically, such full body scanners include an enclosure (shown in Figure 4) in which the individual to be screened is positioned during the inspection and a source of millimeter wave radiation configured for subjecting the individual to radiation to generate the body image data. In use, an individual to be screened is typically instructed to enter the enclosure, to position himself/herself at a predetermined location (generally at the center of the enclosure) and to assume a predetermined body position while the inspection is performed. In a non-limiting example, the predetermined body position involves the individual standing with his/her legs slightly parted and both arms being raised over the head on either side of the body. The body image data generated by subjecting the individual to millimeter waves conveys reflected radiation information, including:

Radiation reflection intensity information, which provides information dependent on the type of surface reflecting incident millimeter wavelength radiation. For example, the reflectivity of skin is somewhat uniform and can be statistically characterized on a per-body region basis. Since water is highly reflective, moist body regions (underarms for example) can have a rather different reflectivity. Contextual information related to the body region can be used to decide whether or not an observed reflectivity is normal or abnormal for a certain area of the body. It is also worth noting that some threats will absorb and look darker (low reflectivity) than skin, while others, for example metallic objects such as knives and guns, will be highly reflective and look lighter than skin; and

- Reflection depth information, which provides information on the depth at which the reflection occurred. Reflection depth information may be useful, for example, in providing information on the geometry of the body of the individual being screened.

Another example of a body scanning device is an X-ray machine that uses X-rays as the source of penetrating radiation. In practice, machines that use X-rays for screening people are seldom used in security applications since X-rays are absorbed by human tissue and may be hazardous from a health perspective when exposures are frequent.

Another example of a body scanning device is a visible light camera which uses visible light with little or no clothes penetrating capabilities to produce an image of the individual being scanned. In non-limiting examples of implementation, this visible light camera may be used in conjunction with a full body scanner (of the type shown in Figure 4). Yet another example of a body scanning device is an infrared camera which senses temperature patterns over the body of the individual being screened. In non-limiting examples of implementation, this infrared camera may be used in conjunction with a full body scanner (of the type shown in Figure 4).

The body scanning device 102 generates body image data 103 which is then released to processing engine 104.

Optionally, the body scanning device 102 may include an interface allowing a human operator to specify gender information conveying whether the individual being screened is more likely to be a male or is more likely to be a female. The interface may be operator by a security screener or, alternatively, by the individual being screened or by another individual. The interface may include any suitable type of input device such as, but without being limited to, a keypad, a touch sensitive screen, a pointing device and a voice recognition module. Figure 5 shows a non-limiting example of a keypad 500 for allowing an operator to specify gender information associated with an individual being screened. The keypad 500 depicted includes two key namely a first key 502, which when selected by an operator specifies that the individual being screened is most likely to be a male, and a second key 504, which when selected by an operator specifies that the individual being screened is most likely to be a female.

It is to be appreciated that, although an interface allowing a human operator to specify gender information has been described as part of the body scanning device 102 (shown in Figure 1), such interface may alternatively be a separate device and may be in direct communication with processing engine 104. Displav device 106

The display device 106 is in communication with the processing engine 104 and receives data conveying anomalies detected in the body image data 103 generated by the body scanning device.

The display device may, for example, render an image of the individual being screened, where the rendered image conveys at least some of the potential anomalies identified by the processing engine 104. Alternatively, or in addition to the display device 106, the system may include an output device (not shown) providing an indication of whether the individual being screened should be subjected to further inspection. Such output device may be in the form of a speaker to provide an audio signal, one or more light signals (e.g. green light = OK; red light = proceed to secondary screening) or any other device suitable for conveying information related to the results obtained by the processing engine 104.

In specific implementations, the data received by the display device may be configured for causing the display device 106 to display an image of the individual being screened. An example of such rendering is shown in Figure 10. The image illustrates two views of the same body, one being from the front and one from the back. The rendering of the surface features appears to the eye as a removal of the outer clothes layer, leaving exposed skin and body features.

In specific implementations, the image rendered may convey at least some of the potential anomalies identified by the processing engine 104 (shown in Figure 1). The image indicates the existence of possible anomalies and the location of those anomalies on the human body. An example of image that would be rendered on the display device is shown in Figure 12. The image provides a rendering of the body features assembled in a complete image and also shows locations of anomalies indicative of possible security threats. The anomalies shown in the image are highlighted as at 22. Any suitable highlighting method can be used such as adding color to the anomalies, overlaying highly visible geometric figures (such as the circles shown in the drawing) or any other suitable image treatment technique.

Optionally, the images rendered may include information conveying the likely gender of the individual being screened. This may be effected for example by displaying an icon on the display device representative of a man or a woman next to the rendered body image. Such information may provide additional information to the security operator in visually assessing the image of the individual being screened. This is particularly useful when the operator is located remotely from the scanning device and does not have the individual under inspection in his line of sight.

In specific implementations, the display device 106 may include a display screen that is part of an operator station, which may be located in proximity to, or remotely from, the body scanning device. Alternatively, the display screen may be part of a portable device, including for example a smart-phone or other portable device that may be carried by a security operator. Processing engine 104

The processing engine 104 is configured to receive and process body image data 103 derived by subjecting an individual to radiation in order to identify potential anomalies conveyed by the body image data 103. In identifying potential anomalies, the processing engine 104 is configured to take into account contextual information including, for example, gender information associated with the individual being screened and/or particular regions of the body of the individual. The processing engine 104 will now be described in greater detail with reference to Figure 3.

As depicted, the processing engine 104 includes an input 300 for receiving body image data 103 derived by the body scanning device 102. The body image data 103 may originate from a single image source, for example from a millimeter wave full body imager, or alternatively the body image data may originate from multiple image sources, for example a millimeter wave full body imager, a visible image camera and an infra-red camera. The body image data 103 received may be stored in a memory unit (not shown in the Figure) that is part of, or associated with, processing engine 104.

Optionally, the input 300 can also receive data conveying gender information associated with the individual being screened, the gender information conveying whether the individual being screened is more likely to be a male or a female. The gender information may originate from an interface associated with the body scanning device (for example an interface of the type depicted in Figure 5) or alternatively may originate from another suitable system configured for generating information conveying the likely gender of the individual under inspection. The processing engine 104 also includes a memory unit 306 for storing a context database 306 and functional components namely an image reconstruction and information extraction module 302, an anomaly detector module 304 and an image rendering module 308. The processing engine 104 also includes an output 310 in communication with the image rendering module 308 for releasing data configured for causing an image of the individual being screened to be rendered on display device 106 (shown in Figure 1).

The image reconstruction and information extraction module 302, the anomaly detector module 304 and the image rendering module 308 are described in greater detail below. Image reconstruction and information extraction module 302

The image reconstruction and information extraction module 302 is in communication with the input 300 and is configured to reconstruct an image (which may be a 3D or 2D image) of the individual being screened.

In a specific implementation, the reconstruction/extraction process is software implemented and includes processing of the body image data to produce a virtual representation of the human body or portions thereof. Any suitable method for processing image data to reconstruct an image may be used.

When the body scanning used to perform the scan relies on millimeter wave radiation, the source of the topology information that is conveyed by the image data is obtained from the reflections of the electromagnetic radiation. More specifically, the interrogation source produces electromagnetic radiation that easily penetrates the clothing but is reflected back by the skin. By successively sending bursts of the electromagnetic radiation and measuring the time of flight of the reflections, it is possible to generate a (three dimensional map) of the body surface. The module 302 receives a multitude of data points in the body image data 103 which represent topographic information of different points of the body surface. The module 302 then combines these data points in a continuous map that provides a three-dimensional representation of the skin surface. In specific implementations, a wideband cylindrical imager may be used in order to perform image reconstruction. An implementation of a wideband cylindrical imager is described in "Effective Reconstruction Approaches to Millimeter- Wave Imaging of Humans", by J. Detlefsen, A. et al., of the Technische Universitat MUnchen. This document was presented at the following conference: VHIth URSI General Assembly, New Delhi, India; October 23rd to 29th, 2005. It is available at the following link http://ursiweb.intec.ugent.be/ ^/ India05/GA05index.htm. The contents of the above documents are incorporated herein by reference. Another example of implementation of a wideband cylindrical imager is described in U.S. Patent No. 5,859,609 entitled "Wideband cylindrical Holographic Surveillance System", by Sheen et al. issued on January 12, 1999. The contents of the above documents are incorporated herein by reference. The reconstructed image data is released by the image reconstruction and information extraction module 302 and provided as an input to the image rendering module 308 and the anomaly detector 304.

Anomaly detector 304

The anomaly detector 304 (shown in Figure 3) processes the reconstructed image data released by the image reconstruction and information extraction module 302 in order to identify potential anomalies in the image. The identified anomalies may be indications of security threats, such as concealed weapons, explosives or contraband material.

As is explained further below, to identify potential anomalies, the anomaly detector 304 analyses the reconstructed image data image to identify certain features, which may be any identifiable aspect of the image data or associated image, such as the shape, configuration, arrangement, texture, location of one or more objects relative to a person's body, or features of the person's body, such as orientation, position, texture, specific body parts, size, shape, configuration, symmetry, or other appropriate aspect. In accordance with a preferred embodiment, the anomaly detector 304 in particular considers contextual information, including amongst others gender information and/or body location specific information, when identifying a potential anomaly.

In accordance with a specific example, the anomaly detector 304 makes use of contextual information to estimate what should be expected of the body image of the person being scanned. If the observed ("real") body image generally matches what is expected, the anomaly detector 304 infers that no anomalies exist. However, if the observed ("real") body image includes features that are not expected to be there, given the context, those features may be indicative of a security threat such as a concealed weapon or explosive. The contextual information on the basis of which the anomaly detector 304 produces the expected (reference) information can be derived from different sources.

One source is the reconstructed body image data received from module 302 (shown in Figure 3). In a specific example, this reconstructed body image data conveys three dimensional information of the skin surface (topology information) derived from body image data generated by a full body scanner. In such case, the reconstructed body image is the observed ("real") body image, which can be analyzed to extract a number of parameters that can be used to generate expected (reference) topology information. For instance, the observed ("real") body image can be analyzed to extract (1) dimension information such as the height of the person, his/her shoulder size, his/her waist size, crotch height and thigh size and (2) the gender of the individual being screened. On the basis of those parameters a standardized body outline can be generated and used for building the expected (reference) topology. In one possible example of implementation, the parameters extracted from the observed ("real") body image are communicated to a context database that maps parameters with corresponding expected body outlines. In this example, the output of the database conveys expected (reference) topology information providing a general indication of what the image of a person having those parameters should look like. Therefore, the expected (reference) topology information provides a basis against which the observed ("real") body image can be compared to determine the presence of anomalies indicating possible security threats.

Another possible source of contextual information is an image of the individual produced by a scan using a different spectral source such as a visible light camera. The visible light information picks up information on facial features, clothing features that are unlikely to be fully visible in the real topology information generated as a result of the scan. For example the visible light camera can convey the location on large shirt or coat buttons, belt buckles or other features that may be large enough or made of denser material such that they will continue to show to at least some extent in the real topology information. It other words they will not be fully erased. This contextual knowledge can then be used to further customize the expected (reference) topology information, by integrating them to the expected (reference) topology information to avoid that they appear as anomalies suggesting a security threat. The integration process may involve extracting the signature of those objects in the real topology information and copying it in the corresponding location on the expected (reference) topology information. In another example, a facial image of the individual taken with a visible light camera and may be used to extract information such as for example gender information by processing facial features.

Yet another source of context information is an X-ray scan or an infrared camera that senses temperature patterns over the body of the person being scanned. The temperature patterns may indicate hot or cold spots that may suggest the presence of material covering the skin. For example a thin layer of foil covering an area of the body is not likely to have the same temperature as the remainder of the skin surface. The layer of foil will be picked up by the infrared camera on the basis of the temperature difference.

Yet another source of context information is a manual input made by a user. For example, as described above, the body scanning device 102 (shown in Figure 1) may include an interface allowing a human operator to specify gender information conveying whether the individual being screened is more likely to be a male or is more likely to be a female. The interface may include any suitable type of input device such as, but without being limited to, a keypad, a touch sensitive screen, a pointing device and a voice recognition module.

Yet another source of context information may be provided through another type of input. For example by providing the individual's passport information, driver's license or other document, gender information as well as other information related to the individual (e.g. age and height) can be obtained. It is to be appreciated that the above examples of sources of contextual information may be used independently from one another or in various combinations. In addition certain ones may be omitted from some implementations and other types of contextual 5 information may be added depending on the particular application and/or circumstances.

Figure 6 shows a functional block diagram of an example of an anomaly detector module 304.

10 As depicted, the anomaly detector 304 includes an input 601 for receiving a reconstructed body image from the image reconstruction and information extraction module 302. The anomaly detector also includes functional modules including a gender information extraction module 600, a body part segmentation module 602, a feature detection module 604 and an anomaly assessment module 606. The functional modules 600 602 604 and

15 606 make use of information stored in a context database 306 (shown in Figure 3), which defines expected parameters on the basis of contextual information. The context database 306 will be described later on in the present document. The anomaly detector 304 also includes an output 608 for releasing data releasing data related to potential anomalies that were identified. The data may convey, for example, whether anomalies where detected

20 and location information related to the anomalies identified. Optionally, the data may also provide information conveying the type of anomalies that were identified. For example, the data may convey whether the anomaly is likely to be composed of metal, of plastic or of another material. In another example, the data may convey information as to whether the anomaly is likely to constitute a threat. The data may also convey 5 information related to the gender of the individual being screened as derived by the gender information extraction module 600. It will be readily apparent to the person skilled in the art in light of the present description that additional information related to potential anomalies identified may be released at the output 608. The gender information extraction module 600 may be configured to derive gender information associated with the individual under inspection at least in part by processing the reconstructed image derived from the body image data 103 by the image reconstruction and image extraction module 302 (shown in Figure 3). Different approaches can be taken by the gender information extraction module 600 to derive the likely gender of the individual being screened. Specific example of such approaches will be described later on in the present document. The gender information extraction module 600 releases an indicator conveying whether the individual is likely to be male or female. The indicator may simply indicate "Male" or "Female" or alternatively the indicator may provide a likelihood value that the individual is male or female. For example, the indicator may convey that the individual has an 80% chance of being male. This gender indicator is provided to the body part segmentation module 602, to the feature detection module 604 and to the anomaly assessment module 606. It will be appreciated that in certain alternative implementations (not shown in the Figures), the gender information extraction module 600 may be omitted and the gender indicator may be provided by a device external to the anomaly detector module 304.

The body part segmentation module 602 may be configured for associating portions of the body image data with respective body regions at least in part by processing the reconstructed image derived from the body image data 103 by the image reconstruction and image extraction module 302 (shown in Figure 3). Different approaches can be taken by the body part segmentation module 602 to segment the body image data into body regions, some of which may optionally make use of the gender information derived by the gender information extraction module 600. Specific example of such approaches will be described later on in the present document. The body part segmentation module 602 releases information associating portions of the body image data with respective body regions. The particular body regions with which portions of the body image data may be associated may differ between specific implementations. In a specific example of implementation, the body of an individual is divided into the following body regions: - head region

(2) arm regions

Chest region

- Waist and groin region

- (2) leg regions

(2) foot regions

It is to be appreciated that variants of the inventions may include fewer than or more than the above body regions. It is also to be appreciated that a body region may overlap with other body regions. The information associating portions of the body image data with respective body regions is provided to the feature detection module 604 and to the anomaly assessment module 606.

The feature detection module 604 may be configured for processing the reconstructed image derived from the body image data 103 to identify features of interest on the body of the individual, which could potential signal the presence of an anomaly. In a specific example of implementation, the processing performed by the feature detection module is performed on a per body region basis based on the segmentation performed by the body part segmentation module 602. What constitutes a "feature of interest" is built into the software logic of the feature detection module 602. The features of interest that are being searched are likely to be different depending on the likely gender of the individual being screened and/or depending on the body region being examined and thus information derived by the body part segmentation module 602 and the gender information extraction module 600 may be used. In a specific implementation, the features of interest that are being searched are specified by parameters found in context database 306 (shown in Figure 3).

The detection of the features of interest may also be dependent on the type of body image data generated by the body scanning device 102 (shown in Figure 1). Image processing is performed to detect the features of interest using known image deconstruction and feature extraction techniques to identify the presence of a feature of interest (a missing element is considered as a feature of interest in the image).

In the case where the body image data is derived using electromagnetic radiation in the frequency range of about 200 MHz to about 1 THz, which reveals features under the clothes, examples of features of interest may be:

1. Pixel formations indicating the presence of a mass where no such mass is expected to exist on a normal human body. For instance, the system may be designed to search for such pixel formations in regions of the human body where a threatening object is likely to be hidden, such as the crotch region, underarms, the belt and feet (the shoes). If a pixel formation is identified which is of a shape and/or size that is not expected in a normal human body on that region of the human body, it is classified by the processing as a "feature of interest".

2. Geometric shapes that do not occur normally in a human body.

The system may search for certain atypical shapes or shape fragments suggesting an extraneous object. The human body normally has a gently curvilinear outline and any sharp edged objects or objects having an outline that made up of segments that are straight or of certain geometric form, such as circles, triangles, rectangles, etc are considered suspect and classified as "features of interest". In addition, the shape of the human body varies based on a gender. Thus an individual classified as female by the gender information extraction module 600 but for which an atypical shape appears in the groin area, wherein the shape could be considered normal for a male, would be considered suspect and classified as "features of interest".

In the case of a visible source examples of features of interest may be:

1. Clothing features which may register in some confusing way on a scan performed with another source. Examples include jewelry, such a large necklace, watches, heavy bracelets, large belt buckles, heavy shoes or shoe features (large buckles or straps), or heavy clothing that may not be fully erased by the millimeter wave scan. Examples of heavy clothing include bulky pants, such as cargo pants and heavy sweaters or clothing with certain decorative features such as chains or buckles that the millimeter wave scan is unlikely to erase.

2. Distinguishing external body features, such as a large abdomen in the case of an overweight person or any other oversized body part. In this instance, it would be possible to co-relate the presence of the large body part into the other spectral scan (such as the scan using electromagnetic radiation in the frequency range of about 200 MHz to about 1 THz). If they do not match, in other words, no large body part shows where the first scan finds one, this suggest the presence of some padding material on the body that would require some further investigation.

3. The general outline of the clothed individual may be in itself a "feature of interest" since it can be matched against the image of the unclothed body obtained by a different scan to show if both match. If there is a significant disparity between the scans, that disparity may suggest the presence of extraneous material on the body suggesting further investigation.

In the case of an infrared source examples of features of interest may be: 1. Cold or hot spots on the person body that can be used to confirm the presence of certain objects (such as jewelry or a watch) that appear on another spectral scan such as to confirm what they are, namely normal objects unlikely to be threatening. The cold or hot spot information can be used also to identify the presence of hidden objects on the body by showing abnormal temperature patterns in areas of the body where no such pattern should normally be present. For instance if the infrared scan shows an area that is relatively colder in the thighs or legs of the person being scanned, that lower temperature area may suggest the presence of an extraneous object.

In the case of an X-ray source, an example of features of interest may include an X-ray absorption patterns showing high density objects on the person that are likely to be extraneous objects.

The processing of the body image data may include generating, for each feature of interest, location data to specify where that feature of interest is. In this fashion, it is possible to correlate the features of interest from different data source by determining if the location data matches.

The identification of the features of interest may be made by performing an image analysis. In the case of the millimeter wave scan, known image analysis techniques are used to identify the pixel formations which are atypical. For instance a bulge having a surface area above a certain threshold and located along one of the arms or legs is considered as a "feature of interest". Also, certain edge shapes or edge patterns in the image that are atypical for an unclothed human body are classified as "features of interest". Those edge or edge patterns are also identified in the image by performing image processing with known algorithms. Similar techniques may be used for identifying the features of interest in the scans using different types of body scanning device using different spectral sources.

Optionally, the feature detection module 604 may also apply a variety of filters to the reconstructed body image. Some or all of the following filters may be applied to obtain a filtered version of the reconstructed body image:

• Match filter identify spot-lit parts (pixels similar to each other)

• Symmetry checker (symmetry in the context of a human body)

• 3D cavity search engine (for instance we expect cavities in certain areas and not others, unexpected cavities will stand out)

• Contextual edge highlighter

It is also to be appreciated that, in the example depicted in Figure 6, the reconstructed body image is segmented into body regions by the body part segmentation module 602 prior to the detection of features of interest by the feature detection module 604. It is to be appreciated that, in alternative examples, these two functions may be reversed. For instance, the detection of features of interest may be performed on the reconstructed body image as a whole rather than on a "per body region" and the segmentation into body region may be performed after the features of interest have been identified. In such alternative, once the features of interest have been identified, they may be associated with the body regions (region segmentation) they correspond to.

In a specific example, the output of the feature detection module 604 may include features of interest data that specifies:

1. whether features of interest have been identified (based on data originating from one type of body scanning device or several types of body scanning devices);

2. if features of interest have been identified, any characterization information of the features of interest such as location information, size such as surface area, body region where it was located (close to arms, legs, etc), temperature (for infrared scans) among others.

The feature of interest data may then be provided as an input to the anomalies assessment module 606.

The anomalies assessment module 606 receives and examines the features of interest data derived by the feature detection module 604 and determines whether one or more of the features of interest should be identified as potential anomalies.

More particularly, the feature detection module 604 essentially performs a first pass to identify features of interest in a somewhat broad manner. The anomalies assessment module 606 then applies a filter to eliminate from the list of identified features of interest those that are unlikely to be anomalies and to constitute a threat. In performing this filtering, the anomalies assessment module 606 can makes use of contextual information, such as for example the likely gender of the individual and/or the body region in or near which a given feature of interest was identified, to determine whether the feature of interest should be identified as an anomaly.

The anomalies assessment module 606 performs a comparison between observed information, derived by processing the body image data, and expected information. The expected information conveys features that would be expected based on the context in which such features appears and is derived at least in part based on the gender information associated with the individual and/or the body region in or near which the feature appeared. What is "acceptable" in terms of context is defined by information stored in a context database 306 (shown in figure 3).

In a specific example of implementation, the anomalies assessment module 606 determines if a given feature of interest fits into an acceptable context given the gender information and/or the body region information. More specifically, within the groin body region for an individual that is male what are the things that could be confusing us? That is, what are we expecting to see in that particular body region and do the features of interest identified by the feature detection module 604 match these expectations? We may see a blob in a body region, however within this context, having a blob there isn't out of context. For instance, horizontal lines that are typical of a belt will be expected in the waist body region of the individual being screened. However the same type of horizontal lines located on the leg region may not be expected and may thus signal an anomaly.

In a specific example, the anomalies assessment module 606 identifies a potential anomaly by performing a comparison between observed information derived by processing the body image data and expected information derived at least in part based on the gender information associated with the individual. In accordance with this specific example of implementation, if the gender information conveys that the individual is more likely to be a male, the image data is processed to identify anomalies at least in part based on male contextual information. Conversely if the gender information conveys that the individual is more likely to be a female, the image data is processed to identify anomalies at least in part based on female contextual information.

In a first non-limiting example, the anomalies assessment module 606 performs a comparison between observed body proportions derived by processing the reconstructed body image and expected body proportions derived at least in part based on the gender information derived by the gender information extraction module 600. A potential anomaly is identified when the observed body proportions differ from the expected body proportions by an amount exceeding a threshold amount. The threshold amount may be selected using any suitable technique, which may be based on heuristic rules. The body proportions may be compared on a per body region basis and optionally may be limited to only certain areas of the reconstructed body image, for example the groin region and/or the torso body region. In a second non-limiting example, the anomalies assessment module 606 performs a comparison between an observed body geometry derived by processing the reconstructed body image and an expected body geometry derived at least in part based on the gender information derived by the gender information extraction module 600. A potential anomaly is identified when the observed body geometry differs from the expected body geometry by an amount exceeding a threshold amount. The body geometry may be compared on a per body region basis and may be limited to only certain areas of the body image. For example, if the gender information conveys that the individual is more likely to be a female, if the observed body geometry of the crotch area is not consistent (or differs from the expected body geometry by an amount exceeding a threshold amount) with the expected body geometry of the crotch area for a female, then an anomaly can be identified.

In a third non-limiting example, the anomalies assessment module 606 performs a comparison between observed items worn derived by processing the reconstructed body image and expected items worn derived by the gender information extraction module 600. Examples of expected items worn may include, without being limited to, shoe types and undergarment types. A potential anomaly is identified when the observed items worn are inconsistent with the expected items worn. The observed and expected items worn may be compared on a per body region basis and may be limited to only certain areas of the reconstructed body image.

In a fourth non-limiting example, the anomalies assessment module 606 performs a comparison between observed reflectivity information derived by processing the reconstructed body image and expected reflectivity information obtained from the context database 306 based on the gender information extraction module 600. As such, for the crotch body region, if the observed reflectivity information of the crotch area is not consistent (or differs from the expected reflectivity information by an amount exceeding a threshold amount) with the expected reflectivity information of a crotch area, then an anomaly can be identified in the crotch area.

The logic for determining if a feature of interest is an "anomaly", in the sense that it does not fit within a certain context and thus may signal a potential security threat, relies on rules such as:

1. The shape and dimension data for a certain feature of interest can be used to compute volume information which can be a trigger as to what is considered to be a security threat or not. For example, anomalies having a volume beyond a certain threshold can be considered to be likely security threats and are treated as such. Volumes below the threshold are dismissed;

2. Factors such as the location of the feature of interest, for example the body region on which it was identified, can also be used as indicators to determine whether the feature of interest should be classified as an anomaly. For example, a large feature of interest on the leg is likely to be a security threat since leg shapes in the general population do not vary much from one another. Another example is the case of "extraneous" items. Consider the image at Figure 11 which shows several anomalies appearing as bulges at several locations along the legs and thighs of the person. The bulges 62 are of a shape and size that is unlikely to be a normal component of the leg regions of a human body, hence they are classified as anomalies indicative of a security threat. Parameters used in determining whether a feature of interest should be classified as an anomaly based on the body region on or near which is was identified can be stored in a database, for example context database 306 (shown in Figure 3). Factors such as the gender of the individual under inspection can also be used as indicators to determine whether the feature of interest should be classified as an anomaly. For example, the presence of two oval elements near the lower half of the breast area may indicate the presence of a bra with push-up inserts in a female but would not be expected to be found on a male. Similarly, the presence of an elongated member in the groin area of the individual, depending on the size, may correspond to the male genitals when the gender of the individual is male while it would not be expected to be found in a female and thus would be potentially classified as an anomaly in a case where the likely gender of the individual is female. Thus bulges that are of a shape and size that is unlikely to be a normal (expected) component of a human body given the gender of the individual, may be classified as anomalies. Parameters used in determining whether a feature of interest should be classified as an anomaly based on the gender may be stored in a database, for example context database 306 (shown in Figure 3).

Curvature Analysis. This is particularly useful when the body image data is derived using radiation that penetrates clothing but that is reflected by the skin, such as for example millimeter- wavelength radiation. Depending on context (body region and/or gender) a given curvature variation across the body region would be expected. For example, a male chest would be expected to be somewhat flat. Some weight estimation could indicate to expect a more or less intense curvature variation. The direction of the variation may also important, i.e. mostly outwards (towards the observer) and slow. A powdery substance placed in a plastic bag taped to the chest region of a male would typically absorb and reflect incident millimeter-wavelength radiation differently than human tissue. The observable effect in the depth data is such situations is often that of a depression with rapid curvature variation on the sides which is quite unexpected on a male chest. When such effect is observed, this area may be tagged as an anomaly. Figure 9 of the drawings illustrates renderings of a crotch area of a male individual derived at least in part using data generated by a body scanner using millimeter-wave radiation. The male genitals are expected to be centered in the crotch and be somewhat elongated. A convex hull around the bump thus created should generally have the same area as the outline of the bump, otherwise, an anomaly may be identified in the crotch area. In Figure 9, the left image is normal while the right image indicates the presence of an anomaly in the crotch conveyed by an unexpected curvature in that area.

5. Reflectivity Analysis may also be useful when the body image data is derived using radiation that penetrates clothing but that is reflected by the skin, such as for example millimeter-wavelength radiation. The reflectivity of skin is generally somewhat uniform and can be (statistically) characterized on a per-body region basis. Since water is highly reflective, moist body regions (underarms for example) can have a rather different reflectivity. Some threats will absorb millimeter-wavelength radiation and thus have a darker appearance than skin in the image. Others, such as metallic objects (knives and guns), will be highly reflective. By using context base on body region, it can be determined whether observed reflectivity in a feature of interest is normal or abnormal. Therefore reflectivity information may provide an additional indication of whether a body region should be indentified as an anomaly

Correlation between features of interest obtained from different sources of body image data may also be useful. More sparticualrly, if features of interest have been identified in two or more of the scans, the anomalies assessment module 606 will first determine if the locations of the features of interest match. If they do, this may mean that the features of interest observed in two or more different scans likely relate to a same object or thing. Accordingly, the information about the features of interest seen in several scans generated by different body scanning devices are accretive, in other words they can be combined to better assess if a given feature of interest is an anomaly and should be considered to be a potential threat. For example, if a feature of interest in the scan generated with a visible source shows a large belt buckle, which is identified as such by image processing techniques, the residual signature of the belt buckle that may still be observed in the millimeter wave scan now considered to be unlikely a threat since it is much more likely a belt buckle. In this case the correlation between multiple features of interest negates false positives.

The anomalies assessment module 606 may also perform a shape recognition search on the basis of the shape of the feature of interest as seen in the image as well as material information if such information is available. The shape recognition search can be performed on images produced by the millimeter wave scan or scans from other spectral sources. Examples of objects that can be identified as threatening on the basis of their shape include guns or stabbing objects that have recognizable shapes. The anomalies assessment module 606 essentially performs an image analysis to determine if the features of interest identified by the feature detection module 604 have a shape corresponding to the shapes of objects in a reference database of threatening objects. The reference database of threatening objects stores characteristic shape information for a variety of possible threatening objects. The anomalies assessment module 606 extracts from the database the various shapes and processes the area of the reconstructed image where the feature of interest was found to determine the likelihood of presence of this shape on the person being scanned. Note that the database may store characteristic shapes information identifying the shapes from different perspectives. For example, in the case of a hand gun, the database may store generate data conveying what the handgun looks like from different angles, for example from the side, from the bottom, from the top or any other intermediate position.

Once the features of interest data derived by the feature detection module 604 have been processes and classified as either being normal/expected or identified as potential anomalies, the anomalies assessment module 606 releases to output 608 data related to the potential anomalies that were identified so that the data may be processed by the image rendering module 308 (shown in Figure 3). Context database 306

Specific implementations of the anomaly detector 304 (shown in figure 3) may make use of information stored in a context database 306 (shown in Figure 3). The context database 306 stores parameters pertaining to expected features of an image of a human body based on contextual information. The contextual information provided influences what one would expect to see in a body image. In the particular examples here, the contextual information includes gender information associated with the individual as well as body portion information. Other types of contextual information may include, without being limited to, the age of the individual, the ethnic origin (for example African American, Caucasian, Asian, Indian etc ..) or other. For the purpose of simplicity, only gender and body portions are considered in greater detail in the present application. The context database 306 may provide entries for each feature likely to be part of a body image including parts of the human body itself as well as items likely to be found on a human body (e.g. clothing, jewelry, accessories, shoes etc .).

Figure 14 depicts a hierarchical representation of contextual information for the human body in accordance with a non-limiting example. In the example shown, the centre portion of the body is the torso. All other portions of the body are linked from there.

Articulated anchor points may be identified (for example, the wrist, the knees, the ankles, the waist, the elbows etc .). Some features are mostly surface features which can be easily identified and can be used to position other elements in the scene (ex: the ear). Some items may vary between the genders and within the genders themselves (ex: breasts). Some elements are unique on the body, like the head; while others come in pairs, like the feet.

Figure 15 depicts a simplified representation of contextual information for the crotch area 1500 in accordance with a non-limiting example. Context information for other body regions may be configured using a similar top down (hierarchical) approach.

In the simplified representation depicted in Figure 15, the crotch area 1500 is characterized by a set of parameter defining expected surface curvature, which may be defined separately based on gender (male crotch curvature 1504; female crotch curvature 1506). High reflectivity objects are not expected to be found in the crotch area except maybe for belt metallic buckles, zippers and pants buttons. As such, the crotch area context information may specify parameters associated with likely high reflectivity items 1508, such a belt 1514 and a buckle 1516. Although not explicitly shown figure 15, the parameters of high reflectivity items may be further refined based on gender. In addition to the above, it may be expected that there will be lines in the crotch area that signal the presence of surface underwear. To enable detection of these lines, parameters pertaining to lines such as the orientation, the search area, the expected widths, the expected uniformity of reflectivity, the amount of acceptable completeness and the position. As such the crotch area may specify parameters associated with such lines 1540 such at to be able to distinguish whether those likes are likely to be associated with underwear 1512. Although not explicitly set out in the figure, the parameters related to underwear may be further refined based on gender since it would be expected that underwear for men and women would be different.

The context database 306 provides information pertaining to the boundaries of what would be considered "normal" in a given context. As such, given the crotch body portion and given a female gender, the expected curvature (including tolerance values) of that body portion would be provided by item 1506. This expected value may be compared against an observed curvature of that body portion (the crotch area) to determine whether what is being observed qualifies as an anomaly.

It will be appreciated by the person skilled in the art that the above simplified representation was presented for the purpose of illustration only and that practical implementations may include many more parameters for characterizing features of the crotch areas of a human body. In addition, it will also be appreciated many other configurations and contents for the context database 306 may be possible and that the example provided above was intended for the purpose of illustration only. Image rendering module 308 (Figure 3)

The image rendering module 308 receives a reconstructed image derived by the image reconstruction and information extraction module 302 and also receives information conveying anomalies detected by the anomaly detector 304.

The image rendering module 308 generates data for causing a display device, such as display device 106 (shown in Figure 1), to display an image of the individual being screened. The data generated may also cause anomalies detected by anomaly detector 304 to be conveyed to the user. For example, the data generated may cause the image to indicate the existence of possible anomalies and the location of those anomalies on the human body. An example of an image that would be rendered is shown in Figure 12. The image provides a rendering of the body features assembled in a complete image and also shows locations of anomalies indicative of possible security threats. The anomalies shown in the image are highlighted as at 22. The highlighting effect can be obtained by placing in the image a symbol such as a circle or arrow to visually draw the attention of the operator, by changing the color of image locally or by blurring the image around the feature of interest in order to make it more visually prominent. It is to be appreciated that any suitable highlighting method can be used such as adding color to the anomalies, overlaying highly visible geometric Figures (such as the circles shown in the drawing) or any other suitable image treatment technique.

The data generated by the image rendering module 308 is release at output 310 of the processing engine 104, which is in communication with display device 106. Segmenting a reconstructed body image into body regions

As mentioned above with reference to Figure 6, a body part segmentation module 602 may be provided for associating portions of the body image data with respective body regions.

In a specific example of implementation, the body part segmentation module 602 is part of the anomaly detector module 304 (shown in Figures 3 and 6) and is configured for associating portions of the body image data with respective body regions at least in part by processing the reconstructed image derived from the body image data 103 by the image reconstruction and image extraction module 302 (shown in Figure 3).

The particular body regions with which portions of the body image data may be associated may differ between specific implementations. In a specific example of implementation, the body of an individual is divided into the following body regions:

- head region

- (2) arm regions

- Chest region

- Waist and groin region

(2) leg regions

(2) foot regions

It is to be appreciated that other segmentations may be possible. For example, the arm regions may be further divided into upper and lower arm regions and hand regions. Similarly, the leg regions may be further divide into upper leg (thighs), lower leg and ankle regions. A way that this segmentation can be handled may be to segment the larger most visible parts of the body first. It will be easy enough to identify the head and segment that out, as well as the torso. The groin area may also be segmented. The arms and legs can also be segmented out, and within their segmentation we can segment from the joint for more thorough analysis for abnormalities. An objective in performing such segmentation is to facilitate the detection of distorted areas of the human body so that these can be analyzed in more detail. This would not only include the groin area, but joint areas, and potential zones of heavy fat deposits. In specific implementations, areas commonly used for hiding places, such as the back, the groin area and the legs, may be segmented from the body image to facilitate the detection of anomalies. This may not only be good for detecting devices and weapons but drugs that are hidden on the person. In a non-limiting implementation, the process for segmenting the reconstructed body image may begin by locating the head region. The system may then initiate the segmentation based on the initial assumption that the body of the individual is eight "heads" high. The first head height is obviously the starting point. That is the person's actual head. The next head will extend down to around the chest at the nipples, or there about. The third head down will be at the waistline around the bellybutton. The forth head will be the groin area, and the fifth head will be just above the knee. The sixth head length is going to be just below the knee area, while the seventh head will be above the ankle. The last head length will be the feet, this is be around seven and ¾ or the eighth head. When the arms are at the side of the person's body, the wrist bone aligns with the groin area. The elbow will align with the waistline, at or above the bellybutton. The shoulder width from side to side will be between 2 and 2 1/3 heads wide. Figure 13 illustrates the segmentation of a body using the head height as a basis of measurement. It is to be appreciated that this segmentation provides rough estimates of the body portions and may be particularly useful in performing a first pass for the segmentation. Following this, the segmentation may be refined, if needed, to provide a better segmentation using any suitable processing technique.

Optionally, anchor points may also be identified and used to segment the body image data into body regions. For example, the head, elbows, wrists, ankles, knees, hips and the belly button may be identified as anchor points.

It is to be appreciated that the segmentation of a body image into body regions may be facilitated in situations where the individual under inspection is in a pre-determined positioned. As is often the case, when using a full body scanner of the type depicted in Figure 4, the individual being screened is instructed to enter the enclosure and to assume a predetermined body position while the inspection is being performed. The predetermined body position typically involves the individual standing with his/her legs slightly parted and both arms being raised over the head on either side of the body.

U.S. patent no. 7,386,150 entitled "Active subject imaging with body identification", which was issued on June 10, 2008 to Fleisher, describes other manners in which a body image may be segmented into body regions. The contents of the aforementioned document are incorporated herein by reference.

It is to be appreciated that the body part segmentation module 602 may associate portions of the body image data with respective body regions using a number of different approaches, which may be used either independently or in combination with one another. The specific manner in which the body image data is segments into body regions is not critical to the invention and thus will not be described in greater detail here. Determining likely gender based on body image data

As mentioned above with reference to Figure 6, a gender information extraction module 600 may be provided for deriving gender information associated with the individual being screened based on body image data.

In a specific example of implementation, the gender information extraction module 600 is part of the anomaly detector module 304 (shown in Figure 3 and Figure 6) and is configured to derive the gender information at least in part by processing the reconstructed image derived from the body image data 103 by the image reconstruction and image extraction module 302 (shown in Figure 3).

Optionally, the gender information extraction module 600 may derive the gender information taking into account the manner in which portions of the body image data are associated with respective body regions. Under such an option, the gender information extraction module 600 may be in communication with the body part segmentation module 602 (described previously) and may make use of the reconstructed image derived from the body image data 103 based on the segmentations derived by the body part segmentation module 602.

The gender information extraction module 600 may derive the gender information using a number of different approaches, which may be used either independently or in combination with one another. A few examples of such approaches are presented below.

A first approach for distinguishing between men and women is the body shape, also referred to as body type. This approach is most useful in situations in which the body image data conveys information on the shape of the upper body, including the hip area of the individual under inspection, and in which the clothing of the individual has minimal impact on the body image data. Amongst others, this approach is particularly useful in situations where the body image data is derived using a full body scanner that uses millimeter waves since millimeter waves easily penetrate clothing but are reflected by the skin.

This approach considers that some body types are more common to women while others are more common to men. Generally speaking, it can be considered that there are five (5) different body types, which are generally illustrated in Figure 8.

Pear, Spoon, or Bell (800): This body type tends to be more common in females. This body type is characterized by the hip section being wider than the upper body. As the name suggest, this body type mimics the shape of a pear or bell.

V-Shape (802): This body type tends to be primarily associated with males. However, you cannot rule out that a female may have this shape. This body type is characterized as having proportionately smaller buttocks, a bigger chest, and a wider torso. To visualize this better, think of the V in the name, the individual will be wider from the shoulder and narrower around the waist and buttocks area.

Rectangle, Straight, or Banana (804): This body type is mainly associated with males or young (pre-pubescent) adults. This body type is characterized by the hip, waist, and shoulder sections being relatively similar in size. Thus the appearance of a rectangle, as if one could draw the lines of a rectangle around the body of the individual and touch all sides.

Hourglass (806): This body type tends to be more common in females. This body type is characterized as the body being significantly narrower in the waist in both the frontal and profile views. The waist is narrower than the chest region due to the breasts and narrower than the hip region due to the width of the buttocks. As the name suggest, this body type will resemble an hourglass, being most narrow in the center of the individual. Apple (808): This body type tends to be more common in males. This body type is characterized by the stomach region being wider than the hip region. This is going to give the appearance of an apple shape around the mid-section. More commonly, there will be more prominent love handles.

In a non-limiting implementation, each body type is associated with a likelihood score that the body type is associated with a male or female. For example, the "Pear, Spoon, or Bell (800)" body type may indicate that the individual is a female 90% of the time and is a male 10% of the time. In other words, of all individuals in a reference group considered to have Pear, Spoon, or Bell (800) body types, 90% were women and 10% were male. The "V-Shape (802)" body type may indicate that the individual is a male 85% of the time and is a female 15% of the time. In other words, of all individuals in a reference group considered to have "V-Shape (802)" body types, 85% were male and 15% were female. It is to be appreciated that the likelihood scores associated with the body types depend on the reference group considered. For example, the body types of Caucasian North Americans are significantly different from those of native Malaysians and therefore it would be expected that the likelihood that a given body type is associated with a male or a female would also be different for each of these groups. As such, the likelihood scores associated with the body types may be conditioned depending on the geographic location in which the practical systems would be used by using reference groups representative of the individuals that will be scanned when assigning likelihood scores.

Figure 7 shows a process implemented by the gender information extraction module 600 in order to make use of body type as an indicator for gender.

As step 700, the gender information extraction module 600 receives the reconstructed body image. In the specific example of implementation presented, the reconstructed body image is received from the image reconstruction and information extraction module 302 (shown in Figure 3). As step 702, the reconstructed body image is processed to derive geometric information associated with the individual under inspection. The geometric information includes parameters that may be used in determining gender, such as for example but not limited to:

- Shoulder dimension;

- Chest dimension

- Waist dimension

- Hip dimension

In addition to the above, or alternatively thereto, the geometric information may also include relative dimensions, or ratios, such as the following:

- Chest-to-waist ratio

- Hip-to-waist ratio

- Hip-to-chest ratio

- Hip-to shoulder ratio

In situations where the reconstructed body image includes multiple views of the individual (for example frontal/back views and side views), one or more of the views may be used to extract derive geometric information associated with the individual. The geometric information derived from the different view may also be compared to determine whether such information is consistent.

At step 704, the geometric information is processed to derive gender information conveying whether the individual is more likely to be male or more likely to be female. In a non-limiting example, the geometric information is used to derive which of the body types depicted in Figure 8 and described above is most likely to correspond to the individual. A likelihood score may associated with each body type conveying the likelihood that the geometric information corresponds to each body type. For example, based on the chest dimension, waist dimension and hip dimension (and/or ratios between these measurements such as the chest-to-waist ratio and the hip-to-waist ratio), the likelihood score may convey that the geometric information has:

- an 80% likelihood of corresponding to the "Pear, Spoon, or Bell (800)" body type;

- a nearly 0% likelihood of corresponding to the V-Shape (802) body type;

a nearly 0% likelihood of corresponding to the Rectangle, Straight, or Banana (804) body type;

- a 20% likelihood of corresponding to the Hourglass (806) body type; and

- a nearly 0% likelihood of corresponding to the Apple (808) body type.

In view of the above scores, since the "Pear, Spoon, or Bell (800)" body type and the Hourglass (806) body type are most often associated with females, it would follow that the individual being screened is more likely to be a female than a male.

Optionally, a likelihood score associated with the individual being a male (or female) may be derived. In a non-limiting implementation where each body type is associated with a likelihood score that the body type is associated with a male or female, based on the body type(s) derived, it may be determined that there is an 82% probability that the individual is a female (and thus an 18% probability that the individual is a male). The specific manner in which the probability (likelihood) score is derived may vary between implementations without detracting from the spirit of the invention.

It is also to be appreciated that the use of body type alone may not provide a conclusive and reliable indication as to whether the individual is male or female. In practical implementations, the body type may be used as one indicator amongst other indicators when determining the likely gender of the individual being screened. A second approach for distinguishing between men and -women is to focus the information conveyed by the reconstructed imase in the zroin area. Similarly to the above described approach based on body type, this second approach may be useful in situations in which the body image data conveys information on the shape of the individual under inspection in which the clothing of the individual has minimal impact on the body image data. Amongst others, this approach is particularly useful in situations where the body image data is derived using a full body scanner that uses millimeter waves since millimeter waves easily penetrate clothing but are reflected by the skin. This approach can consider that there is a difference between men and women in the positioning of the lowest part of the crotch, between the front and back views of the individual. Women tend to have these two measurements more equal. Men have a disceraable offset between the position of the lowest part of the crotch in the front and back view. This offset can be identified on the basis of a reconstructed body image derived from body image data by processing the reconstructed body image to derive geometric information associated with the individual under inspection. The geometric information includes parameters associated with the crotch area, in particular to the position of the lowest part of the crotch, between the front and back views of the individual. This geometric information is then used to derive indicator conveying whether the individual is likely to be male or female.

Alternatively or in addition to the above, this approach can consider that the shape of the male and female genitalia is different. In particular, male genitals are expected to be centered in the crotch and be somewhat elongated. The presence of male genitals can be identified on the basis of a reconstructed body image derived from body image data by processing the reconstructed body image to derive geometric information including parameters associated with the crotch area. The geometric information conveys whether there is the presence of an elongated shape corresponding to the penis and the presence of protuberances corresponding to the testicular sack. It is noted that size will vary as will whether or not some males have both testicles. The average flaccid (non-erect) penis length for a male is between 8.5cm and 10.5cm. This geometric information is then used to derive indicator conveying whether the individual is likely to be male or female. The absence of detection of male genitals (penis and/or testicular sack) may indicate that the individual is likely to be female.

It is to be appreciated that the use of information conveyed by the reconstructed image in the groin area may provide a somewhat reliable indication as to whether the individual is a male when the geometric information conveys an elongated shape (corresponding to the penis) and/or the presence of protuberances (corresponding to the testicular sack). The absence of detection of such shapes is not always conclusive that the individual is likely to be female (for example it may be a male with underdeveloped genitals). In practical implementations, the indicator related to the groin area may be used as one indicator amongst other indicators when determining the likely gender of the individual being screened.

A third approach for distinguishing between men and women is the detection o f gender specific items conveyed by the reconstructed image. This third approach may be useful in situations in which the body image data conveys information what the individual is wearing or carrying on his/her person (including but not limited to clothing, undergarments, shoes, jewelry, etc .). Amongst others, this approach can be useful in situations where the body image data is derived using a full body scanner that uses millimeter waves since millimeter waves easily penetrate clothing but are reflected by the skin. This approach may also be useful in situations in whch the body image data is derived using a visible light camera that uses visible light with little or no clothes penetrating capabilities to produce an image of the individual being scanned. More particularly, this approach considers that some items worn or carried by an individual are somewhat gender specific. As such, certain items may more commonly be found on women while others may more commonly be found on men. Examples such gender specific items may include certain articles of clothing (example undergarments, skirts, dresses). For example, the presence of a brassiere provides an indication that the individual is more likely to be female as it would be rare for a male to be wearing something like this. The presence of a brassiere can be identified on the basis of a reconstructed body image derived from body image data by processing the reconstructed body image to derive geometric information including parameters related to the presence of generally horizontal lines surrounding the chest area, for example.

Other examples of such gender specific items may include certain types of shoes. For example high heel shoes are worn (almost) exclusively by women and thus the presence of a high heel show provides an indication that the individual is more likely to be female as it would be rare for a male to be wearing something like this. The presence of a high heel shoe can be identified on the basis of a reconstructed body image derived from body image data by processing the foot body region of the reconstructed body image to derive geometric information related to the foot body region.

Other examples of such gender specific items include certain types of accessories. For example, the presence of hair clips/pins/brooches provides an indication that the individual is more likely to be female as it would be rare for a male to be wearing something like this.

Other examples of such gender specific items include gender specific hygiene products such as menstrual pads. It is also to be appreciated that the use of gender specific items may not necessarily provide a conclusive and reliable indication as to whether the individual is male or female in particular since women can wear essentially the same clothing types as men. Other than some exceptions (e.g. high heels, dresses, menstrual pad ...), very few items can truly be considered to be gender specific. In practical implementations, the gender specific items may be used as indicators amongst other indicators when determining the likely gender of the individual being screened.

There are many other secondary characteristics that can be used in the determination the likely gender of the individual and which may be derived based on geometric information extracted from the body image data.

One of these is the presence of breasts. The presence breasts, as conveyed in the geometric information by a marked change in size between the chest area and the waist area (in contrast with a gradual change in size), would be an additional indicator that the individual is likely to be female.

Another indicator of gender is the presence in the neck area of an Adam's apple. The presence of an Adam's apple, as conveyed in the geometric information by a marked protuberance in the neck area of the individual, would be an additional indicator that the individual is likely to be male.

Another indicator of gender is the average height of individuals. The average height for a Canadian male is 1.736m (5'8 ^{1 2} ") while the average female is about 1.595m (5'3"). The average American male is around 1.706m-1.789m (5'7"-5' 10 ^{1 2} "). Although not all individuals will have heights that are close to the respective average heights, the height can be used as an additional indicator of gender when combined with other indicators.

Other indicators of gender include the following: Males:

Greater mass of thigh muscle in front of the femur

Greater stature (taller in height)

Heavier Skull/ bone structure

Broader shoulders and chest

Higher waist to hip ratio

Females: - Greater development of thigh muscles in back (behind) femur than in front

Wider hips

Lower waist to hip ratio

- Upper arms approximately 2cm longer for given height In its generalized form, the determination of gender can be considered as a function F( ) of multiple indicators, including one or more indicators of the type described above. The function F( ) may assign different weight to different indicators depending on their nature and how conclusive that can be considered in defining gender. For example, the presence of an elongated shape in the groin area is likely to be considered as having more weight in determining that the individual is more likely to be a male in comparison to an indicator based on the individual's height.

The function F( ) may be expressed as follows: ( 1 ) LIKELY GENDER -

F (body type, groin information, detected artifacts, breasts, ..., others)

Where LIKELY GENDER can be {Male, Female, indeterminate} The function F() may be incorporate logic/heuristics rules in the determination of the likely gender of the individuals. The Function F() may also be configured to account for inconsistencies. For example, if clearly female gender specific artifacts are detected (say high heel shoes) and the information in the groin area clearly indicates the presence of male genitalia, an inconsistency indicator may be raised, which could signal the presence of a potential anomaly. This may be conveyed, for example, by setting the LIKELY GENDER to "indeterminate".

It is to be appreciated that, although some examples of approaches for deriving gender information conveying whether the individual being inspected is more likely to be a man or a woman have been presented above, other suitable approaches are possible and will become readily apparent to the person skilled in the art in view of the present application.

It is also to be appreciated that although in the specific example described the body image data on which the anomaly detection is performed and the body image data used to derive the gender information associated with the individual under inspection are the same, this may not be the case in alternative examples of implementation. For example, a full-body imager using millimeter waves may be used to generate body image data for use in anomaly detection while a different type of image data may be used to derive the gender information. For example, the gender information may be derived by processing an image taken by a visible light camera, the image conveying a representation of the face of the individual under inspection. In such an implementation, any suitable method for deriving gender information for a facial image may be used. A non-limiting example of such a method is described for example in U.S. Patent No. 7,912,246 issued on March 22, 2011 to Moon et al. and entitled "Method and system for determining the age category of people based on facial images". The contents of the aforementioned document are incorporated herein by reference. Several other variants are also possible and will become apparent to the person skilled in the art in light of the present description. In addition to the above, although the gender information extraction module 600 has been described as being part of the anomaly detector 304 (shown in Figure 3 and Figure 6), it will be appreciated that, in alternative examples of implementation, this module may be a separate component and be external to the anomaly detector 304 or even a component external to processing engine 104 (shown in Figure 1).

Specific Practical Implementations

Certain portions of the system for performing a security screening at security checkpoint on an individual can be implemented on a general purpose digital computer 1600, of the type depicted in Figure 16, including a processing unit 1602 and a memory 1604 connected by a communication bus 1614. The memory 1604 includes data 1608 and program instructions 1606. The processing unit 1602 processes the data 1608 and the program instructions 1606 in order to implement the functionality described above with reference to processing engine 104 (depicted in Figure 1 and Figure 3) including the image reconstruction and information extraction module 302 and automated anomaly detection module 304. The digital computer 1600 may also comprise one or more I/O interfaces for receiving or sending data elements to external devices. For instance, the digital computer may include an I/O for receiving body image data derived by the body scanning device 102 (shown in Figure 1) by subjecting an individual to radiation (millimeter wave) and another I/O in communication with display device 106 (also shown in Figure 1), such as for conveying information to a security operator.

Alternatively, portions of processing engine 104 (shown in Figures 1 and 3) can be implemented on a dedicated hardware platform where electrical/optical components implement the functions described in the specification and depicted in the drawings. Specific implementations may be realized using ICs, ASICs, DSPs, FPGA or other suitable hardware platform. Other alternative implementations of the processing engine 104 can be implemented as a combination of dedicated hardware and software.

It will be appreciated that the system for performing a security screening at security checkpoint on an individual 100 that is depicted in Figure 1 may also be of a distributed nature where the body image data is obtained at one location (or more than one location) and transmitted over a network to another entity implementing the functionality of processing engine 104 described above. Another unit may then transmit a signal for causing one or more display devices to display information to the user, such as a reconstructed image of the individual being screened and optionally information conveying the presence of detected anomalies. The display device(s) may be located in the same location where the body image data was obtained or in the same location as the server unit or in yet another location. In a non-limiting implementation, the display device(s) may be part of a centralized screening facility.

Figure 17 illustrates a network-based system 1700 for screening individuals in accordance with a specific example of implementation of the invention. The system 1700 includes a plurality of operator stations 1702, 1704, 1706 and 1708, as well as body scanning (inspection) devices 1760 A and 1760B connected to a shared processing entity 1710 through a network 1712. Communication links 1714 between the operator stations 1702, 1704, 1706, 1708, the body scanning (inspection) devices 1760A, 1760B and the shared processing entity 1710 can be metallic conductors, optical fibers or wireless, for example. The network 1712 may be any suitable network including, but not limited to, a global public network such as the Internet, a private network and a wireless network. The shared processing entity 1710 may be adapted to process information received from the body scanning (inspection) devices 1760A and 1760B and issue signals conveying results to the operator stations 1702, 1704, 1706 and 1708 using suitable methods known in the computer- related arts. The shared processing entity 1710 includes a program element 1716 for execution by a CPU (not shown). Program element 1716 includes functionality to implement the functionality of processing engine 104 (shown in Figures 1 and 3) described above. Program element 1716 also includes the necessary networking functionality to allow the shared processing entity 1710 to communicate with the operator stations 1702, 1704, 1706 and 1708, as well as the body scanning (inspection) devices 1760A and 1760B over the network 1712. In a specific implementation, the operator stations 1702, 1704, 1706 and 1708 include display devices responsive to signals received from the shared processing entity 1710 for displaying screening results derived by the shared processing entity 1710.

Although the present invention has been described with reference to certain preferred embodiments thereof, variations and refinements are possible without departing from the spirit of the invention.