FIELD OF THE INVENTION

[0001] The present invention relates to a video based access control system, in particular a system for controlling access to a premises based on detecting a non- obscured face.

BACKGROUND TO THE INVENTION

[0002] Business premises desire to provide unimpeded access to clients in order not to deter would be clients, particularly where the choice of business by the client is discretionary. Unfortunately unimpeded access to potential clients also provides unimpeded access to armed bandits.

[0003] To help keep out undesirables some premises lock their front doors and only admit people who look non-threatening. A suitable determinant of admissibility is the ability to see a person's face in full. Whilst not foolproof, if a person allows their face to be seen and thus be recorded by surveillance cameras and be recognised by shop attendants one would have some confidence that they are unlikely to rob the premises. In such a situation staff would need to be constantly monitoring the door for

approaching customers in order to not delay and potentially lose customers as they wait to be admitted. This has obvious cost implications.

[0004] The applicant has recognised a need for a system that can recognise if a person has a clearly visible face and control access to a premises accordingly

[0005] The object of this invention is to provide an access control system that automatically provides access to premises to people with a non-obscured face and bars access to people with an obscured face.

SUMMARY OF THE INVENTION

[0006] In a first aspect the invention comprises an access control system with a video stream input, image processing function and an access control output wherein the video stream in analysed by the image processing function for the presence of an unobscured human face and upon detection of an unobscured face the access control output is asserted. Preferably the image processing function is a neural network employing Haar descriptors. [0007] Preferably the system further comprises an indicia output, wherein the video stream is further analysed for the presence of an obscured human face and the indicia output is asserted when an obscured human face is detected. Preferably the further analysis processing function is performed by a further neural network employing Haar descriptors.

[0008] In preference the video stream is further analysed for the presence of a live human face and the access control output is only asserted if a live human face is detected. The presence of a live human face may be determined by temporal variations in features of the face or by performing a difference abstraction on the colours of a series of frames of the video stream to produce a colour difference and comparing the colour difference with a predetermined threshold.

[0009] It should be noted that any one of the aspects mentioned above may include any of the features of any of the other aspects mentioned above and may include any of the features of any of the embodiments described below as appropriate.

BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Preferred features, embodiments and variations of the invention may be discerned from the following Detailed Description which provides sufficient information for those skilled in the art to perform the invention. The Detailed Description is not to be regarded as limiting the scope of the preceding Summary of the Invention in any way. The Detailed Description will make reference to a number of drawings as follows.

[0011] Figure 1 is a schematic diagram of a premises incorporating an access control system according to the current invention.

DETAILED DESCRIPTION OF THE INVENTION

[0012] The following detailed description of the invention refers to the

accompanying drawings. Wherever possible, the same reference numbers will be used throughout the drawings and the following description to refer to the same and like parts. Dimensions of certain parts shown in the drawings may have been modified and/or exaggerated for the purposes of clarity or illustration.

[0013] The present invention provides a video based access system that can recognise that a face is not obscured and subsequently grant access to a system. It differs from conventional face detection access systems in recognising the presence of a face, not the identity of the face.

[0014] Face detection, as opposed to face recognition, is a relatively well known technology with applications found in diverse areas such as video and camera auto- focus and auto-exposure and face tracking in video games. Such applications seek to find a face, often when the face is obscured. The present system also detects faces, but then goes one step further in determining if the face is obscured.

[0015] A premises incorporating an access control system 10 according to the present the present invention is shown in schematic form in Figure 1. A person 20 approaches the door 25 of the premises and has their image captured by video camera 30. With suitable interfaces video in a variety of formats can be accepted, e.g.

analogue video, digital video or video over IP. The video signal 32 is processed by the access control system 10 which determines if the person should be admitted and if so asserts a control signal 42 to the door controller 40 which in turn opens the door. An optional alert signal 72 for driving an indicia 70 is asserted when a face is detected that is obscured. This allows an operator to manually override the system (by means not shown such as a push button). Such a need could arise for instance when a person wears a peaked hat particularly low or is heavily bearded.

[0016] The access control system may be implemented by various different means, however all implementations encapsulate the core concept of determining if a face is obscured and controlling access accordingly. A preferred embodiment is presented below.

[0017] The access control system 10 is logically split into a control logic function 65 and an image processing function 50. Both functions may be implemented on a variety of hardware platforms, but in a preferred embodiment the image processing function 50 is performed by a general purpose "off the shelf processor module incorporating an Intel i7 processor running a Linux based operating system, whilst the control logic function 65 is implemented on a custom hardware module to allow for flexible

interfacing to the door controller 40 and indicia 70. The custom hardware module is preferably a microcontroller driving relay contacts, but could be implemented by any suitable means. Where the door controller and indicia provide suitable interfaces the control function can instead be integrated into the same processor module as the image processing function. For a custom hardware version of the controller a generic PC interface such as USB to the image processing processor module can be used. The video input signal 32 can be delivered by various means, such as composite video, IP or USB, with the use of suitable hardware interfaces and software drivers as are well known in the art.

[0018] The access control system 10 is centred around the image processing function 50. This again may take on a variety of software implementations, with a preferred embodiment being based on the Open CV image processing software. The image processing function performs three main tasks: face detection and extraction 51 ; obscuration detection 52; and liveness detection 53. Optionally images are captured and stored in an image store 54 with a record of whether or not a face was deemed as being obscured. Post event analysis of this data allows for retraining of the system to better discriminate obscured faces.

[0019] In the first instance the video signal is passed to the face detection and extraction function 51 which detects if a face is present in the video image and if so extracts the portion of the image containing the face 33 for further processing by the other image processing functions. This is preferably implemented by a first neural network, although any suitable technique may be used.

[0020] The face portion of the image is processed by the obscuration detection function 52 to determine if the face is obscured or not. Such obscuration may be by a mask, dark glasses or a scarf etcetera. In a preferred embodiment the obscuration detection is performed by a second neural network using Haar cascades 55 as identifiers. Having the correct Haar cascades is key to the system performing reliably, that is accurately determining if a face is obscured or not.

[0021] The Haar cascades 55 are stored in the image processor 50, but may originate from a second source. The Haar cascades are determined by training a neural network with a variety of positive and negative images, the positive images showing faces that are deemed to be obscured and the negative images showing faces that are not obscured.

[0022] The third image processing function is the liveness detection function 53 which determines if the face that has been detected is a live face. Such a function is well known and may be implemented by a variety of means. Typically temporal variances in the face image such as movement of a mouth or blinking of an eye are looked for. This allows a live face to be distinguished from a photo held up to the camera or a rubber mask worn by a would be robber.

[0023] Confidence in the liveness detection may be further strengthened by capturing a series of frames and performing a difference abstraction on the colours of the frames. When the difference is multiplied by a large value, typically ~ 200x, a significant colour difference of sectors of the face must be observed as well as the face detectors being satisfied in order to assert a live face. Comparing the colour

differences to a predetermined threshold can be used to determine if a face is a live face.

[0024] The outputs from the image processing blocks: face detection 61 ;

obscuration detection 62 and liveness detection 63 are fed to the control function 65 which in turn asserts the access signal 42 and alert signal 72. The access signal is asserted if a face is detected that is not obscured and the alert signal 72 is asserted if a face is detected that is obscured. The face detection signal may also be used to suspend the operation of the other image processing functions in order to reduce processor usage.

[0025] The reader will now appreciate the present invention which provides an access control system that can automatically control access to a premises based on detecting if a person's face is obscured or not.

[0026] Further advantages and improvements may very well be made to the present invention without deviating from its scope. Although the invention has been shown and described in what is conceived to be the most practical and preferred embodiment, it is recognized that departures may be made therefrom within the scope and spirit of the invention, which is not to be limited to the details disclosed herein but is to be accorded the full scope of the claims so as to embrace any and all equivalent devices and apparatus. Any discussion of the prior art throughout the specification should in no way be considered as an admission that such prior art is widely known or forms part of the common general knowledge in this field.

[0027] In the present specification and claims (if any), the word "comprising" and its derivatives including "comprises" and "comprise" include each of the stated integers but does not exclude the inclusion of one or more further integers.