TECHNICAL FIELD

The invention relates to a luminaire, a system and a method for adaptive illumination.

BACKGROUND ART

Intelligence of lighting system is steadily increasing to address new efficiency and simplicity requirements. For example, to achieve relatively sophisticated lighting behaviors, adapted to the intended use of an environment, it is possible to manually set up different preset settings for the lighting infrastructure at the installation phase of a lighting system. With regard to the energetic efficiency, solutions exist, which typically adapt presence detection sensors that automatically turn on or off lights. However, this simple binary presence detection does not take into account more complex information, such as about the activity of people. Moreover, these lighting systems are not adaptive with respect to new or changing situations.

US 2008/0265799 Al relates to providing energy efficient and intelligent illumination using distributed processing across a network of illuminators to control the illumination for a given environment.

SUMMARY OF THE PRESENT INVENTION

It is an object of the present invention to provide a luminaire, a system and a method for illuminating a surrounding area, wherein the illumination is autonomously and flexibly adapted according to changing requirements, while increased energetic efficiency, higher flexibility and easy user control are provided.

The object is solved by the features of the independent claims.

The invention is based on the idea to adjust lighting configuration based on data mining, e.g. to use a typical pattern or history of detected data for adapting an illumination behavior. Preferably, the data correspond to activity data detected or sensed in an environment or surrounding area to be illuminated. Here, the activity data may be detected at a single point in the vicinity of a luminaire or illumination unit. Alternatively, the activity data may be detected within a larger predefined surrounding area of the luminaire or illumination unit. Thus, information for lighting configuration may be generated depending on a spatial segmentation of an area of interest based on statistically learned patterns in the area of interest from the spatial information provided by a sensor. By these means, the illumination is adaptable to changes of the usage of a particular place with time, without human intervention or cumbersome installation requiring knowledge of the expected use of the environment.

According to one aspect of the present invention, a luminaire for adaptive illumination is provided comprising an illumination unit, an activity sensor unit and a control unit. The illumination unit is adapted to illuminate the environment or surrounding area of the luminaire. For this, the illumination unit may comprise one or more light sources with adjustable lighting properties, such as light color, light intensity, light quality, color rendering index, correlated color temperature or the like. For instance, LED elements, fluorescence lamps, incandescent lamps, HID lamps, halogen lamps, etc. or a combination thereof may be used. Moreover, the illumination unit may be configured to change the direction and/or shape of the light beam. Then, at least one of shape, size or location of the illuminated area may be adjustable. The activity sensor unit is adapted to sense or detect activity within the environment of the luminaire or of the illumination unit. Therefore, the activity sensor unit is preferably provided or associated with the luminaire or illumination unit, so that the activity within the area to be illuminated can be measured. Thus, the sensitive area of the activity sensor unit may correspond to the direct vicinity or to an area below or around the luminaire (or illumination unit). Possibly, the sensitive area of the activity sensor unit corresponds even to a single point. The activity sensor unit may comprise one or more sensors adapted to sense activity, including at least one of a conventional camera, a range imaging sensor (such as stereo camera, time-of- flight sensor, structured light 3D scanner, coded aperture camera etc.), a microphone array, an ultrasound sensor, a microwave sensor, a laser radar sensor, an infrared sensor or any combination thereof. Preferably, the activity sensor unit is adapted to sense motion, speed of motion and/or direction of motion. Furthermore, the control unit is adapted to generate at least one history from the activity data sensed by the activity sensor unit. The history of activity data may refer to activity data that are sensed over a

predetermined data recording interval. This data recording interval may be adjustable or set during commissioning. Based on the generated history (or histories), the control unit may further be adapted to adjust operation characteristics of the luminaire. Therefore, the luminaire is capable of learning about the activity in its surroundings and to adjust its functionality or operation accordingly. Thus, the luminaire is capable of learning an appropriate illumination without requiring manual re-configuration or updates.

Preferably, it is differentiated between different daytimes, so that a history may be recorded, respectively, for night time, in the morning, during lunch time etc.. Thus, in a preferred embodiment, the control unit may be capable of generating histories of activity data for different daytimes, respectively, and adapt the operation characteristics accordingly. Alternatively or additionally, it may be likewise differentiated between weekdays and weekends or holidays.

Possibly, the predetermined data recording interval defining the detection time interval of activity data included in the history may be adjustable. For instance, the data recording interval may be set to an hour, a day, a week or the like. Thus, the history may be based on activity data of the recent past. For updating the history, a sliding window algorithm or the like may be used, so that new data replace old data. However, also any other algorithm for obtaining a current history comprising activity data of the recent past may be used. By these means, also the speed of adaptation may be adjusted. In some circumstances, a slow adaptation may be preferred avoiding consideration of temporary changes. In other situations, a fast reactivity may be required in order to quickly adapt the operation of the luminaire.

In a preferred embodiment, the adjusted operation characteristics are used as new operation parameters of the luminaire, the illumination unit, the activity sensor unit and/or the control unit. For this, the new operation parameters may be stored for future use. Therefore, the operation characteristics of the luminaire may be considered as adaptive operation parameters or adaptive default parameters. The basic operation features and thus the functionality of the luminaire may be permanently or continuously adapted.

The luminaire may further comprise a memory unit for storing at least one of activity data, one or more histories of activity data, adjusted operation characteristics, or the like. Possibly, the history may be stored as a single activity value, which is updated using new activity data provided by the activity sensor unit. For instance, this single activity value may correspond to a mean, average, or the like. Preferably, the memory unit stores activity data recorded for a predetermined time.

In addition, the luminaire may further comprise means for wireless communication, so that the luminaire may be capable of communicating or exchanging data with other luminaires or external control centers in its surroundings. Also, the luminaire may comprise a user interface, so that a user can manually change at least one of operation characteristics, parameters corresponding to a current illumination, or the like. Preferably, the operation characteristics refer to parameters of at least one of the luminaire, the illumination unit, the activity sensor unit and the control unit, the operation characteristics characterizing the respective operation or functionality. For instance, operation characteristics include at least one of: a turn-on time relating to the time for reaching the final illumination state from an off-state (no light); a turn-off time relating to the time for switching off the illumination from an on-state; an activation time relating to the time interval, during which the luminaire, the illumination unit, the activity sensor unit and/or the control unit is active; an activation schedule relating to a daytime or date, when the luminaire, the illumination unit, the activity sensor unit and/or the control unit is activated; an illumination reactivity relating to the reaction speed with respect to current changes of activity within the surroundings (e.g. change of occupancy or presence); an adjustment reactivity relating to the adaptation speed for adjusting operation characteristics with respect to more continuous changes of activity, i.e. based on the history; the predetermined data recording interval for sensing activity data for the history; a lighting level; a maximum light intensity; a minimum light intensity; one or more intermediate intensity levels; a light color; a color rendering index; a light quality; a correlated color temperature; a shape of light beam; a direction of light beam; and an illumination zone, i.e. the size of the illuminated area. Of course, also more than one value may be set for these parameters, e.g. for different situations. For instance, several minimum intensity levels may be set or the like. Thus, operation characteristics may relate to any parameters used by the luminaire during operation, but not only to parameters relating to illumination per-se. Preferably, these operation characteristics can be individually adjusted. Alternatively or additionally, predefined sets of operation characteristics may be adjusted together, e.g. in dependency of each other. This may be advantageous for inter-correlated operation characteristics.

In a further preferred embodiment, the activity data include at least one of an occupancy level relating to the number of persons detected within the surroundings, speed of motion, direction of motion, maximum speed, average speed and frequency of motion detection, i.e. how often motion is detected within a predefined time interval. Here, at least one of the speed of motion, the direction of motion, the maximum speed, the average speed and the frequency of motion detection may be determined for each person separately and/or simultaneously. Alternatively, the speed of motion, the direction of motion, the frequency of motion detection and/or the maximum speed may be determined for a plurality of persons together. Also, the motion of one or more persons may be observed over a predetermined time interval in order to determine one or more of these parameters. For instance, the average speed or the frequency of motion detection may be defined by a predetermined time interval. Of course, any other quantitative measures for characterizing motion may be applied. By these means, the luminaire may be adapted to differentiate between surroundings having different activity based on occupancy and/or motion detected therein. Thus, according to a preferred embodiment, a luminaire is provided that can flexibly adapt its characteristics depending on typical speed patterns that are observed in its vicinity for a certain amount of time.

Furthermore, the control unit may be capable of considering at least one of user control, ambient noise, ambient brightness, installation settings, time of day and date for adjusting the operation characteristics. For this, the luminaire may also comprise a light sensor and/or a sound sensor.

Additionally or alternatively, the control unit may determine an activity pattern of the surrounding area based on the detected activity data. For instance, the control unit may be capable of differentiating between desk activity and corridor activity and to adjust the operation characteristics accordingly. This is highly advantageous, since a person working at a desk will most likely require a higher illumination quality than a person walking down a corridor. Possibly, a plurality of activity patterns is predefined, so that the activity data or the history of activity data can be discretely classified. Moreover, one or more operation characteristics or one or more sets of operation characteristics may be predefined corresponding to an activity pattern. For instance, a desk activity pattern may correspond to a set of operation characteristics including high lighting level, high light quality, slow illumination reactivity and long turn-on time and turn-off time.

Possibly, a histogram is generated from the detected activity data for adjusting the operation characteristics of the luminaire. Here, the histogram may correspond to the history of activity data sensed for the predetermined time. A sliding window method may be used for updating the histogram with respect to new activity data provided by the activity sensor unit. Therefore, the histogram comprises activity data of the recent past. For analyzing the histogram, a fit function may be used. Alternatively, a discrete quantization may be performed, e.g. counts of activity below or above a certain threshold, higher/lower speed of motion than a predetermined value etc. Possibly, the histogram is also used for determining the activity pattern.

The operation characteristics may be adjusted based on a predefined mapping function, which links the activity data, the history of activity data, the histogram of activity data and/or the activity pattern to one or more operation characteristics. Preferably, however, this mapping function can be used without prior determination of activity patterns. The mapping function may relate to a fit function, which can be either applied directly to the activity data, to pre-processed activity data, to the history of activity data and/or to the histogram of activity data. In one embodiment, the mapping function may relate to a family of curves defining a fit function for one or more operation characteristics, respectively. For instance, the illuminated area or surface A may be a function of the occupancy level:

A=A(n). In another example, the lighting intensity I may be a function of the detected ambient brightness I ambient, the daytime t and the detected speed of motion v:

I=I(I_ambient, t, v).

In a further embodiment, the luminaire, the activity sensor unit, the illumination unit and/or the controller is activated at predefined daytimes and/or in regular time intervals and/or when the luminaire or the illumination unit is manually activated by a user. Alternatively, the luminaire, the activity sensor unit and/or the illumination unit may be continuously activated. By these means, the operation of the luminaire or of the activity sensor unit may be adapted to the requirements, thus optimizing the energy efficiency.

In a preferred embodiment, a system for adapted illumination is provided comprising a plurality of luminaires according to one of the above-described embodiments, wherein the luminaires are capable of wirelessly communicating with each other and/or with a control center. In this embodiment, the luminaires may adapt their operation characteristics to each other.

According to a further aspect of the present invention, a system for adaptive illumination is provided, comprising at least one illumination unit, an activity sensor unit associated with the illumination unit and a control unit. For instance, the activity sensor unit may be integrated in or attached to an illumination unit. Similarly, the illumination unit and the activity sensor unit may be provided together, e.g. both comprised in a luminaire unit. Furthermore, the activity sensor unit is capable of sensing activity data in the surroundings of the illumination unit. The control unit is capable of generating at least one history of activity data for adjusting operation characteristics of the system based on the generated history of activity data. In one embodiment, the operation characteristics of one or more illumination units are adjusted. Alternatively or additionally, the operation characteristics of one or more activity sensor units are adjusted. Preferably, the system comprises only one control unit, which may be wirelessly connected to the illumination units and/or to the activity sensor units. Furthermore, the control unit may be adapted to receive control commands from a remote user control interface. Wireless communication may be performed using infrared communication, Bluetooth, ZigBee, WLAN, or radio communication.

Preferably, the luminaire or system according to one of the above-described embodiments is adapted to sense the activity data in a spatially resolved manner. For instance, data detected at one point of the surrounding area may be distinguished from another point of the surrounding area. This spatially resolved activity data may also be referred to as spatial information of the observed surroundings or area of interest. The spatial information may be used for learning patterns of activity in the surrounding area of interest based on statistical methods in order to spatially divide the area into segments. As described before, the spatial information may be recorded for some time, so that a history of spatial information can be derived for adjusting operation characteristics of the illumination unit accordingly.

A yet further embodiment of the invention relates to a system for lighting configuration, wherein the system comprises at least one sensor for providing spatial information of an observed an area of interest or surrounding area, and a control unit being configured to process the spatial information provided by the at least one sensor by performing the following acts of statistically learning patterns in the area of interest from the spatial information for a predefined time intervall, spatially segmenting the area of interest based on the statistically learned patterns, and generating information for lighting

configuration depending on the spatial segmentation of the area of interest.

The control unit of any embodiment of a system according to the present invention is preferably configured to perform a method of the invention as specified below.

According to another aspect of the present invention, a method for adaptive illumination is provided, comprising the steps of: sensing activity data in a surrounding area; generating at least one history of activity data sensed for a predetermined time; and adjusting operation characteristics based on the history of activity data. Furthermore, the method may further comprise the step of illuminating the surrounding area based on the adjusted operation characteristics. Also here, the area, in which activity data are detected, corresponds to the area to be illuminated. The operation characteristics may relate to a luminaire as described above or to the operation characteristics of a system as described above. Thus, the method according to the present invention may be performed for realizing a luminaire or system for adaptive illumination according to one of the above-described embodiments of the present invention. According to a further aspect of the present invention, an illumination or lighting is configured depending on a spatial segmentation of an area of interest based on patterns, which were statistically learned in the area of interest from the spatial information provided by a sensor for a predefined time intervall. The patterns may be for example patterns of activity in the area of interest, patterns of changes, patterns of colors, patterns of appearance features or the like. The predefined time intervall may last few hours or even days, so that a large amount of data may be available for the statistical learning of patterns, e.g. based on a data mining technique. In particular, a data mining algorithm may be employed in order to analyze the statistically learned patterns and/or to extract high-level information from a large number of observations of the area of interest. The so learned patterns may be used to spatially segment the area of interest, for example in segments with high and low activity. This segmentation may then be used for generating information for lighting configuration, e.g. for displaying the segmentation on a computer screen, so that a user may configure the lighting in accordance with the processed segmentation, or for creating light settings for light units located in the area of interest so that a fully automatic lighting configuration in an environment can be achieved. By the pattern learning based segmentation of an area of interest, the lighting configuration may be better adapted to the usage of an environment.

Therefore, according to a further embodiment of the invention, a method for lighting configuration is provided, comprising the steps of receiving spatial information of an observed area of interest provided by at least one sensor; statistically learning patterns in the area of interest from the sensor information for a predefined time intervall; spatially segmenting the area of interest based on the statistically learned patterns; and generating information for lighting configuration depending on the spatial segmentation of the area of interest. Here, the spatial information may be information from the area of interest, which contains an assignment of a detected information in the area of interest, for example an activity, to a location in the area of interest. Spatial information may be for example obtained from a camera capturing pictures from the area of interest, a microphone array, an array of PIRs (passive infrared receivers), ultrasound arrays, TOF (Time of Flight) cameras or radar sensor sampling the area of interest for changes or activities.

Preferably, the statistically learning of patterns in the area of interest from the spatial information for a predefined time intervall may comprise analyzing the statistically learned patterns for partitions of the area of interest with different features, and the spatially segmenting of the area of interest based on the statistically learned patterns may comprise spatially segmenting the area of interest depending on the analyzed partitions with different features. Thus, the area of interest may be spatially segmented with regard to different features, such as average motion detection, particularly slow and fast motion, direction of motion, speed of motion, detection of a change and/or color(s). For example, a partition with a clear trend in the direction of motion may be assigned to a walking area, where many people walk through (e.g. entrances in rooms). A partition with a slow motion and frequent detection of changes may be assigned to a working area, where people usually move slowly, and things such as files and books on a desk are frequently moved from one location to another. Also, colors may analyzed for determining partitions, for example a partition with a many color changes may be assigned to a walking area, since many people with differently colored clothes walk through this partition within a short time period.

Moreover, the segments of the area of interest may be classified. Such a classification with three different regions "region with no activity", "walking area", "desk area" can be aleady performed after only a few hours of obervation of the are of interest. Another embodiment of the invention provides a computer program enabling a processor to carry out the method according to the invention and as specified above.

According to a further embodiment of the invention, a record carrier storing a computer program according to the invention may be provided, for example a CD-ROM, a DVD, a memory card, a diskette, internet memory device or a similar data carrier suitable to store the computer program for optical or electronic access.

A yet further embodiment of the invention provides a computer programmed to perform a method according to the invention such as a PC (Personal Computer) and comprising a first interface for receiving spatial information of an observed area of interest from at least one sensor and a second interface for outputting a generated information for lighting configuration. The computer may execute a program with a graphical user interface, allowing a user to comfortably configure a lighting created with one or more controllable light units of a lighting system.

The invention will be described in more detail hereinafter with reference to the exemplary embodiments.

BRIEF DESCRIPTION OF DRAWINGS

Fig. 1 shows a luminaire according to an exemplary embodiment of the present invention. Fig. 2 shows a system for adaptive illumination according to an exemplary embodiment of the present invention.

Fig. 3 shows a set of operation characteristics for different activity patterns according to an exemplary embodiment of the present invention.

Fig. 4 shows histograms of motion speed according to an exemplary embodiment of the present invention.

Fig. 5 shows a flow diagram of a method for adaptive illumination according to an exemplary embodiment of the present invention.

Fig. 6 shows a simple block diagram of another embodiment for a system for lighting configuration according to the invention.

Fig. 7 shows a top view of a small office environment as area of interest as captured with a camera embedded in the ceiling of the room, wherein zones of different activity of people are marked.

Fig. 8 shows a flowchart of an embodiment of the method for lighting configuration according to the invention.

Fig. 9 shows a flowchart of an embodiment of the act of spatially segmenting the area of interest according to the invention.

Fig. 10 shows a flowchart of a further embodiment of the act of spatially segmenting the area of interest according to the invention.

Fig. 11 shows a flowchart of an embodiment of the act of computing histograms according to the invention.

DESCRIPTION OF PREFERRED EMBODIMENTS

In Fig. 1, a luminaire 100 according to an exemplary embodiment of the present invention is shown. The luminaire 100 comprises an illumination unit 110, an activity sensor unit 120, a control unit 130 and a memory 140. The illumination unit 110 can comprise one or more light sources, such as an LED, a halogen lamp, a fluorescence lamp or the like. The illumination unit 110 can further be configured to change the direction or shape of the light beam, so that the shape, size or location of the illuminated area can be adjusted. Also the activity sensor unit can comprise more than one sensor element, e.g. a presence sensor, a motion sensor, a speed sensor or the like. The activity sensor unit 120 is configured to measure the activity in the direct neighborhood of the luminaire 100, e.g. at a single point or in a small area below or around the luminaire 100. Furthermore, the luminaire 100 can comprise further sensors (not shown), e.g. for sensing ambient brightness, ambient sound, etc.

In a further embodiment as exemplarily shown in fig. 2, a system for adaptive illumination is provided. Here, the illumination unit 110 and the activity sensor unit 120 are co-located. In the example shown in fig. 2, they are both included in a luminaire unit 150. However, the activity sensor unit 120 can also be included in the housing of the illumination unit 110 or vice versa. The controller 130 is provided separately from the luminaire unit 150 and can comprise a memory 140. Alternatively, the memory 140 can be provided in the luminaire units 150. In any case, the controller 130 can differentiate between activity data measured for the different luminaire units 150. In a preferred embodiment, the controller 130 and the luminaire units 150 can communicate wirelessly with each other in order to exchange operation characteristics and activity data. Hence, the system may be employed as a wireless network. Additional sensors, e.g. for sensing the ambient brightness, can either be provided in the luminaire units 150, in the control unit 130, or separately.

In the following, the invention is explained using the example of the luminaire 100 shown in fig. 1. However, the invention is not limited thereto, but the embodiments described below can also be transferred to the system 200. Furthermore, the principles of the invention are described using the example of an activity sensor unit 120 capable of measuring velocity and direction of motion. Thus, a conventional camera with a computer vision algorithm, a range imaging sensor (such as stereo camera, time-of- flight sensor, structured light 3D scanner, coded aperture camera etc.), a microphone array with a speed estimation algorithm, an ultrasound based sensor, a microwave or lased radar sensor and the like or any combination thereof can be used as sensor elements of the activity sensor unit 120 in order to assess direction or speed of motion in the vicinity of the luminaire 100. Again, the invention is not limited to activity data related to speed or direction of motion, but also any other activity data may be used.

According to a preferred embodiment, the velocity is measured when activating the luminaire 100, e.g. whenever an object or a person moves in the range of the luminaire 100. Then, the velocities measured in the recent past are exploited to extract simple yet reliable information about the environment in the direct vicinity of the luminaire 100. This information is then used to modify the operation characteristics of the luminaire 100, of the activity sensor unit 120 or of the illumination unit 110. For instance, as shown in fig. 3, the operation characteristics of the illumination unit 110 can include activation parameters, e.g. a detection time to, a light activation start time t _ls a light activation end time t ₂ , a light deactivation start time t ₃ , a light deactivation end time t ₄ and a final illumination value I _max .

In fig. 3, examples for illumination characteristics of areas with different usage are shown. Fig. 3A illustrates an example of an illumination pattern for an area showing "desk activity", while fig. 3B illustrates an example of an illumination pattern for an area showing "corridor activity". In the following, two luminaires 100 installed in an office are compared in order to illustrate the adaptation of illumination patterns based on the typical activity in the surroundings of the respective luminaires 100. The first luminaire 100 is installed on top of a desk area, while the other luminaire 100 is arranged in a passage area or corridor. After installation and a few hours of a typical working day, the luminaire 100 mounted in the desk area will have observed mainly a large series of slow movements, with the exception of few fast movements representing the instances, in which a person arrived or left the working area. Thus, this luminaire 100 will detect a "desk activity" as typical activity in its surroundings. In the same span of time, the luminaire 100 in the corridor will have observed a series of fast movements, corresponding to the many people passing by, and a few instances of slow movements corresponding to the situations of people stopping to discuss. Hence, the luminaire 100 located in the corridor will sense a "corridor activity" being the typical activity in its vicinity. However, a person working at a desk and a person walking down a corridor have very different requirements with respect to illumination. While the person at the desk will require a high quality illumination for avoiding fatigue and degrading concentration, the person in the corridor will only require a basic illumination of the corridor in his walking direction.

Consequently, the illumination for the desk area should be set to be intense and slowly reactive to changes in order to maintain a good level of light required for optimal visibility, as shown in fig. 3A. Thus, the interval between the first detection of motion or presence (arrow in fig. 3A) at time to and the light activation start time ti can be longer than for areas with other activity. Moreover, the time t ₂ , until which the final illumination level Imax is reached, can be set to be later. This final illumination level I _max will relate to a high intensity or high quality illumination. Also, the time for deactivating the illumination, i.e. the interval between the light deactivation start time t ₃ and the light deactivation end time t ₄ , can be set longer. Advantageously, the illumination unit 1 10 of the luminaire 100 installed in the desk area does also not turn off itself in case that no or only few movement is detected for a short time. In contrast, the illumination behavior of the luminaire 100 located in the corridor will be set to be reactive, both in activation and deactivation, and to a lower intensity I _max , since this is sufficient for people walking by. An example for such an illumination pattern is shown in fig. 3B. Here, the detection time to and the light activation start time are

substantially simultaneous. Moreover, the time until reaching the final illumination level I _max or for deactivation is shorter than in the desk area. Likewise, when detecting no activity in the vicinity, the luminaire 100 in the corridor may be faster switched off or switched to a standby state. In addition, the direction of illumination can be adjusted to the direction of motion or the size or diameter of the illuminated area can be adapted to the number of detected people. These parameters will be adjusted for the luminaire 100 in the corridor rather than for the luminaire 100 in the desk area.

In fig. 3C, another example for an illumination pattern of a corridor area is shown. Again, the time for light activation, i.e. the interval between light activation start time ti and light activation end time t ₂ , and for light deactivation, i.e. the time interval between the light deactivation start time t ₃ and the light deactivation end time t ₄ , are relatively short. Moreover, in this example, an additional intensity level I _m i _n is defined for situations, when no activity is sensed for a predefined time period. Thus, after a first detection of motion, the light activation is immediately started at ti. Then, illumination of the intensity level I _max is provided for a predetermined time interval (from t ₂ until t _a ). If during this time interval, no further motion is detected, i.e. when no further passenger walks by, the illumination is switched at time tb to a standby illumination with a lower illumination level I _m i _n . At time t _c , new motion is detected leading to re-activation of the illumination unit 110 to the higher illumination level I _max for the predefined time interval (from td until t ₃ , possibly being identical to the interval from t ₂ until t _a ). Then again, if no motion is detected, the illumination is switched to the standby illumination. In case that also in a second predefined time interval for the illumination with the lower intensity I _m i _n , no motion is detected, the deactivation is started at time t ₃ . Alternatively, deactivation starts, if the illumination unit 110 is manually deactivated.

The illumination rules or presets can be set a priori or can be adapted after installation, e.g. by facility management or by the users themselves. Moreover, the control unit 130 can also consider other additional parameters for adapting the operation

characteristics of the luminaire 100, e.g. current ambient brightness, current daytime, current user control input or a history of user control inputs. For this, the luminaire 100 can comprise further sensors, such as a brightness sensor or the like.

In figs. 4A and 4B, histograms for the detected speed of motion are shown for the luminaire 100 installed in the desk area and the luminaire 100 installed in the corridor, respectively. As shown in fig. 4A, the luminaire 100 in the desk area will mainly detect motion with low speed, whereas the luminaire 100 installed in the corridor will mainly detect motion with high speed (shown in fig. 4B).

For generating these histograms, the detected speed of motion is quantized into a number of bins, e.g. low velocity, middle velocity and high velocity. The size or number of bins may be adjustable or set during installation. The histograms comprise speed data of the resent past, i.e. recorded during the predefined data recording interval. New detected speed data replace old data in the histogram, so that the histogram is permanently updated. This can be achieved, for instance, using a sliding window algorithm, wherein a window having a predefined width, e.g. corresponding to the predefined data recording interval, is slid over the data recorded over time and only data within the window are considered for the histogram. Based on the histogram shape, the control unit 130 of the luminaire 100 can determine a typical activity pattern of the surroundings of the

luminaire 100. Instead of histograms, the control unit 130 may also use any statistical parameter determined from the detected activity data for determining the typical activity pattern in the surroundings. In another embodiment, the control unit 130 fits the raw activity data, preprocessed activity data or the histogram with a predefined fit function in order to determine the typical activity pattern. Based on the determined activity pattern, the control unit 130 adapts the operation characteristics of the luminaire 100 for obtaining a functional and comfortable illumination behavior. For this, different sets of operation characteristics may be defined for a particular activity pattern, e.g. for a desk lamp, a corridor lamp, an entrance lamp, etc.

Alternatively, the operation characteristics can be directly determined from the histogram without prior classification of the area with activity patterns. Here, the definition of a mapping function from the histogram space to the operation characteristics space is required. This mapping function can relate to a simple step function, e.g. desk activity for an average velocity below a certain threshold and corridor activity for an average velocity above this threshold. Another example for a mapping function can be based on the frequency or counts of high and low velocity.

In a still further embodiment, the operation characteristics are determined from the raw or preprocessed activity data themselves without generation of histograms and the like. For instance, the operation characteristics of light intensity I can be set as depending on the detected occupancy level (i.e. the number of people present in the vicinity) and on the average speed of motion. Thus, the mapping function for the light intensity I is a function of these data determined from the sensor data.

In another embodiment, the control unit 130 can differentiate between daytimes. In other words, the control unit 130 determines histograms or typical activity patterns for different daytimes, respectively. For instance, the activity next to the luminaire 100 in the morning may be different from the activity during lunchtime. Therefore, the control unit 130 is adapted to adjust the operation characteristics of the luminaire 100 based on the activity detected during a respective daytime.

From fig. 4, the difference between the proposed invention and a luminaire controlled by simple binary presence detection becomes obvious. In binary presence detection, no information about the motion of objects or people is provided. However, activity patterns can be distinguished for a corridor, where ten people pass by within one minute and stay only for a very limited time under the luminaire 100, and an office, where a person moves on his chair for 10 times in this minute. Therefore, in contrast to binary presence detection and subsequent illumination control, the present invention provides a luminaire 100 that can learn the typical activity in its surroundings and adapt its operation characteristics accordingly.

In fig. 5, a method for adapting the illumination of the surroundings based on the detected activity is illustrated. In step S510, activity data is sensed in the direct vicinity of the luminaire 100. These data are used for generating a history of activity (S520), wherein only activity data detected within the predefined data recording interval are used. Then, the activity data can be quantized in order to generate a histogram (S530). The history of activity data is updated continuously or with predefined time periods for new activity data (S540). Based on the history of activity, the operation characteristics of the luminaire 100 are adapted (S550). Then, the surroundings of the luminaire 100 can be illuminated based on the adjusted operation characteristics (S560).

Preferably, in addition to the detected activity, also further parameters may be considered for adjusting the operation characteristics of the luminaire 100. For instance, the current day time or ambient brightness can be used for enabling a more appropriate adaptation of the operation characteristics. Furthermore, installation settings or user control can be taken into account. Likewise, the operation characteristics of the luminaire 100 do not only include activation characteristics of the illumination unit 110 (see fig. 3), but can also relate to illumination properties, such as light color, light temperature, light effects, periodic light amplitude fluctuations, color changes, the direction of the light beam, the shape or size of the illuminated area and the like. In addition, the operation characteristics of the activity sensor unit 120 can be adjusted, e.g. the time, in which activity data are taken into account for the adaptation process, a bin size of histograms, a time of activation of the activity sensor unit 120 and the like.

Fig. 6 shows a simple block diagram of another embodiment of a system for lighting configuration according to the present invention. The system 600 is capable of learning how an environment such as a small office is used and configuring lighting in the environment, particularly adjusting the lighting settings accordingly. The system 600 may comprise at least one sensor or sensor unit 120 for observing an area of interest 14 (the office environment) and providing spatial information from the area of interest, a processing/control unit 130, for example a computer configured by software to process the spatial information provided by the sensor 120, and at least one controllable light unit or illumination unit 110. Thus, as shown in fig. 6, the sensor unit 120 may collect data for several illumination units 110, wherein the sensor unit 120 is adapted to sense data with spatial resolution, also referred to as spatial information. The control unit 130 may receive the spatial information from the sensor 120 via a wired or wireless communication connection, for example via a ZigBee™ connection. The illumination units 110 may be controlled by the control unit 130 also via a wired or wireless communication connection such as a ZigBee™ communication connection.

The sensor 120 of the system 600 is able to deliver spatial information from the area of interest 14, which in general corresponds to the surrounding area of one or more illumination units 110. Several types of sensors that can provide spatial information of the observed environment can be used. Examples include cameras, arrays of PIRs as described in "Video Scene Understanding Using Multi-scale Analysis", Yang, Y., Liu, J. and Shah, M. 2009, IEEE International Conference on Computer Vision, ultra-sound radar arrays, microphone arrays, thermopile arrays etc.

The sensor 120 collects measurements of the environment 14. The processing unit 130, using data mining techniques, automatically learns partitions of the observed space 14 that reflect the usage that is made of the environment. Based on this knowledge, the control unit 130 can adapt the type of illumination (shape, color, intensity etc.) created by the illumination units 110 according to the way the different areas are typically used.

In one preferred embodiment of the present invention, usage patterns are determined using statistical learning for a predefined time interval, collecting simple and robust features for a long time. These long-term observations allow building a high-level segmentation of the environment, which was found being tightly related to the activities carried out in the different zones. This allows for a robust and flexible lighting configuration system that is able to provide the right amount and quality of light given the situation, allowing energy saving (e.g. more light on the desk than on the corridor) and better fit to user's needs. The statistical learning can be done only once, after the system is installed, or can be updated (continuously, periodically or in certain occasions, e.g. when a room is refurbished) so that the behavior and usage changes can be automatically taken into account.

Fig. 7 shows in a top view a small office 14 as an example of an environment (area of interest), which can be observed with the system 600 according to this embodiment of the invention. In the left part of the office 14, two desks 143 with computers and chairs 144 for office workers are located. In the right part of the office 14, office furniture such as shelves 145 and the door 146 to the office are located. As spatial information providing sensor, a camera 147 is embedded in the ceiling, which is configured to capture images of the entire office environment and to transmit the captured images to the control unit of the system. After enough information is collected (typically few hours), the system 600 is able to "understand" which areas of the office 14 are used and how, and may adapt the light settings in the room accordingly. The "understanding" is performed by applying algorithms as will be described later in detail. For example, the system "recognizes" a desk area or partition 141 with slow motion (dashed-dotted box), and lights are consequently set to be for example intense and slowly reactive to changes. The system 600 also automatically "learns" the area or partition 142 with fast motion where people walk (dashed box) and illuminates it for example with reactive light only when users are detected there. The rest of the room is not used and thus illumination can be, for example, low and diffuse. These illumination "rules" can be set a priori or adapted, after the installation, by facility management or by the users themselves.

As outlined above, the system extracts context information from the environment, which can be used in two different ways:

Context information can be fed to presence detection systems to improve their performances, e.g. lowering detection threshold in areas where activity has typically occurred more frequently, as in the partitions 141 and 142 in Fig. 7.

Given the presence of a user (e.g. detected by the presence detection unit), the lighting system can activate the appropriate lighting according to the learned context information. This approximates a real-time activity recognition system, which is robust and simple. The robustness comes from the fact that the context is learnt from long-term statistics of reliable features, and the presence can be robustly detected using available solutions. The system is simple because it does not require the modeling of human activities and their reliable detection in unconstrained environments.

Besides, information extracted from long-term observations of the environment can be used for applications that are not directly related to lighting. For example, usage patterns can be used to provide statistics of people behavior in shops or public spaces for marketing purposes or security, or to automatically extract abnormal events in video surveillance systems.

For simplicity of deployment and installation, the following embodiment of the invention makes use of a vision sensor. Vision sensors provide accurate spatial information with only one sensor, and are already deployed in smart lamps to sense light levels and occupancy levels, as in the Philips Mini300 LED luminaire. The sensor can be embedded in a light fixture or can be installed as a separate module with the lighting infrastructure, for example in the ceiling.

Next, the algorithm for processing the images captured with a vision sensor according to the invention is described in detail with regard to Figs. 8-11, which show flowcharts of embodiments of the algorithm. The algorithm is implemented in the control unit 130 of the system 600. The control unit 130 may be for example implemented by a computer with interfaces for receiving spatial information from the sensor 120 and for controlling the illumination units 110. The interfaces may be for example implemented by ZigBee™ or any other modules suitable for communication in a lighting system. The algorithm may be implemented as part of a program for controlling and configuring a lighting system.

Fig. 8 shows in a flowchart the steps S10-S16 of an embodiment of the algorithm. In step S10, spatial information, namely images captured with a vision sensor from the area of interest 14 are received. In the following step S12, patterns in the area of interest 14 are learned from the received spatial information. In step S14, the area of interest 14 is then spatially segmented into two partitions 141 and 142 based on the statistically learned patterns. Finally, information for lighting configuration depending on the spatial segmentation is generated in step SI 6. The generated information can for example be displayed on a monitor coupled with the control unit 130, so that a user may configure the lighting system by using this information, or the generated information can configure the lighting in the area of interest 14 by controlling the illumination units 110. In an embodiment of steps S12 and S14 of the algorithm, the statistically learned patterns are analyzed for partitions 141, 142 of the area of interest 14 with different features (step S121 in Fig. 9) and the area of interest 14 is then spatially segmented depending on the analyzed partitions with different features (step S141 in Fig. 9).

An embodiment of the analysis in step S121 and of the spatially segmenting in step S14 is shown in Fig. 10:

In a step S1211, images of the area of interest 14 are divided into blocks, and then very simple video features for each block can be computed (in Fig. 7, the division of an image into blocks is shown by a grid over the image of the entire office 14). Features can include average motion detection such as slow or fast motion, direction of motion such for example from the door 146 to a desk 143, speed of motion, detection of a change such as for example the change of the location of a chair 144, color etc. The advantage of using such simple features is that they are very basic, making the system flexible and general, and easy and robust to compute.

Features can be collected for a long time (e.g. some hours or days) and histograms of these features are computed in step S1212. Histograms provide a very compact representation of data, while capturing the main statistical properties of the observed phenomena.

The computation of histograms is now described for an embodiment of the algorithm shown in Fig. 11, where an image is divided into blocks of 8x8 pixels and for each of them it is counted how many time there was slow motion, fast motion and change detected (step S12121 in Fig. 11). Motion fields can be computed with the 3DRS algorithm (Sub-pixel motion estimation with 3-D recursive search block-matching, de Haan, G. and Biezen, P. 6, 1994, Signal Processing: Image Communication, pp. 229-239.), while for change detection a simple thresholding of the filtered difference between consecutive frames can be used. For each image block a 3 -bin histogram for motion (no motion, slow motion, fast motion) and a 2-bin histogram for change detection (no change detected, change detected) can be built (steps S 12122 S 12123 in Fig. 11) in the following way:

if the average speed of motion > 1 and < 10 in the block centered at (x,y) then SlowMotion(x,y) = SlowMotion(x,y)+l

if the average speed of motion > 10 and < 20 in the block centered at (x,y) then FastMotion(x,y) = FastMotion(x,y)+l

if change detected in the block centered at (x,y) then ChangeDetection(x,y) = ChangeDetection(x,y)+ 1 SlowMotion, FastMotion and ChangeDetection can be visualized as 2D density maps and each point into these histograms represents one image block. Here, the size of these density maps is HightImage/8 WidthImage/8.

Referring again to Fig. 10, in step S1411 blocks with the same characteristics are grouped together. There are many data mining techniques that can serve to the purpose, in particular clustering methods such as hierarchical clustering (Hastie, T., Tibshirani, R. and Friedman, J. H. The Elements of Statistical Learning: Data Mining, Inference, and Prediction, s.l. : Springer, 2009; Hierarchical grouping to optimize an objective function. Ward, J. H. 301, 1963, Journal of the American Statistical Association, Vol. 58, pp. 236-244), diffusion maps (Diffusion maps. Coifman, R. R. and Lafon, S. 1, 2006, Applied and Computational Harmonic Analysis, Vol. 21, pp. 5-30), hierarchical Bayesian models (Latent dirichlet allocation. Blei, D. M., Ng, A. Y. and Jordan, M. I. 2003, Journal of Machine Learning Research, Vol. 3, pp. 993-1022; Hierarchical dirichlet process. Teh, Y. W., et al. 476, 2006, Journal of the American Statistical Association, Vol. 101, pp. 1566-1581). Herein, image blocks may be clustered based on the similarity of features' histograms using the minimal variance hierarchical clustering method (Hierarchical grouping to optimize an objective function. Ward, J. H. 301, 1963, Journal of the American Statistical Association, Vol. 58, pp. 236-244). Using hierarchical clustering, the number of clusters does not have to be specified in advance. Herein, the number of clusters can be automatically computed in such a way that the average solidity (area/convex area) of all connected regions in the labeled image is maximal. This is done to ensure spatial compactness of the clusters.

Using this procedure, an analysis of an office room such as the one shown in Fig. 7 found essentially 4 clusters. These clusters are shown in Fig. 7 and are denoted with reference numerals 147-150. The segmentation of the environment reflects the function of different regions: non-moving areas (147), walking areas (148), door/corridor (149) and desk areas (170). The spatial grouping (partitions 141 and 142) of similarly behaving image regions is obtained without using the information of blocks position in the image. This labeling of the observed scene is clearly related to the usage that is made of the room. This image may be referred to as activity map.

The activity map of the room can be used in different ways:

1. Fully automatic: Rules for predefined regions are set a priori. For example, in an office environment one might want to set 3 different regions:

a) Region where nothing happens Slow motion LOW, Fast motion LOW, Change detection LOW; b) Walking area Slow motion HIGH, Fast motion HIGH, Change detection LOW;

c) Desk area Slow motion HIGH, Fast motion LOW, Change detection HIGH;

Using these simple rules, the environment shown in Fig. 7 can be automatically subdivided into these pre-defined areas (partitions 141 and 142) automatically after few hours of observation. The 4 clusters in Fig. 7 can be reduced to 3 clusters, where the door 149 and the walking areas 148 are merged (they are both classified as walking area according to rule (b) above). The third cluster with the regions where nothing happens is not denoted and shown in Fig. 7. For each of these areas, a predefined light setting is stored in the system and is applied to the corresponding part of the room.

2. Semi-automatic: The clusters are areas where similar activities are carried out and that are described in a language (feature representation) that is understandable for the computer. Thus a user or facility manager can select one cluster (e.g. the desk area 170 in Fig. 7) and define some lighting setting for it (e.g. intense light for reading). Then all regions observed by the system that are similar in activity can be automatically set to the same lighting setting. In this way, the proposed method translates complex semantic concepts (area where people walk, area where people work etc.) into commands that are easily

understandable for the lighting system. A clear advantage of this approach is that, after the facility management has defined lighting for a certain activity cluster, any new zone behaving similarly will be automatically assigned to the same light setting.

3. Manual: Facility management understands, how the building is used and exploits this information to optimize the building layout and the illumination settings. For example, in a shop context this information can be used to check where people mostly walk and where they stop more frequently, and adapt lighting settings accordingly to attract the attention of customers to "neglected" areas.

Therefore, according to the present invention, a luminaire, a system and a method for smart reactive illumination are provided, wherein a sensor unit senses activity data in the surroundings of the luminaire (unit) or of an illumination unit, a control unit generates a history of the activity data sensed for a predetermined time and adjusts operation characteristics of luminaire (unit) or of the illumination unit based on the history of activity data for illuminating the surroundings. By these means, a luminaire, system and method for adaptive illumination are provided capable of flexible, autonomous and automatic adaptation of operation characteristics (e.g. how quickly illumination is turned on/off, length of activation time, maximum intensity, shape of light beam, etc) depending on typical speed patterns observed in the vicinity of the luminaire or luminaire unit for a certain amount of time. Thus, a surrounding area can be illuminated with increased energetic efficiency, higher flexibility, improved user convenience and operation comfort, the illumination being autonomously adapted according to changing requirements. The luminaires, systems and methods for adaptive illumination according to the present invention can be used for any professional or consumer illumination, e.g. in cafeterias, libraries, museums, office buildings, common spaces, corridors, and private homes in order to improve the lighting configuration, particularly to better adapt lighting to the activities in regions with different activities of an environment. The invention can also be used to assist users in lighting configuration, for example by means of computer executing a computer program implementing the present invention and processing spatial information of an environment collected over a long time, such as images captured with a video camera installed in the ceiling of shop from the activity in the shop over one or more days. Shop personnel and lighting designers can then obtain lighting configuration information as output of the program, which can be visualized and allow to recognize spatial segments or partitions of the environment with different activity.

At least some of the functionality of the invention may be performed by hard- or software. In case of an implementation in software, a single or multiple standard microprocessors or microcontrollers may be used to process a single or multiple algorithms implementing the invention.

It should be noted that the word "comprise" does not exclude other elements or steps, and that the word "a" or "an" does not exclude a plurality. Furthermore, any reference signs in the claims shall not be construed as limiting the scope of the invention.