KR20080005265A - Pattern based occupancy sensing system and method - Google Patents

Abstract

A pattern based occupancy sensing system and method includes an image array (24) generating an image signal (42) in response to an image, a microcontroller unit (MCU) (26) responsive to the image signal 42 and generating a communication signal (46), and a communication interface (38) responsive to the communication signal (46) and generating a system signal (50). The MCU (26) processes the image signal (42) to calculate area occupancy by analyzing an object for a pattern, determining whether the pattern matches a classifier, and processing the object when the pattern matches the classifier to calculate area occupancy.

Description

Pattern-based Occupancy Detection System and Method {PATTERN BASED OCCUPANCY SENSING SYSTEM AND METHOD}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention generally relates to lighting control, and more particularly, to lighting control according to pattern-based occupancy detection.

To conserve energy, the lights should be turned off when no one is in the room. For security and safety, the light should be on when someone enters the room. To this end, lighting control for a lighting system often includes occupancy sensors for turning the lights on and off depending on who is in the room. Currently, analog detectors wired to lighting systems such as passive infrared (PIR) detectors or ultrasonic motion detectors are used to detect room occupancy.

Unfortunately, such analog sensors often cause problems with modern lighting systems that are part of building automation systems. Building automation systems are typically digital systems and make it difficult to include analog detectors without expensive adapters or dedicated interfaces. Analog detectors also consume relatively large amounts of power, waste power, and require wiring to the power supply. Wiring limits design flexibility and increases installation costs. In addition, analog detectors lack the desirable characteristics in modern lighting systems such as embedded programmable and occupant counting.

It would be desirable to have a pattern based occupancy detection system and method that overcomes the above drawbacks.

One aspect of the invention provides an image array that generates an image signal in response to an image, a microcontroller device (MCU) that responds to an image signal and generates a status signal and a communication that generates a communication signal in response to a status signal. It provides a pattern based occupancy detector with an interface. The MCU analyzes the object for the pattern, determines if the pattern matches the classifier, and processes the object to calculate the area occupancy when the pattern matches the classifier so that the area occupancy is calculated. Process the image signal.

Another aspect of the invention includes obtaining an image, detecting an object in the image, analyzing the object for a pattern, determining whether the pattern matches the classifier and the pattern matches the classifier. The method provides a state counting method comprising the step of processing an object to calculate an area occupation state.

Another aspect of the invention provides a means for obtaining an image, means for detecting an object in the image, means for analyzing the object for the pattern, means for determining whether the pattern matches the classifier and the pattern matches the classifier. And an occupancy counting system comprising means for processing the object to calculate an area occupancy state.

The above and other features and advantages of the present invention will become more apparent from the following detailed description of the preferred embodiments made in conjunction with the accompanying drawings. The detailed description and drawings are merely illustrative of the invention rather than limiting the scope of the invention, the scope of the invention being defined by the appended claims and their equivalents.

1 is a block diagram of a pattern based occupancy sensor made in accordance with the present invention.

2 is a flow chart for a counting method employing a pattern based occupancy sensor made in accordance with the present invention.

3A and 3B are schematic diagrams illustrating boundary arrangements and images, respectively, for a pattern based occupancy sensor made in accordance with the present invention.

4 is a flow chart for a boundary counting method employing a pattern based occupancy sensor made in accordance with the present invention.

5A and 5B are schematic diagrams respectively illustrating region placement and images for a pattern based occupancy sensor made in accordance with the present invention.

6A and 6B are flowcharts respectively illustrating a method of counting an area and an occupancy state using a pattern-based occupancy state sensor made in accordance with the present invention.

1 is a block diagram of a pattern based occupancy sensor made in accordance with the present invention. The pattern-based occupancy sensor 20 includes a lens 22, an image array 24, a microcontroller unit (MCU) 26 and a communication interface 38.

Pattern based occupancy sensor 20 receives light 40 refracted from lens 22 as an image on image array 24. Image array 24 generates image signal 42 in response to the image. Microcontroller unit (MCU) 26, which includes an image processor 32, a memory 36, and a communication stack 34, receives an image signal 42 for processing. In accordance with the instructions stored in the memory 36, the MCU 26 compares the pattern of the object detected in the image signal 42 with the classifiers stored in the memory 36 to see if the pattern matches the classifier and the object is counting. Determine if it is of interest. The MCU 26 generates a state signal 46 that includes occupancy state information in response to the image signal 42. Occupancy status information may indicate whether an area is occupied and / or the number of objects of interest in the area. State information is passed through the communication stack 34 and converted into communication packets according to the communication protocol. Status information including the occupation status information is included in the status signal 46.

The communication interface 38 includes a radio frequency (RF) transceiver 28 and an antenna 30. RF transceiver 28 receives status signal 46 and provides modulated signal 48 to antenna 30, which communicates via pattern 50 with other pattern-based occupancy sensors, local lighting controls, other Communicate with local controls and / or building automation systems. The power supply 40 supplies power to the image array 24, the MCU 26, and the RF transceiver 28. In one embodiment, MCU 26 provides power control signal 52 to power supply 40.

Those skilled in the art will appreciate that the signals of the pattern-based occupancy sensor 20 may be bidirectional signals that communicate information between components coupled with them. For example, the pattern based occupation status detector 20 sends a communication signal 50 to other pattern based occupation status detectors, local lighting controls, other local controls and / or building automation systems, and the pattern based occupation status detector 20. ) Receives other communication signals from other pattern-based occupancy sensors, local lighting controls, other local controls and / or building automation systems. The communication signal may include data and instructions such as occupancy status information, operation instructions, queries, programming instructions, and the like. Image signal 42 and status signal 46 may also be bidirectional signals, and image signal 42 transfers data and instructions between image array 24 and MCU 26, and status signal 46 is MCU ( Data and commands are passed between 26 and the communication interface 38.

Lens 22 may be a lens such as a plastic lens or a glass lens suitable for viewing an object and presenting an image on image array 24. In one embodiment, lens 22 may be a lens used in a mobile phone camera. Those skilled in the art will appreciate that features of the lens, such as focal length, aperture, and magnification, may be selected to suit the particular application in which the pattern based occupancy sensor 20 is used.

Image array 24 may be any detector suitable for converting an image received from lens 22 into image signal 42 while maintaining low power consumption. In one embodiment, image array 24 is a complementary metal oxide semiconductor (CMOS) image sensing chip. In another embodiment, image array 24 is a charge coupled device (CCD). The image array 24 detects the image on the pixel array and converts the sensed image into an image signal 42. Typically, image array 24 is a low power image array. Image array 24 may include on-chip functions such as calibration, region selection, decimation, power management, pixel correction, and color correction. In one embodiment, image array 24 includes a lens 22. In other embodiments, image array 24 may include MCU 26 and / or RF transceiver 28. One example of a suitable image array is the VV6501 VGA CMOS color image detector manufactured by STMicroelectronics of Geneva, Switzerland.

Microcontroller device (MCU) 26 may be any microcontroller suitable for storing and processing instructions and data. MCU 26 analyzes the object for the pattern, determines if the pattern matches the classifier, and processes the object to calculate the area occupancy when the pattern matches the classifier so that the image occupancy state is calculated. 42). Typically, MCU 26 includes an image processor 32, memory 36, and communication stack 34. Instructions may be programmed into the memory 36 during manufacturing of the pattern based occupancy sensor 20 and may be programmed via the communication signal 50 during operation. Special features of the MCU 26, such as n-bit architecture, clock speed and memory size, can be selected for special applications. Examples of suitable microcontrollers include the 8-bit HCS08 and 16-bit HCS12 series manufactured by Freescale Semiconductor, Inc., Austin, Texas, Atmel Corporation, San Jose, California AVR 8-bit RISC flash microcontroller manufactured by Atmel Corporation and STV0767 imaging digital signal processor manufactured by Estima Microelectronics, Geneva, Switzerland. Those skilled in the art will appreciate that the MCU 26 may be a single chip or may be combined on a single chip with one or both of the image array 24 and the RF transceiver 28. In addition to performing image processing, in various embodiments, the MCU 26 may manage the communication that the communication pattern based occupation status detector 20 sends and receives over the communication interface 38 and the pattern based occupation status detector 20. Manage power usage at

The radio frequency (RF) transceiver 28 may be any transceiver for communication between the pattern based occupancy sensor 20 and other pattern based occupancy sensors, local lighting controls, other local controls, and / or building automation systems. Typically, the RF transceiver 28 operates at low voltage at low power consumption and may include power management features. In one embodiment, the RF transceiver 28 communicates at 2.4 GHz in accordance with the IEEE 802.15.4 near field wireless standard and the ZigBee networking standard protocol. In another embodiment, the RF transceiver 28 communicates at 15 GHz. Examples of suitable transceivers are MC 13193 short range, low power, 2.4 GHz ISM band transceivers manufactured by Freescale Semiconductor Inc., Austin, Texas, manufactured by Ember Corporation, Boston, MA. EM2420 transceiver and CC2420 RF transceiver manufactured by Chipcon AS of Oslo, Norway. Those skilled in the art will appreciate that the RF transceiver 28 can operate in a variety of frequencies and various protocols as desired for a particular application. The RF transceiver 28 may be a single chip or may be combined on a single chip with one or both of the image array 24 and the MCU 26.

The power supply 40 may be any wired or wireless power supply suitable for powering the components of the pattern based occupancy sensor 20. The wireless power supply may include a scavenging power supply, such as a battery or a power supply for scavenging solar, vibration, electromagnetic energy, and the like. Vibration scavenging power supplies can be fabricated using microelectromechanical devices (MEMS) and nanotechnology. In one embodiment, MCU 26 provides power supply 40 with a power control signal 52 for managing power consumption in pattern based occupation status sensor 20. In another embodiment, the entity components of the pattern based occupancy sensor 20 include power management features.

2 is a flow chart for a counting method employing a pattern based occupancy sensor made in accordance with the present invention. The counting method acquires an image at block 100, detects an object in the image at block 102, analyzes the object for a pattern at block 104, and determines whether the pattern matches the classifier at block 106. And, at block 108, processing the object to calculate the area occupancy when the pattern matches the classifier. In one embodiment, the method further includes obtaining training images and generating a classifier from the training images. Area occupancy can be sent to other pattern-based occupancy sensors and / or building automation systems for emergency call, safety, facility control, lighting control, and the like.

3A and 3B are schematic diagrams illustrating boundary arrangements and images, respectively, for a pattern based occupancy sensor made in accordance with the present invention. 3A shows an example of the boundary arrangement of a pattern based occupancy sensor. 3B shows an example of the image detected by the pattern based occupancy sensor.

Referring to FIG. 3A, a pattern based occupancy sensor 120 is installed near the doorway 122 to observe the boundary 124 between the first area 126 and the second area 128. Objects with a pattern such as a person, animal, machine or part, etc. move across the boundary 124 as shown by arrow 132. In this example, the object 130 is a human and the pattern based occupation sensor 120 is installed near the top of the doorway 122 to observe the head and shoulder patterns of the object 130. In other embodiments, the pattern based occupancy sensor 120 may be installed to suit a particular application. In an alternative embodiment, the pattern based occupancy sensor 120 is installed below the doorway 122 to observe a human leg or hip pattern. In another embodiment, a plurality of pattern based occupancy sensors 120 may be installed to handle a plurality of directions, ie, looking down and crossing the doorway 122. Multiple pattern-based occupational state detectors 120 may communicate with each other or with a building automation system and may cross-check the region occupational state determined by each pattern-based occupational state detector 120.

3B, the image 150 is divided into a first region 152 and a second region 154 by an image boundary 156. The first region 152 and the second region 154 correspond to the first region 126 and the second region 128 of FIG. 3A, respectively. Image boundary 156 may be any continuous curve dividing image 150 into two regions. The pattern 158 is a pattern corresponding to the object 130 of FIG. 3A shown in the first region 152. In this example, pattern 158 is a human head and shoulder pattern.

4 is a flow chart for a boundary counting method employing a pattern based occupancy sensor made in accordance with the present invention. The boundary counting method initializes a pattern-based occupancy detector at block 400, obtains an image with first and second regions at block 402, detects an object within the image at block 404, and The object is analyzed for a pattern at 406, and a block 408 is determined if the pattern matches the classifier. When the pattern does not match the classifier, the method returns to acquiring an image at block 402. When the pattern matches the classifier, the method continues to identify the object as being in one of the first area and the second area at block 410, obtain a subsequent image at block 412, and block 414. ) Determines whether there is an object in the other of the first region and the second region in the subsequent image. If it is not in the other of the first region and the second region within the subject second image, the method returns to step 402 of obtaining an image. When the object is in the other of the first area and the second area within the second image, the method continues to change the occupancy counter at block 416. The method continues to step 402 to acquire an image. The time for subsequent acquisition of the image at block 402 may vary depending on the expected velocity of the object across the image. In one embodiment, image acquisition at 402 may be repeated approximately hundreds of milliseconds, such as about every 100 milliseconds.

Initializing the pattern-based occupancy sensor in step 400 initializes a counter, defines boundaries and image regions, obtains classifiers by downloading and / or obtains classifiers by training. It may include. Counters, such as occupancy counters, can be set to zero to start a counting session. Image boundaries and regions can be automatically generated or manually selected from natural features in the sample image, such as floor color changes or entrance edges. Image boundaries and regions remain constant for the pattern based occupancy detector per image. If a library of classifiers is available, at least one pattern checking classifier, such as the head and shoulder classifiers, may be downloaded from the library to the pattern based occupancy sensor. If a library of classifiers is not available or a custom classifier is required for a particular application, the classifier can be obtained by training a pattern-based occupancy sensor. Training includes obtaining training images, extracting a feature from objects of interest in the training image, and generating a classifier from the features. Classifiers can include shapes as key features, regardless of size. Other embodiments may include color or grayness as a feature for determining the classifier. Classifiers may be individual classifiers that identify particular individuals or group classifiers that identify different groups such as people and dogs.

Analyzing the object for the pattern at block 406 is similar to generating classifiers. That is, the characteristics of the objects in the image are determined and analyzed to determine the pattern for the object, such as the head and shoulder patterns. In determining whether the pattern matches the classifier at block 408, a tolerance may be applied to determine that a match has occurred as long as the pattern falls within the tolerance.

At block 410, the object is identified as being in one of the first and second regions, the subsequent image is obtained at block 412, and the remaining of the first and second regions in the subsequent image at block 414. Determining whether there is an object in one checks to see if the matched object moves across the boundary between one region and the other region between one image and subsequent images. The movement can be any movement from the first region to the second region or from the second region to the first region, ie across a boundary that requires a change to the occupancy counter.

The Occupancy Counter can record matched objects in any area, such as border crossings or rooms. Regarding boundary crossings, the occupancy counter increments each time the matched object moves from the first area to the second area or from the second area to the first area. As for the matched object in the area, the occupancy state counter includes an IN counter and an OUT counter. The Occupancy Counter increments the IN counter if the subject is in the first region in the first image and the second region in the second image, and the OUT counter if the subject is in the second region in the first image and the first region in the second image. Change by increment. Area occupancy, such as the number of people in a room, can be calculated from the difference between the IN and OUT counters. Boundary crossings and / or area occupancy can be sent to other pattern based occupancy sensors and / or building automation systems for emergency call, safety, facility control, lighting control, and the like.

Classifiers may be individual classifiers that identify particular individuals or may be group classifiers that identify different groups such as people and dogs. Individual classifiers can be used to track individuals from one area to another. Group classifiers can be used to track the number of group crossings from one region to another. In another embodiment, group classifiers can be used to track an entity by assigning an entity tag to an identified object of the group and tracking that entity tag from one area to another.

5A and 5B are schematic diagrams respectively illustrating region placement and images for pattern based occupancy sensors made in accordance with the present invention. 5A shows an example of area placement of a pattern based occupancy sensor. 5B shows an example of the images detected by the pattern based occupancy sensors.

Referring to FIG. 5A, pattern-based occupancy sensors 220, 221, 222, and 223 face a layer 240 that includes a first region 226, a second region 228, a third region 230, and a fourth region 232. It is installed near the ceiling of the room 242. The pattern-based occupancy sensors 220, 221, 222, and 223 observe the first region 226, the second region 228, the third region 230, and the fourth region 232, respectively. Each pattern-based occupancy sensor observes a single area. Objects with a pattern, such as a person, animal, machine, or part can be in the room 242. In this example, the subjects 210 are human and the subject 212 is an animal. Pattern-based occupancy sensors 221, 222 observe the head and shoulder patterns of objects 210 in second area 228, and pattern-based occupancy sensor 222 detects object 210 in third area 230. Observing the head and shoulder patterns, the pattern based occupancy sensor 223 observes the body pattern of the object 212 in the fourth region 232. In other embodiments, pattern-based occupancy sensors may be installed to suit a particular application. In an alternative embodiment, pattern based occupancy sensors may be installed below the wall to observe the human leg or hip pattern or the side of the animal. In another embodiment, multiple pattern-based occupational state detectors 120 may be installed to cover multiple directions. Multiple pattern based occupation status detectors can communicate with each other or with the building automation system and can cross-check the area occupation status determined by each pattern based occupation status detector.

Referring to FIG. 5B, the room image 250 is divided into a first image 266, a second image 268, a third image 270, and a fourth image 272. Each pattern-based occupancy sensor has a single image. The first image 266, the second image 268, the third image 270, and the fourth image 272 include the first region 226, the second region 228, and the third region 230 of FIG. 5A. ) And the fourth region 232, respectively. The patterns 280 correspond to the objects 210 of FIG. 3A, and in this example, the patterns 280 are human head and shoulder patterns. Pattern 282 corresponds to object 212 in FIG. 5A, and in this example, pattern 282 is the back pattern of an animal.

6A and 6B are flowcharts illustrating a method for counting areas and occupying states employing pattern-based occupancy sensors made in accordance with the present invention, respectively. Occupancy checking method may optionally be included in the area counting method to check and correct the area occupancy state.

Referring to FIG. 6A, the area counting method initializes a pattern-based occupancy detector at block 600, obtains a background image at block 602, and calculates target space P0 and target ranges RL to RU at block 604. Set the occupancy counter N = 0 at block 606, obtain a target image at block 608, count the counting target M in the target image at block 610, and target image at block 612. Detecting the object i in the genus, analyzing the object i for the pattern i in block 614, and determining in block 616 whether the pattern i matches the classifier. When the pattern i does not match the classifier, the method returns to step 612 of detecting the object i in the object image. When the pattern i matches the classifier, the method continues to divide the area occupancy state counter N at block 618 and determines whether object i = Mth object at block 620. If the object i is not the Mth object, the method returns to step 612 of detecting the object i in the object image. If the object i is the M-th object, the method continues to examine the area occupancy state at block 622 and determines at block 624 whether time = background update time. If the time at block 624 is a background update time, the method returns to step 602 of obtaining a background image. If the time at block 624 is not a background update time, the method continues to determine if time = watch time at block 626. If the time is a monitoring time, the method returns to step 606 where the occupancy state counter N = 0 is set.

Initializing the pattern based occupancy sensor at block 600 may include initializing counters, obtaining classifiers by downloading, or obtaining classifiers by training. Counters, such as occupancy counters, can be set to zero to start a counting session. When a library of classifiers is available, at least one pattern checking classifier, such as the head and shoulder classifiers, may be downloaded from the library to the pattern based occupancy sensor. When a library of classifiers is not available or when a custom classifier is required for a particular application, the classifier can be obtained by training a pattern based occupancy sensor. Training includes obtaining training images, extracting a feature from the object of interest in the training image, and generating a classifier from the features. Classifiers can include shapes as main features, regardless of size. Another embodiment may include color or grayscale as a feature for determining the classifier.

Obtaining the background image at block 602 and calculating the target space P0 and the target ranges RL to RU at block 604 are optional and may be used in the selective occupancy check method 622. Acquiring a background image at block 602 includes acquiring a background image where no objects exist. The background image provides a criterion for determining the accuracy of the displayed area occupancy. The background image is also used in step 604 of calculating the object space P0 and the object ranges RL through RU. The object space P0 is the space of pixels changed by the shape of the object of interest. The object space P0 can be determined from measuring pixel changes from sample objects of interest for the background image and averaging the measured pixel changes for multiple sample objects. The target range, that is, the lower range RL to the upper range RU, may be a band around the target space P0, such as ± 10% of the target space P0 value.

Objects in the object image are counted and the objects of interest that match the classifier are counted by incrementing the area occupancy counter between blocks 606 and 620. Examining the area occupancy indicated by the area occupancy counter may be checked at block 622. Examining the area occupancy state indicated by the area occupancy counter may indicate a change in light and shadow around an error, such as counting an object that is similar to the object of interest but not of interest. In alternative embodiments, checking the area occupancy at block 622 may be omitted. In block 624 the background update time is selected to be approximately a few hours to update the background image as required and depending on the particular installation. The monitoring time in step 626 is chosen to be approximately a few minutes to update the target image as required and depending on the particular installation. Area occupancy can be sent to other pattern-based occupancy sensors and / or building automation systems for emergency call, safety, facility control, lighting control, and the like. In one embodiment, area occupancy states can be added from multiple pattern based occupancy sensors to calculate the occupancy state for a larger area. For example, area occupancy from multiple pattern-based occupancy sensors responsible for specific portions of the room may be added to calculate occupancy for the room.

6B, the selective occupancy check method 622 begins at block 700, calculates a pixel change P between the target image and the background image at block 702, and N * RL < RTI ID = 0.0 > Determining whether P ≦ N * RU. If N * RL ≦ P ≦ N * RU, then NFinal = N at block 706, and the method ends at block 712. If N * RL ≤ P ≤ N * RU, the method continues to set NCalc = Int (P / P0) at block 708 and NFinal = Min (N, NCalc) at block 710. Set, and the method ends at block 702.

Computing the pixel change P between the target image and the background image at block 702 uses the target space P0 and the target ranges RL to RU calculated at block 604. The pixel change P is the number of pixels changed from the background image to the target image. The target range for the displayed area occupancy state is between the displayed low range RL pixels per area occupied state N times the object and the upper range RU pixels per displayed area occupancy state N times the object. If the pixel change is within the target range for the displayed area occupancy state, the displayed area occupancy state is acceptable, and the final area occupancy state NFinal is taken as the displayed area occupancy state N. Otherwise, the displayed area occupancy state can be corrected, and the final area occupancy state NFinal is set to the lesser of the displayed area occupancy state N and an integer value obtained by dividing the pixel change P by the object space P0.

Although the embodiments of the invention described herein are given in view of the good, various changes and modifications may be made without departing from the scope of the invention. It is intended that the scope of the invention be expressed by the claims appended hereto, and that such modifications come within the meaning and range of equivalency thereof.

Claims (24)

In the pattern based occupancy sensor, An image array 24 for generating an image signal 42 in response to the image; A microcontroller device (MCU) 26 responsive to the image signal 42 and generating a status signal 46; A communication interface 38 responsive to said status signal 46 and generating a communication signal 50, The MCU 26 analyzes an object for a pattern, determines whether the pattern matches a classifier, and processes the object to calculate the area occupancy state when the pattern matches the classifier, thereby occupying the area. A pattern based occupancy sensor which processes the image signal (42) such that a state is calculated. The method of claim 1, And the image array (24) is selected from the group consisting of complementary metal oxide semiconductor (CMOS) image sensing chips and charge coupled devices (CCDs). The method of claim 1, The MCU (26) has an image processor (32), a memory (36) and a communication stack (34). The method of claim 1, The communication interface 38 has a radio frequency (RF) transceiver 28 and an antenna 30, the RF transceiver 28 generating a modulated signal 48 in response to the status signal 46. An antenna (30) is a pattern based occupancy sensor for generating a communication signal (50) in response to a modulated signal (48). The method of claim 1, And said system signal (50) communicates at 2.4 GHz in accordance with the IEEE 802.15.4 near field wireless standard and the ZigBee networking standard protocol. The method of claim 1, Further comprising a power supply 40 for supplying power to the image array 24, the power supply 40 is a pattern-based occupancy state detector selected from the group consisting of a battery and a scavenging power supply . The method of claim 5, The scavenging power supply unit pattern scavenging power from a source selected from the group consisting of solar energy, vibration energy, electromagnetic field energy. The method of claim 5, The power supply 40 is a pattern-based occupation state detector in response to the power control signal (52) from the MCU (26). In the state counting method, Acquiring an image (100), Detecting 102 an object in the image; Analyzing the object for a pattern (104), Determining 106 if the pattern matches a classifier, and And processing (108) an object to calculate an area occupancy state when the pattern matches the classifier. The method of claim 8, Obtaining training images, Generating the classifier from the training images. The method of claim 8, Each image comprises a first region and a second region, and the processing step includes Identifying (410) the object as being in one of the first region and the second region, Acquiring a second image (412); Determining (414) whether the object is in the other of the first area and the second area in the second image; And occupying (416) the occupancy state counter when the subject is in the remaining one of the first region and the second region. The method of claim 10, The occupancy state counter includes an IN counter and an OUT counter, and the modifying step increments the IN counter if the object is in the first area of the first image and in the second area of the second image. And incrementing the OUT counter if the object is in the second area of the first image and in the first area of the second image. The method of claim 11, And determining a difference between the IN counter and the OUT counter to calculate an area occupancy state. The method of claim 10, And wherein the classifier is an individual classifier. The method of claim 10, And the classifier is a group classifier, further comprising assigning an entity tag to the identified subject. The method of claim 8, The detecting step includes detecting each object in the image, Said analyzing step comprises analyzing each said object for a pattern, The determining step includes determining whether each of the patterns match the classifier, And said processing step increments an area occupancy counter each time each of the patterns matches the classifier. The method of claim 15, And checking an area occupancy state indicated by the area occupancy state counter. The method of claim 16, The image is a target image, the inspection step, Obtaining a background image, Calculating an object space and an object range from sample objects for the background image; Calculating a pixel change between the target image and the background image; Determining whether the pixel change is within the target range for the displayed area occupancy state; Allowing the displayed area occupation state when the pixel change is within the target range for the displayed area occupation state, and When the pixel change is not within the target range for the displayed area occupancy state, designating the area occupancy state as the lesser of the displayed area occupancy state and the integer value of the pixel change divided by the target space. Occupancy status counting method comprising a. In an occupied counting system, Means for obtaining an image, Means for detecting an object in the image; Means for analyzing the subject for a pattern, Means for determining if the pattern matches a classifier; And means for processing the subject to calculate an area occupation when the pattern matches the classifier. The method of claim 18, Means for obtaining training images, Occupancy counting system further comprising means for generating the classifier from the training images. The method of claim 18, Occupancy counting system further comprising means for powering the means for obtaining the image. The method of claim 18, The image comprises a first region and a second region, The processing means, Means for identifying the object as being in one of the first region and the second region; Means for obtaining a second image, Means for determining whether the object is in one of the first area and the second area of the second image, and And means for changing an occupancy state counter when the object is in the other one of the first region and the second region. The method of claim 18, The means for detecting comprises means for detecting each object in the image, Said analyzing means comprises means for analyzing each said object for a pattern, The determining means comprises means for determining whether each of the patterns matches a classifier, And the processing means includes means for incrementing an area occupancy counter each time each of the patterns matches a classifier. The method of claim 22, And means for checking an area occupancy state indicated by the area occupancy state counter.

Images