JP2001008197A - Device for analyzing human action and recording medium for recording human action analytic program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To precisely analyze human action by automatically detecting the point under consideration of a person under an environment in which the position/direction of the person is arbitrarily changed. SOLUTION: Which region in a space can be observed by each line of sight detecting device is preliminarily stored in an observable region storing part 2, and the layout of space to be monitored and the position of an article to be investigated for judging whether or not the article is noticed by a person are written in a space map storing part 3, and the space to be monitored is divided according to the size of the observable range of the line of sight detecting device 1 so that all the lines of sight to each article to be investigated can be observed even when the person stands at any position and in any direction, and the plural line of sight detecting devices are arranged as necessary by removing any dead angle so that the lines of sight can be detected even when the person stands in any direction in each region.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for analyzing the behavior of a person and a medium on which a program is recorded. In particular, the present invention relates to a trajectory of a person in a space to be monitored and a person's movement at each point in the space. The present invention relates to a person behavior analysis device that automatically collects a target of interest and a recording medium that records a person behavior analysis program.

[0002]

2. Description of the Related Art Conventionally, this kind of human behavior analysis method is used to detect the gaze direction of a person and analyze the degree of attention or interest in goods, advertisements, etc. existing in a space to be monitored. ing.

[0003] Conventional techniques for detecting the direction of the line of sight of a person are roughly classified into two types: contact type and non-contact type.

[0004] As a contact-type conventional technique, Japanese Patent Laid-Open No.
A technique using an eye camera as described in JP-A-6-347866 is known. This eye camera is
In order to detect the direction of the user's line of sight, the eyeball is irradiated with near-infrared light and the reflected light is captured by a video camera to detect the direction of the line of sight. Because of this, the movement of the eyeball, that is, the gaze direction can be detected regardless of the movement of the head.

[0005] Further, as a non-contact type method, many methods have been proposed for estimating the line of sight of a person from an image picked up by an image pickup device such as a CCD camera. For example, a method of estimating the orientation of a face from the inclination of its principal axis using the fact that the face region is close to an ellipse was published at the 2nd International Symposium on Automatic Face and Gesture Recognition held in April 1998. Pp.83-93 (pp.88-93,
The 2nd International Symposium on Automatic Face
and Gesture Recognition, Nara, Apl. 1998)
hida) et al. “3D Head Pose Estimation Without Feature Tracking
(3D Head Pose Estimation without Feature Trackin
g) ”. Also, a method of estimating a gaze direction based on an apparent shape change of a positional relationship between feature points on a face such as an eye, a nose, and a mouth was disclosed in October 1998 by a workshop on application of computer vision by iTripleE. (Pp.256-257, IEEE workshop on Applications of C)
omputer Vision, Princeton, Oct. 1998) by George Stockman et al., "Real-time Tracking of Face Features and Gaze Direction Termination (Real-time Tracki).
ng of Face Features and Gaze Direction Determinati
on) ".

[0006] As described above, the method of detecting the direction of the line of sight of a person is roughly classified into a contact type and a non-contact type. However, it is necessary to analyze the degree of attention or interest in a product or advertisement existing in a certain space. In consideration of this, it is not preferable to use a contact-type method of attaching some kind of sensor to the observer because it places a burden on the observer. Therefore, it is preferable to use a non-contact gaze direction detection method as the human behavior analysis method of this kind. As a conventional technique of a person behavior analysis technique using a non-contact type gaze direction detection technique, a technique described in Japanese Patent Application Laid-Open No. H11-24603 is known.

FIG. 11 is a block diagram showing a configuration disclosed in Japanese Patent Application Laid-Open No. H11-24603.
Hereinafter, a conventional technique will be described with reference to FIG.

[0008] The presentation information storage section 14 stores an image (mainly a commercial) for analyzing the degree of attention and the degree of interest, and this image is displayed via the display section 18. next,
The camera 19 captures an image of a person watching the image displayed on the display unit 18, and the line-of-sight detection unit 12 detects the line-of-sight direction of the person using the above-described non-contact type line-of-sight detection method. The determination unit 20 calculates which part of the video displayed on the display unit 18 the gaze direction obtained from the gaze detection unit 12 corresponds to, and stores the calculation result in the collection information storage unit 16.

[0009]

SUMMARY OF THE INVENTION The first of the prior arts described above.
The problem is that if the observer moves away from the imaging device (camera) to some extent, it is extremely difficult to perform accurate gaze detection. The reason is that the gaze direction of the person is calculated based on the resolution of the imaging device, so that if a person is located at a certain distance from the imaging device, the correct gaze direction cannot be detected.

A second problem of the prior art is that the direction of the line of sight is not detected when the direction of the face of the observer is shifted with respect to the image pickup device, so how the observer moves. It is not always possible to analyze the behavior of the observer. The reason is that the conventional technology only detects the gaze direction and counts the point of interest in the gaze direction, and there is no means to know what action was taken for an observer who could not detect the gaze direction. It is.

Accordingly, an object of the present invention is to provide a human behavior analysis device which solves the above-mentioned problems of the prior art.

[0012]

According to the first aspect of the present invention, there is provided a human behavior analysis apparatus that covers a line of sight for each article to be investigated as much as possible, regardless of the position / orientation of a person in a space to be monitored. A plurality of spaces to be monitored are divided and arranged in accordance with the size of the observable range of the eye-gaze detecting device (1 in FIG. 1) so that observation is possible. It is desirable that this line-of-sight detection device is arranged as necessary so that the line-of-sight detection device does not have a blind spot so that the line of sight can be detected regardless of the direction of the person in each area. In the observable area storage unit (2 in FIG. 1), which area in the space each gaze detecting device can observe is stored, and in the spatial map storage unit (3 in FIG. 1), The layout of the space to be used and the position of the article to be checked whether or not the person is paying attention are described.

Next, a person position detecting section (4 in FIG. 1) detects where the person is in the space by using a surveillance camera / infrared sensor / pressure sensor / wireless, etc., and a device selecting section (FIG. 1). In 1-5), it is determined based on the information in the observable area storage unit which gaze detection device the person at the position detected by the person position detection unit can observe, and one or more necessary gaze detection devices are determined. Select multiple. The gaze determination unit (6 in FIG. 1) integrates the processing results of the gaze detection device selected previously to determine the final gaze direction of the person. The person-of-interest target detection unit (7 in FIG. 1) estimates, based on the information in the spatial map storage unit, the articles present in the line-of-sight direction detected by the line-of-sight determination unit from the person position detected by the person position detection unit. Then, the article is extracted as a person's attention target. The person attention target storage unit 8 (8 in FIG. 1) records the attention target detected by the person attention target detection unit along with the measurement time for each person.

Therefore, the first human behavior analysis device of the present invention detects the direction of the observer's line of sight using a plurality of line-of-sight detecting devices, and detects the position of the observer in the space to be observed. The gaze detection device corresponding to the detected position of the observer is selected, and the target of the observer is detected from the gaze direction information of the selected gaze detection device. The first problem that it becomes extremely difficult to accurately detect the line of sight when the sight line is separated to some extent.

[0015] The second person behavior analysis apparatus of the present invention comprises a first person behavior analysis apparatus.
In addition to the configuration of the person behavior analysis device of the above, a person movement line storage unit 9 that records the person position information obtained by the person position detection unit together with the measurement time is provided.

Therefore, the second person behavior analysis device of the present invention can analyze how the observer moves in the space based on the person position information. The second problem that it is not possible to know what action the observer has taken when the direction in which the observer is displaced with respect to the imaging device can be solved.

According to a third person behavior analysis device of the present invention,
Alternatively, the image selected by the device selection unit of the second person behavior analysis device is displayed on a monitor (10 in FIG. 3), and individual person images are stored in a person image storage unit (11 in FIG. 3). Can have the same role as the monitoring system.

[0018]

Next, an embodiment of the present invention will be described in detail with reference to the drawings.

[First Embodiment] First, the first embodiment of the present invention will be described.
DESCRIPTION OF THE PREFERRED EMBODIMENTS OF THE PERSON ANALYSIS APPARATUS (Description of Configuration) FIG. 1 is a block diagram showing an embodiment of the configuration of the first human behavior analysis apparatus of the present invention.

Referring to FIG. 1, a first person behavior analysis device according to the present invention comprises a gaze detection device 1, an observable area storage unit 2, a spatial map storage unit 3, a person position detection unit 4, and a device. Selection unit 5, gaze determination unit 6, person attention target detection unit 7
And a person attention target storage unit 8.

A plurality of eye-gaze detecting devices 1 are arranged in a space to be observed and detect the gaze direction of an observer (person). Further, it is preferable that the line-of-sight detection device 1 is arranged so that the line-of-sight direction can be detected regardless of the direction of the observer in the space to be monitored. In addition, the eye-gaze detecting device 1 may use the method of Yauchida or Stockman described in the related art as it is, or may use another device if there is another substitute.

The observable area storage unit 2 stores in which area of each gaze detection device 1 the gaze direction can be detected from the position information.

The space map storage unit 3 describes the layout of the space to be monitored and the position of the article to be checked whether the observer is paying attention.

The person position detector 4 detects the position of the observer in the space to be observed and outputs the position information. Specifically, it is only necessary that the observer can detect where in the space the observer is located using a known technique such as a monitoring camera, an infrared sensor, a pressure sensor, and a wireless communication.

The apparatus selecting section 5 refers to the observable area storage section 2 and detects the line of sight of the observer in accordance with the position of the observer detected by the person position detecting section 4. Select 1.

The line-of-sight determining unit 6 finally determines the line-of-sight direction of the observer with reference to the line-of-sight direction information detected by the line-of-sight detecting device 1 selected by the device selecting unit 5. In addition, the number of gaze detection devices 1 selected by the device selection unit 5 is not limited to one, and a plurality of cases may be considered. In this case, the gaze direction information may be integrated by, for example, averaging gaze direction information obtained from the plurality of gaze selection devices 1.

The person-of-interest target detecting section 7 refers to the spatial map storage section 3 and, based on the position of the person detected by the person position detecting section 4, identifies an article present in the line of sight detected by the line of sight determining section 6. Extract as a target of interest.

The person attention target storage unit 8 stores the attention object extracted by the person attention target detection unit 7.

(Explanation of Operation) Next, the operation in the present embodiment will be described based on a flowchart showing an embodiment of the operation of the first person behavior analysis apparatus shown in FIG.

As an example given here, an elongated space such as a passage as shown in FIG. 6 is used as a space to be monitored. In FIG. 6, Si (i = 1 to M) is a divided space to be monitored, and each of the divided spaces has a size that can be observed by the visual line detection device 1.
Ci1 to Ci4 (i = 1 to M) are eye-gaze detecting devices 1 arranged in four directions with respect to each Si, and the observer's face can be captured by any of the observing devices with respect to the observer existing in the Si. , The observer's gazing point can be estimated. However, this Ci, Si
Is only an example. Si may be automatically obtained from the observable range of Ci. Also, aj (j = 1
-N) indicate the position of each article to be investigated to see if it is noticed by the observer. Further, CM3 and CM4 in FIG.
, There is no aj (j = 1 to N) in the set of points of interest Tref that will be detected by the line of sight observed due to these. Such a line-of-sight detection device 1 does not need to be particularly arranged. Also,
The layout of the space to be monitored shown in FIG.
The information on the position of the article to be investigated whether the observer is paying attention is stored in the space map storage unit 3 and the line-of-sight detection device corresponding to the position information of each area Si (i = 1 to M) performs observation. It is assumed that information on the possible area is stored in the observable area storage unit 2 in advance.

In the space to be monitored as shown in FIG. 6, first, the person position detecting unit 4 allows the observer to use any existing technology such as a monitoring camera, an infrared sensor, a pressure sensor, and wireless communication to determine the position in the space. Is detected (step 1 in FIG. 5). When an infrared sensor, a pressure sensor, or a wireless communication is used, the sensors may be arranged so as to detect the location of a person inside each area Si. In this embodiment, a method of detecting the position of an observer in a space using a surveillance camera (not shown) arranged above the space will be described.

FIG. 4 is a block diagram showing a detailed configuration example of the person position detecting section 4 in this case.

The image input unit 41 is a camera (Hyper Omni Vision (see: Yamasawa, Yagi, Yanaida: “All for navigation of mobile robots”) such as a fish-eye camera and a wide-angle camera that can capture a wide range of space. Orientation sensor Hype
r Proposal of Omni Vision ”IEICE Transactions D-II, V
ol. J79-D-II, No. 5, pp. 698-707 (1996)) or the like, and inputs an image showing the entire space to be monitored.

The person detecting section 42 extracts a person image from the image obtained by the image input section 41. A method for extracting a human image is described in the second international symposium on automatic face and gesture recognition in April 1998, pp. 222-227.
(pp.222-227, The 2nd International Conference on Fa
ce and Gesture Recognition) Haritaogl
u) et al. “Double 4: Who? Fen? Fair? Fat?
A Real-time system for detecting
(W4: Who? When? Where? What? A Realtime System fo
r Detecting and Tracking People) '', but even if multiple people are overlapped by the frame, any method can be used as long as each person can be extracted separately based on the past movement history .

The image / spatial correspondence storage unit 43 describes where each pixel in the image obtained by the wide-angle camera corresponds in the real space. The information stored in the image / space correspondence storage unit 43 is a table showing the correspondence between each pixel P (u, v) of the wide-angle camera and the position P ′ (x, y) in the real space as shown in FIG. It is described in. When calculating the correspondence between each pixel of the wide-angle camera and the position in the real space, for example, if the layout of the facility is known, the position in the space of a feature point such as a corner on the floor can be known. Pixels indicating points are manually extracted from the image, and for the remaining pixels,
By performing estimation by linear interpolation or the like based on the position in the space of the pixel obtained earlier, a more accurate correspondence can be obtained. However, if the correspondence between P and P 'can be obtained analytically, an expression describing the correspondence may be obtained in advance without saving the data in a table. Note that P and P '
Is performed only for the point on the floor for convenience.

The person position estimating section 44 includes a person detecting section 42
In the image obtained by cutting out the region of the person in FIG. 7, a point in the region indicating the observer (indicated by S1, S2, and S3 in FIG. 7A) is closest to the center of the image (M1 in FIG. , M2, M
3) are extracted. And M1, M2, M
3 is estimated in the space based on the value of the table stored in the image / space correspondence storage unit 43,
The position of each person in the space (M′1, M′2,
M′3) is obtained. The reason why the points M1, M2, and M3 were estimated as the position of the person is that the IEEE Workshop on Visual Surv.
Indoor Monitoring Via the Collaboration Betw described in eillance (Jan. 1998, Bombay, India) Proceedings pp.2-9
een a Peripheral Sensor and a Foveal Sensor (Y.cui,
As described in (S. Samarasekera, Q. Huang, M. Greiffenhagen), when a person stands upright or walks, these points point to the feet and accurately grasp the position of the user This is because it is possible.

Next, the device selecting section 5 stores the observable area storage section 2 in which area of the area Si (i = 1 to M) the position detected by the person position detecting section 4 is located. Reference and check (step 2). If there is no observer in any of the regions, the process is terminated (step 7). If there is, the visual line detection device 1 (possible to detect the visual line direction of a person existing in that region) ( Ci1 to Ci4) are selected (step 3), and the gaze direction information obtained from the selected gaze detection device 1 is output to the gaze determination unit 6. still,
In this embodiment, the description is continued assuming that the estimated person position Pm exists in Si as shown in FIG.

Next, the line-of-sight determination unit 6 integrates information detected by all the line-of-sight detection devices 1 indicated by Ci1 to Ci4,
The gaze direction of the person is finally determined (step 4). This does not mean that all the gaze detection devices 1 can detect the gaze direction of the person, but only the gaze detection device 1 arranged at a position very close to the direction of the normal person (in the example of FIG. 8, This is because Ci1) can observe the direction of the line of sight of the person, and cannot be normally observed with the other line of sight detecting devices 1.

The line-of-sight determination unit 6 outputs the detection result of the line-of-sight detecting device 1 that has detected the line-of-sight direction as the final line-of-sight direction. If there are a plurality of gaze detection devices 1 that have been able to detect the gaze direction, the average or the one with a high degree of effectiveness is taken as the final gaze direction.

Next, in the person-of-interest target detecting section 7, an article existing in the line-of-sight direction detected by the line-of-sight determining section 6 from the person position Pm detected by the person position detecting section 4 (aj in the example of FIG. 8)
Is extracted as an observer's attention target with reference to the information in the spatial map storage unit 3 (step 5).

The person attention target storage unit 8 records the attention object detected by the person attention target detection unit 7 together with the measurement time for each person (step 6).

Here, FIG. 9 shows an example of information recorded in the person attention target storage section 8. In FIG. 9, the person ID is for identifying each person existing inside the monitoring area. In FIG. 9, 1, 2,... Are described, but anything may be used as long as it can be identified. The measurement time is indicated by (hour: minute: second), but the unit may be arbitrarily set as needed. In addition, in a case where data is totaled by day, data such as year and month may be included therein. The elapsed time describes the elapsed time since the person entered the monitoring area. This is also shown in (hours: minutes: seconds), but the unit may be freely set as necessary. Finally, the person's attention target shows the article of interest of the person at that time, and aj (j = 1 to N) in FIG.
Are described so that they can be distinguished from each other. However,
Depending on the time, the person may not be paying attention to any of the objects to be observed. In such a case, there is no such item, and NULL is stored in the computer.

Therefore, the first person behavior analysis device of the present invention divides the space to be monitored according to the size of the observable range of the gaze detection device, and in which direction the person faces in each region. Can also detect the line of sight.

[Second Embodiment] Next, a second embodiment of the present invention will be described.
FIG. 2 is a block diagram showing an embodiment of the configuration of the second person behavior analysis device of the present invention. The second person behavior analysis device of the present invention is different from the first person behavior analysis device of the present invention in that a person flow line storage unit 9 is added as a new component.

The person flow line storage unit 9 stores the person position information obtained from the person position detection unit 4 together with the measurement time when the person position detection unit 4 detects the position of the observer in the space. are doing.

Here, FIG. 10 shows an example showing the correspondence between the person position information stored in the person flow line storage unit 9 and the information stored in the person attention target storage unit 8. The person flow line storage unit 9 records the position of the person in the space (M′1, M′2, M′3 in FIG. 7B) obtained by the person position detection unit 4 together with the measurement time. Therefore, it is possible to establish a correspondence relationship with the information stored in the person attention target storage unit 8. The person position information stored in the person flow line storage unit 9 is described in an orthogonal coordinate system based on a specific position on the horizontal plane of the monitoring area. However, if another coordinate system such as a polar coordinate system is more convenient depending on the application, it may be used.

In this embodiment, the person attention target storage unit 8 and the person flow line storage unit 9 have different configurations. However, the same applies to the configuration in which the person attention target storage unit 8 stores person position information. The effect is obtained.

Therefore, in the second person behavior analysis device of the present invention, since it is possible to analyze how the observer has moved in the space by using the person position information, it is not possible to obtain the gaze direction information of the observer. However, it is possible to analyze how the observer behaved.

[Third Embodiment] Next, a third embodiment of the present invention will be described.
FIG. 3 is a block diagram illustrating an embodiment of the configuration of the third person behavior analysis device of the present invention. The third person behavior analysis device of the present invention includes a monitor 10 and a person image storage unit 11 as compared with the configuration of the first person behavior analysis device of the present invention.

The monitor 10 displays an image obtained from the visual line detection device 1 selected by the device selection section 5.

The person image storage unit 11 stores an image obtained from the visual axis detection device 1 selected by the device selection unit 5 into an existing V
It is recorded on a recording medium such as a TR.

Therefore, in the third person behavior analysis device of the present invention, it is possible to make the person behavior analysis device have the same function as the existing monitoring system.

In order to implement the human behavior analysis device of the present invention by a computer, the output of a device (such as an imaging device) having the functions of the gaze detection device 1 and the human position detection unit 4 is input to the computer, and A computer that realizes the functions of the device selection unit 5, the gaze determination unit 6, the person-of-interest target detection unit 7, the observable area storage unit 2, the space map storage unit 3, and the person-of-interest target storage unit 8 A program is created, and the computer program is recorded on a recording medium typified by a CD-ROM, a floppy disk, or a semiconductor memory, and the computer reads the recording medium on which the program is recorded, so that the computer By generating each function, the functions of the present invention can be realized by a computer. Further, the computer program may be, for example, in a form recorded in a recording device in the server, or in a form in which a program included in the server is provided via a network.

[0054]

According to the present invention, the space to be monitored is divided according to the size of the observable range of the gaze detecting device, and the gaze direction is detected regardless of the direction of the person in each region. With this configuration, it is possible to accurately detect where the observer is paying attention.

Further, the present invention uses the person position information to
Since the analysis is performed on how the observer moves in the space, the behavior of the observer can be analyzed without obtaining the gaze direction information.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an example of a configuration according to a first embodiment of this invention.

FIG. 2 is a block diagram illustrating an example of a configuration according to a second embodiment of the present invention.

FIG. 3 is a block diagram illustrating an example of a configuration according to a third embodiment of the present invention.

FIG. 4 is a block diagram showing details of processing performed by a person position detection unit 4 when a person position is detected by a monitoring camera.

FIG. 5 is a flowchart illustrating an example of an operation according to the first exemplary embodiment of the present invention.

FIG. 6 is a diagram illustrating contents of an observable area storage unit 2 and a space map storage unit 3;

FIG. 7 is a diagram for explaining processing of a person position detection unit 4 when a person position is detected by a monitoring camera.

FIG. 8 is a diagram for explaining the processing content of a person attention target detection unit 7 from a device selection unit 5;

FIG. 9 is a diagram showing the content of a person attention target storage unit 9 according to the first embodiment of the present invention.

FIG. 10 is a diagram showing contents when a person flow line storage unit 10 and a person attention target storage unit 9 are associated with each other in the second embodiment of the present invention.

FIG. 11 is a block diagram showing a configuration of a conventional technique related to the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 gaze detection device 2 observable area storage unit 3 spatial map storage unit 4 person position detection unit 5 device selection unit 6 gaze determination unit 7 person attention target detection unit 8 person attention target storage unit 9 person flow line storage unit 10 monitor 11 person Image storage unit

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G06F 15/74 320A

Claims (8)

[Claims] 1. A gaze direction corresponding to a detected position of an observer by detecting an observer's gaze direction using a plurality of gaze detection devices and detecting an observer's position in a space to be observed. A person behavior analysis device, wherein a detection device is selected, and a target of interest of an observer is detected from the gaze direction information of the selected gaze detection device. 2. A line of sight corresponding to the detected position of the observer by detecting the direction of the observer's line of sight using a plurality of line of sight detection devices and detecting the position of the observer in the space to be observed. Selecting a detection device, detecting an observer's attention target from the gaze direction information of the selected gaze detection device, and storing the detected attention target and the observer's position information. Behavior analysis device. 3. The method according to claim 1, further comprising detecting a direction of a line of sight of the observer using a plurality of image pickup devices and detecting a position of the observer in a space to be observed. Selecting a device, detecting an object of interest of an observer from gaze direction information detected using the selected imaging device, and outputting a target imaged by the selected imaging device. Analysis device. 4. A plurality of gaze detecting devices arranged in a space to be monitored and capable of detecting a gaze direction of an observer, and an observable region storing in which region of the space each gaze detecting device can observe. A storage unit, a space map storage unit that stores the layout of the space and position information of articles present in the space, a person position detection unit that detects where the observer is located in the space, Based on the position information detected by the person position detection unit, a device selection unit that selects an available line-of-sight detection device with reference to the observable region storage unit, and a line-of-sight detection device selected by the device selection unit detects A line-of-sight determining unit that determines the line-of-sight direction of the observer by referring to the obtained line-of-sight direction information, and an article present in the line-of-sight direction detected by the line-of-sight determining unit from the position information detected by the person position detecting unit. Determined by reference to 憶部, the personal behavior analysis apparatus the article characterized by having a a person focused target detector for detecting a target of interest of the observer. 5. The method according to claim 4, wherein the observer's attention target detected by the person attention target detection unit and the position information of the observer detected by the position detection unit are stored. The person behavior analysis device described in the above. 6. The gaze detection device detects gaze direction information using an imaging device, and further includes a unit that outputs a target imaged by the gaze detection device selected by the device selection unit. The human behavior analysis device according to claim 4 or 5, wherein: 7. An image input unit capable of inputting an image showing the entire space to be monitored, and a person image of an observer is extracted from the image obtained by the image input unit. Reference is made to a person detection unit, an image / space correspondence storage unit storing which position in the space each pixel in the image obtained by the image input unit corresponds to, and the image / space correspondence storage unit. By doing
The person according to claim 4, further comprising: a person position estimating unit that calculates a position in the space of the person image extracted by the person detecting unit. Behavior analysis device. 8. A gaze detection device which is arranged in a plurality of spaces to be monitored and is capable of detecting a gaze direction of an observer, and a person position detection unit for detecting at which position in the space the observer is present. Input to the computer, and the computer stores in the computer an observable area storage function of storing in which area of the space each line-of-sight detection device can observe, and a layout of the space and an article present in the space A space map storage function that stores the position information of the device, and a device selection function that selects an available line-of-sight detection device by referring to the observable area storage unit based on the position information detected by the person position detection unit. A line-of-sight determination function that integrates line-of-sight direction information detected by the line-of-sight detection device selected by the device selection function to determine the line-of-sight direction of the observer; A person who determines an article present in the line-of-sight direction detected by the line-of-sight determination function with reference to the information stored by the spatial map storage function based on the position information, and detects the article as a target of interest of the observer. A recording medium on which a human behavior analysis program for realizing an attention target detection function is recorded.