JPH10334207A - Human stream measuring instrument - Google Patents

Abstract

PROBLEM TO BE SOLVED: To actualize the human stream measuring instrument which can precisely measure the number and moving directions of pedestrians and the degree of congestion even at the time of congestion. SOLUTION: Image pickup devices 103 installed at a staircase at some elevation angle pick up images of pedestrians 102 who go up and down the stairs 101. The images are inputted to a correspondence part 105 to obtain parallax data. The parallax data are converted by a parallax-distance conversion part 106 into distance data. A recognition part 107 separate the pedestrians individually by using the distance data to recognize their positions. A counting part 110 traces the recognized positions of the pedestrians with the time and counts them when they pass a count line. The pedestrians are recognized and traced by using the stereoscopic images having a small dead angle, so the number and moving directions of the pedestrians and the degree of congestion can accurately be measured even in the case of congestion.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a human flow measuring device used for measuring the number of facility users, and more particularly to a human flow measuring device using stereo image processing.

[0002]

2. Description of the Related Art First, a stereo image measurement technique will be described with reference to FIG. 6 as a technique for measuring parallax data using a plurality of images obtained by a plurality of imaging devices. In the stereo image measurement technique, the same object is photographed by two left and right imaging devices, and the distance to the object point p is measured based on the principle of triangulation. FIG. 6 shows the imaging planes 201 and 202 of two imaging apparatuses whose optical axes are arranged horizontally and in parallel, and an object point p. Actually, the imaging surfaces 201 and 2 of the imaging device
Although 02 exists before the x-axis, it will be described in this way for convenience of explanation. Now, and to measure the distance Z _p to the object point p. Object points p are points P _L and P _R on the left and right images, respectively.
Focused on the point P _L, P _R of the x-direction coordinate on the image each X _L, When X _R, parallax data S is (Formula 1) as sought parallax data S = X _L - X _R (1) From the left and right images, the points P _L (X _L , Y _L ) and P _R (X _R , Y _R )
{Where Y _L = Y _R} the process of finding a called associating process. The association is performed by finding a pair of regions having the highest correlation from the left and right images. From the parallax data S, (Equation 2), the distance Z from the imaging device to the object p
The distance to the object point p can calculate the _{p Z p = B × f /} S (2) where, B: the arrangement interval of the imaging device, f: the focal length of the lens.

Next, a method of “counting method and apparatus for each moving object direction” described in JP-A-8-123935 will be described with reference to FIG. 7 as a person flow measuring apparatus using image processing. In FIG. 7, 301 is a camera, 302 is a moving object, 30
3 is a counting line, 304 is a slit background image, 305 is an input slit image, 306 is a slit-shaped moving image acquisition means, 307 is a background update processing means, 308 is a moving object difference binarized image, 309 is a background difference and binarized The processing means, 310 is a connected area forming processing means, 311 is a moving object image, 312 is a counting processing means, 313 is a passing direction determination processing means, 314 is an upward determination area, and 315 is a downward determination area.

A camera 301 is installed at an appropriate depression angle with respect to a plane such as a road, a floor, a conveyor or the like, and a moving image of a moving object 302 moving on the plane is photographed. The pixel value on the counting line 303 provided in the image, the value sequentially updated in a state where the object does not pass is used as the background value, and the pixel value on the counting line that changes over time due to the passing of the moving object and The moving object difference binarized image 308 is obtained by taking a difference with the background value and binarizing the difference value to extract a moving object region on the counting line.

[0005] Next, a moving object image 311 is formed by connecting (labeling) moving object regions for each counting line in chronological order. The counting processing means 312 counts the number of moving objects. When the formed moving object image 311 is composed of a plurality of moving objects,
The number of persons and the number of persons included in the moving object image are calculated by dividing the moving object image by the reference value. When the moving object image includes a single moving object, the moving object image is simply counted.

The moving direction is determined by a passing direction determining means.
At 313. Upward determination area 314 and downward determination area
At 315, a direction determination slit 316 for performing a difference binarization process similar to that used on the counting line is set, and the moving direction of the moving object is determined from the continuity of the moving object detection.

[0007]

In the conventional people flow measuring device, counting can be performed with high accuracy when a pedestrian passes alone, but when a plurality of people collectively pass, detection is not possible. The size of the detected area is divided by the reference value of the detection area per person, and is calculated as an approximate number of passers. This is because it is difficult to separate and detect a plurality of pedestrians one by one using only the luminance information of the image. Furthermore, regarding image capturing means, roads, sidewalks,
Since a pedestrian on a floor or the like is photographed by an image pickup device installed at a depressed angle above the pedestrian, a situation in which a pedestrian in the back becomes almost invisible to a pedestrian in the foreground has occurred. As described above, especially at the time of congestion, the counting accuracy was not sufficient.

[0008] A first object of the present invention is to provide a portable telephone system in which a plurality of pedestrians pass through a counting line.
An object of the present invention is to separate and detect them individually to improve counting accuracy.

Further, in the conventional method, it is difficult to determine the direction of movement of each of the pedestrians when the pedestrians are congested such that they pass each other.

A second object of the present invention is to determine a moving direction with high accuracy by grasping the exact position of each pedestrian and tracking this.

In the conventional method, it is assumed that the detection target is moving, and it is difficult to detect a stationary pedestrian. For this reason, it was not possible to accurately measure the congestion degree at the time of congestion in which a stationary pedestrian is likely to occur.

A third object of the present invention is to make it possible to accurately measure the congestion degree under any congestion condition.

[0013]

In order to solve the first problem, the present invention obtains an image suitable for image processing with a small blind spot by appropriately setting the positional relationship between a staircase and a plurality of imaging devices. Further, individual pedestrians are accurately separated and recognized and counted using distance data obtained by stereo image measurement. This improves the counting accuracy of pedestrians.

In order to solve the second problem, the present invention provides:
The position of the pedestrian is accurately grasped using the distance data from the imaging device to the pedestrian obtained by the stereo image measurement, and the movement direction is determined by tracking this. This makes it possible to determine the moving direction with high accuracy.

In order to solve the third problem, the present invention provides:
Using the distance data from the imaging device to the pedestrian obtained by stereo image measurement, a stationary pedestrian is also detected and the congestion degree is measured. Thus, the degree of congestion at the time of congestion can be accurately measured.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The invention according to claim 1 of the present invention relates to a method for detecting a correlation between a plurality of imaging devices installed at a predetermined depression angle with respect to a stair and a plurality of images from the plurality of imaging devices. An associating unit that acquires disparity data by associating a region with a higher area, a disparity distance converter that acquires distance data from the disparity data, a recognizer that recognizes a position where a pedestrian is present from the distance data, A counting unit that tracks and counts the position of the pedestrian over time, and installs a plurality of imaging devices on the stairs to acquire a stereo image with a small blind spot, and tracks the pedestrian over time. The counting has the effect of improving the counting accuracy.

According to a second aspect of the present invention, there are provided a plurality of image pickup devices installed at a predetermined depression angle with respect to a stair,
An associating unit that acquires disparity data by associating a region having a high correlation between a plurality of images from the plurality of imaging devices,
A parallax distance conversion unit that acquires distance data from the parallax data, a projection unit that projects the distance data on a two-dimensional plane, a recognition unit that recognizes a position of a pedestrian from a projection image, With a counting unit that tracks and counts the position of the pedestrian in time with a plurality of imaging devices on the stairs to acquire a stereo image with little blind spots, and distance data on the horizontal plane Project and separate pedestrians,
It has the effect of improving the counting accuracy by tracking and counting pedestrians over time.

According to a third aspect of the present invention, there are provided a plurality of image pickup devices installed at a predetermined depression angle with respect to a stair,
An associating unit that acquires disparity data by associating a region having a high correlation between a plurality of images from the plurality of imaging devices,
A parallax distance conversion unit that acquires distance data from the parallax data, a recognition unit that recognizes a position of a pedestrian from the distance data, and a movement that temporally tracks the position of the pedestrian to determine a moving direction It is provided with a direction determination unit,
A plurality of imaging devices are installed on the stairs to acquire a stereo image with a small blind spot, and the pedestrian is tracked temporally to determine the moving direction, thereby improving the moving direction determination accuracy.

According to a fourth aspect of the present invention, there are provided a plurality of image pickup devices installed at a predetermined angle of depression with respect to the stairs,
An associating unit that acquires disparity data by associating a region having a high correlation between a plurality of images from the plurality of imaging devices,
A parallax distance conversion unit for acquiring distance data from the parallax data, a recognition unit for recognizing a position where a pedestrian is present from the distance data, and a congestion degree for measuring a congestion degree of an imaging region from the position where the pedestrian is present It is equipped with a judgment unit, and a plurality of imaging devices are installed on the stairs to acquire a stereo image with a small blind spot, and congestion is determined by determining the degree of congestion based on how many pedestrians exist per unit area. This has the effect of improving the degree determination accuracy.

Hereinafter, embodiments of the present invention will be described with reference to FIGS.
This will be described with reference to FIG.

(First Embodiment) In a first embodiment of the present invention, a plurality of image pickup apparatuses, an associating section for obtaining parallax from a plurality of images, a parallax distance converting section for obtaining a distance from parallax, This is a person flow measuring device including a projection unit that projects distance data on a horizontal plane, a recognition unit that recognizes the position of a pedestrian, and a counting unit that tracks and counts pedestrians.

A person flow measuring device according to a first embodiment of the present invention will be described with reference to FIG. 101 indicates stairs, 1
02 indicates a pedestrian walking on the stairs. Reference numeral 103 denotes a plurality of imaging devices, which are arranged in parallel at a predetermined interval and their optical axes (104) have an angle close to horizontal. Reference numeral 105 denotes an associating unit, which calculates disparity data by associating regions having a high correlation between a plurality of images. Reference numeral 106 denotes a parallax distance conversion unit, which converts the parallax data obtained by the association unit 105 into distance data. Reference numeral 107 denotes a recognition unit, which separately recognizes the position of the pedestrian (108). 110 is a counting unit that counts pedestrians passing through the counting line (109).

In the following, stairs 102, associating unit 105,
The parallax distance conversion unit 106, the recognition unit 107, and the counting unit 110 will be described in more detail.

Stairs (102) These are general stairs for pedestrians ascending and descending, which are found in stations, buildings, sidewalks and the like. This is photographed by a plurality of imaging devices installed so that the optical axis is substantially horizontal. This stair may be an escalator.

The associating unit (105) The associating unit sequentially inputs images obtained by a plurality of image pickup devices. Here, the plurality of imaging devices is a stereo camera including two left and right cameras. The left image (111) is divided into M blocks in the horizontal X direction and N blocks in the vertical Y direction, a total of M × N blocks, based on the left image among the left and right images obtained from the stereo camera. As described above, the parallax data S (X, Y) is obtained for each block (X, Y) in the left image by stereo image measurement.

Parallax distance converter (106) The parallax distance converter uses the parallax data S (X, Y) obtained by the associating unit by using (Equation 2) used in the description of the conventional technique. Convert to distance data Z (X, Y).

Recognition unit (107) In the correspondence unit (105) and the parallax distance conversion unit (106), the distance to the pedestrian and the distance to the background stairs have been treated in the same manner. The measured stairs distance data (background distance data) is compared with the distance data Z (X, Y) obtained at a certain time, and the obtained distance data Z (X, Y) is When indicating the distance, it is not used in the subsequent processing. Next, the pedestrian distance data Z (Z, Y) is projected and labeled on a horizontal two-dimensional plane (108) consisting of the X axis and the Z axis, so that the position WLK ( X, Z, t)
Get. Note that the X-axis constituting the two-dimensional plane indicates a horizontal position on the left image as a reference, and the Z-axis indicates a distance on the optical axis from the stereo camera to the pedestrian.

By performing the projection processing, the position of the pedestrian can be grasped as if it were viewed from directly above. In addition, on the left image (111), pedestrians appear to overlap each other before and after, but since the distances Z from the camera are different, each pedestrian can be separated on the two-dimensional plane (108).

The counting unit (110) In the counting unit, a certain time t on the two-dimensional plane (X, Z)
Pedestrian position WLK separately recognized in
(X, Z, t) is temporally tracked and counted when passing through the counting line (109).

As described above, in the people flow measuring device according to the first embodiment of the present invention, a stereo image with a small blind spot can be obtained by installing a plurality of imaging devices on the stairs with a predetermined depression angle. Furthermore, by projecting distance data on a horizontal two-dimensional plane, pedestrians can be separated and their positions can be individually grasped. (Second Embodiment) A second embodiment of the present invention provides a plurality of imaging devices, an associating unit that obtains parallax from a plurality of images, and a distance from parallax. This is a person flow measuring device including a required parallax distance conversion unit, a recognition unit that recognizes a pedestrian, and a movement direction determination unit that determines a movement direction of the pedestrian.

A person flow measuring device according to a second embodiment of the present invention will be described with reference to FIGS. Plural imaging devices 401, associating unit 402, parallax distance converting unit 403, recognizing unit 4
04 is the same as in the first embodiment.
The moving direction determining unit 405 determines the moving direction of the pedestrian. The determination of the moving direction is performed by temporally tracking WLK (X, Z, t), which is the position 501 of the pedestrian obtained by the recognition unit 404, as shown in FIG. By calculating the amount of movement of the position, it is determined whether the pedestrian is up, down, or stationary.

Since the pedestrians are individually recognized and tracked and the pedestrian's movement amount and movement direction are determined, the pedestrian's movement situation, such as how many pedestrians walk in which direction and at what speed, is determined. Detailed distribution can be obtained. Also,
The trajectory of the pedestrian's movement can be easily obtained.

As described above, according to the second embodiment of the present invention, the human flow measuring device tracks the pedestrian in time with the imaging device, the associating unit, the parallax distance converting unit, and the recognizing unit. (Third Embodiment) The third embodiment of the present invention has a configuration in which a moving direction determination unit that determines a moving direction is provided. A human flow measurement including an imaging device, an associating unit that obtains parallax from a plurality of images, a parallax distance conversion unit that obtains a distance from parallax, a recognition unit that recognizes a pedestrian, and a congestion degree determination unit that measures the degree of congestion. Device.

A person flow measuring device according to a third embodiment of the present invention will be described with reference to FIGS. Plurality of imaging devices 601, association unit 602, parallax distance conversion unit 603, recognition unit 6
04 is the same as in the first embodiment.
The congestion degree determination unit 605 determines how many pedestrians 701 recognized in the imaging range of the imaging device are present. For example, as shown in the upper diagram of FIG. 5, if the number of pedestrians recognized as being within the imaging range at a certain time is larger than a predetermined threshold, it is determined that the pedestrian is congested, and the lower diagram of FIG. As described above, if it is smaller than the threshold, it is determined that the traffic is not congested. Not only the degree of congestion in the entire imaging range but also the degree of congestion for each area can be determined by dividing the imaging range into a plurality of areas.

As described above, according to the third embodiment of the present invention, the human flow measuring device includes an imaging device, an associating unit, a parallax distance converting unit, and a recognizing unit. And the congestion degree determination unit that determines the degree of congestion depending on whether there is only one, so that the accuracy of determining the degree of congestion on the stairs is improved. It is also possible to determine the congestion degree in consideration of the pedestrian's moving speed.

[0036]

As described above, according to the present invention, by installing a plurality of image pickup devices on the stairs, an image with a small blind spot is obtained, and further, by using the distance data obtained by the stereo image measurement, By accurately recognizing the position of each pedestrian, the counting accuracy of the pedestrian is improved, and an effect that the moving direction of the pedestrian can be accurately determined can be obtained.

In addition, by recognizing the position of each pedestrian regardless of whether the vehicle is stationary or moving, the degree of congestion can be accurately determined.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a person flow measuring device according to a first embodiment of the present invention;

FIG. 2 is a block diagram showing a configuration of a person flow measuring device according to a second embodiment of the present invention and a flow of processing for determining a moving direction of a pedestrian;

FIG. 3 is an explanatory diagram for determining a moving direction in a people flow measuring device according to a second embodiment of the present invention;

FIG. 4 is a block diagram showing a configuration of a person flow measuring device according to a third embodiment of the present invention and a flow of processing for determining the degree of congestion on stairs;

FIG. 5 is a diagram illustrating a determination of a congestion degree in the people flow measuring device according to the third embodiment of the present invention;

FIG. 6 is an explanatory view showing the principle of stereo image measurement in a conventional technique,

FIG. 7 is a block diagram showing a configuration and a processing flow of a conventional human flow measuring device.

[Explanation of symbols]

 101 Stairs 102 Pedestrian 103, 401, 601 Multiple imaging devices 104 Camera optical axis 105, 402, 602 Correspondence unit 106, 403, 603 Parallax distance conversion unit 107, 404, 604 Recognition unit 108 Two-dimensional plane 109 Count line 110 Counting unit 201 Left image 202 Right image B Baseline length (camera interval) p Object point 310 Camera 302 Moving object 303 Counting line 304 Slit background image 305 Input slit image 306 Slit moving image acquisition means 307 Background updating means 308 Moving object difference Binary image 309 Background difference processing and binarization processing means 310 Connected area forming processing means 311 Animal image 312 Counting processing means 313 Passing direction determination processing means 314 Upper direction determination area 315 Lower direction determination area 316 Direction determination slit 405 Moving direction determination unit 501, 701 Recognized pedestrian position 605 Congestion degree determination unit

Claims (4)

[Claims] 1. A method of acquiring disparity data by associating a plurality of imaging devices installed at a predetermined depression angle with respect to a stair and regions having a high correlation between a plurality of images from the plurality of imaging devices. Unit, a parallax distance conversion unit that obtains distance data from the parallax data, a recognition unit that recognizes a position of a pedestrian from the distance data, and a counting unit that tracks and counts the position of the pedestrian over time. A human flow measuring device comprising: 2. A method of acquiring disparity data by associating a plurality of imaging devices installed at a predetermined depression angle with respect to a stair with regions having a high correlation between a plurality of images from the plurality of imaging devices. A parallax distance conversion unit that obtains distance data from the parallax data, a projection unit that projects the distance data onto a two-dimensional plane, a recognition unit that recognizes a position of a pedestrian from a projection image, A counting section for tracking and counting the position of the pedestrian on a three-dimensional plane with time; 3. A method of acquiring disparity data by associating a plurality of imaging devices installed at a predetermined depression angle with respect to the stairs with an area having a high correlation between a plurality of images from the plurality of imaging devices. Unit, a parallax distance conversion unit that obtains distance data from the parallax data, a recognition unit that recognizes a position where a pedestrian is present from the distance data, and a movement direction that tracks the position of the pedestrian temporally. A human flow measuring device comprising a moving direction determining unit for determining. 4. A method of associating a plurality of imaging devices installed at a predetermined depression angle with respect to a stair with regions having a high correlation between a plurality of images from the plurality of imaging devices to obtain disparity data. Unit, a parallax distance conversion unit that obtains distance data from the parallax data, a recognition unit that recognizes a position where a pedestrian is present from the distance data, and measures a congestion degree of an imaging region from a position where the pedestrian is present. And a congestion degree determining unit that performs the measurement.