JP5419925B2 - Passing object number measuring method, passing object number measuring apparatus, and program - Google Patents

Description

  The present invention relates to a passing object number measuring method, a passing object number measuring apparatus, which measures the number of objects (objects, vehicles, etc.) passing through the place by image processing from a video taken by an imaging device (for example, a camera), And the program.

  As a conventional technique, there is known a method of measuring the number of persons passing through an area shown in a video by detecting and tracking a person from a video composed of time-sequential frames (for example, a non-image). Patent Document 1).

  Also, by estimating the number of people in each frame image without detecting individuals by weighting each pixel of video composed of time-sequential frames according to the correspondence with the position in the real space There is a known method (see, for example, Non-Patent Document 2).

O. Sidla, Y. Lypetskyy, N. Brandle, and S. Seer, “Pedestrian detection and tracking for counting applications in crowded situations”, Proc. IEEE International Conference on Video and Signal Based Surveillance (AVSS06), p. 70, 2006 Hiroyuki ARAI, Isao MIYAGAWA, Hideki KOIKE, and Miki HASEYAMA, “Estimating Number of People Using Calibrated Monocular Camera Based on Geometrical Analysis of Surface Area”, IEICE Trans. Fundamentals, Vol. E92-A, No. 8, pp. 1932- 1938, August 2009

  However, the passing object number measuring method of Non-Patent Document 1 has a problem that the accuracy of counting the number of people is greatly affected by the tracking performance, and it is difficult to obtain sufficient measurement accuracy particularly in a crowded situation.

  In addition, the method for measuring the number of people in Non-Patent Document 2 does not use information related to the movement of a person shown in an image, and therefore cannot measure the number of people who have passed through an area shown in a video. is there.

  The present invention has been made in view of such circumstances, and the purpose of the present invention is to estimate the number of objects shown in the image and the amount of movement (representative movement amount) as a group. An object of the present invention is to provide a passing object number measuring method, a passing object number measuring apparatus, and a program capable of estimating the number of passing objects in a reflected area.

  In order to solve the above-described problems, the present invention is a passing object number measuring method for estimating the number of objects passing through a predetermined area in a video, wherein partial image generation means is continuously input in time series. A partial image generation step of generating a partial image from the video frame, a step of foreground detection means detecting a foreground area in the generated partial image, and an on-image movement amount detection means in the detected foreground area The step of detecting the local movement amount on the image and the movement amount coordinate conversion / representative movement amount calculating means determine the relationship between the coordinates on the image and the real space coordinates from the detected local movement amount on the image. In consideration, the step of calculating the representative movement amount as a whole of the object group moving in each passing object number measurement direction, and the object number calculation means in the image move in a specific direction shown in the generated partial image. You A step of estimating the number of objects in the image, and a step of calculating the number of passing objects includes a step of calculating the number of objects passing through a predetermined area in the video from the representative movement amount and the number of moving objects in the image. This is a characteristic passing object number measuring method.

  According to the present invention, in the above invention, the step of the movement amount coordinate conversion / representative movement amount calculation means calculates a movement direction on the real space coordinates for each of the calculated local movement amounts on the image, And a first step of extracting a local movement amount on the image whose difference between the passing object number measurement direction of interest and a predetermined threshold value or less, and an absolute value of a real space movement amount for each of the extracted local movement amounts A second step of calculating a passing object number measurement direction component for each calculated real space movement amount absolute value, and a voting space with the real space movement amount absolute value as an axis And a fourth step of calculating a representative movement amount by performing a voting process on

  The present invention is characterized in that, in the above-mentioned invention, the first step removes, from the extraction target, a movement amount whose absolute value of the real space movement amount is inappropriate as the movement amount of the object.

  According to the present invention, in the above invention, the step by the passing object number calculating means further includes adding the calculated inter-frame passing object number for each measurement direction, so that the cumulative passing object number in each measurement direction is obtained. Is calculated.

  According to the present invention, in the above invention, the step by the object number calculation means in the image calculates the number of objects moving in each measurement direction when a plurality of passing object number measurement directions are set. And

  Further, in order to solve the above-described problem, the present invention is a passing object number measuring apparatus that estimates the number of objects passing through a predetermined area in a video, wherein the partial image generating means is continuously in time series. In a partial image generation unit that generates a partial image from an input video frame, a foreground detection unit that detects a foreground region in the partial image generated by the partial image generation unit, and a foreground region detected by the foreground detection unit The relationship between the on-image coordinates and the real space coordinates is calculated from the on-image movement amount detection unit for detecting the local movement amount on the image and the local movement amount on the image detected by the on-image movement amount detection unit. In consideration, the movement amount coordinate conversion / representation movement amount calculation unit for calculating the total movement amount of the object group moving in the passing object number measurement direction, and the partial image generated by the partial image generation unit Certain The number of objects in the image that estimates the number of objects in the image moving in the direction, the representative amount of movement calculated by the movement amount coordinate conversion / representative movement amount calculation unit, and the image estimated by the number of objects in the image A passing object number measuring apparatus comprising: a passing object number calculating unit that calculates the number of objects passing through a predetermined region in a video from the number of inner moving objects.

  In order to solve the above-described problem, the present invention provides a computer of a passing object number measuring apparatus that estimates the number of objects passing through a predetermined area in a video, and a partial image generating means is continuously provided in time series. A partial image generation function that generates a partial image from an input video frame, a foreground detection function that detects a foreground area in the partial image generated by the partial image generation function, and a foreground area detected by the foreground detection function In consideration of the relationship between the coordinates on the image and the real space coordinates from the movement amount detection function on the previous image for detecting the local movement amount on the image, and the local movement amount on the image detected by the movement amount detection function on the image. , A movement amount coordinate conversion / representation movement amount calculation function for calculating a representative movement amount as a whole of the group of objects moving in each passing object number measurement direction, and a partial image generated by the partial image generation function A function for calculating the number of objects in an image for estimating the number of objects in an image moving in a specific direction shown in FIG. 5, a representative movement amount calculated by the movement amount coordinate conversion / representative movement amount calculation function, and a function for calculating the number of objects in the image Is a program that executes a passing object number calculation function for calculating the number of objects that pass through a predetermined area in the video from the number of moving objects in the image estimated by the above.

  According to the present invention, it is possible to estimate the number of passing objects in the region shown in the video by estimating the number of objects shown in the video and the movement amount (representative movement amount) as a group.

It is a block diagram which shows the hardware constitutions for implement | achieving the passing object number measuring method by one Embodiment of this invention. It is a block diagram which shows the function structure of the passing object number measuring apparatus by this embodiment. It is a conceptual diagram which shows the input-output data to the function structure of the passing object number measuring apparatus by this embodiment. It is a flowchart for demonstrating the process of the passing object number measuring method by this embodiment. It is a flowchart which shows the representative movement amount calculation procedure by this embodiment. It is a conceptual diagram for demonstrating the relationship between real space height and a movement amount. It is a conceptual diagram explaining the height h in the real space of the point to coordinate-transform for a person as an example. It is a conceptual diagram which shows an example (in the case of a person) of voting processing by this embodiment.

  The present invention relates to a technique for estimating the number of people or the number of people passing through an area in a video. The feature of the present invention is that, for each moving direction set for each of a plurality of feature points in the foreground in the video, the amount of movement of the feature points in real space is obtained to obtain the peak, and the representative moving amount in each moving direction It is to do. By calculating the number of moving objects between frames for each moving direction, assuming the representative moving amount obtained from the feature point as the moving amount of the captured object set, it is possible to detect each object in the video without tracking it. It is possible to estimate the number of objects that pass through the predetermined area. That is, an object of the present invention is to measure the number of objects that pass through an arbitrary real space region shown in an image.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram showing a hardware configuration for realizing a passing object number measuring method according to an embodiment of the present invention. In implementing the method of measuring the number of passing objects according to the present embodiment, the hardware includes an imaging device 1 such as a camera that captures video, a computer 2 that performs image processing, and an image display device 3 such as a monitor that displays the image processing results. It can be realized using.

  The imaging device 1 uses a camera fixed at a predetermined position. The computer 2 includes an input device 2-1 including a keyboard and a mouse, a ROM 2-2 that stores programs and various parameters, and a CPU 2 that performs predetermined processing and operation control of each unit by executing the programs. -3, RAM 2-4 for storing various data associated with program execution, I / F 2-5 for capturing video captured by the imaging device 1, an external storage device 2-6 such as a hard disk, and a recording medium 2-8 A recording medium driving device 2-7 for controlling data writing and data reading between them, and a recording medium 2-8 such as a flash memory.

  Various computers such as a general-purpose computer, a board on which processing logic is mounted, and a chip are conceivable as the computer 2 that performs image processing, but this is not limited here.

  In the following, it is assumed that images (frame images) captured by the camera 1 are input in time series (still image series, video stream, etc., whether or not processing is performed in real time is not limited). explain. In addition, it is assumed that the direction of flow of an object is determined to some extent at a place where the present technology is used, and a viewing direction can be set in advance when measuring the number of passing objects.

  Further, in order to simplify the processing, an object shape model is introduced, and each object to be measured is assumed to be a rectangle having a height H and a width W. The height H and the width W are values specified in advance, and are stored so that they can be used as necessary in the processing. As specific values of the height H and the width W, for example, when a person is to be measured, H = 1.7 m and W = 0.3 m can be set.

Next, main functions of the apparatus configured as described above will be described.
FIG. 2 is a block diagram illustrating a functional configuration of the passing object number measuring apparatus according to the present embodiment. FIG. 3 is a conceptual diagram showing input / output data to the functional configuration of the passing object number measuring apparatus according to the present embodiment. As described above, the passing object number measuring apparatus according to the present embodiment is realized using the imaging device 1, the computer 2, and the image display device 3 shown in FIG. The functional configuration of the passing object number measuring apparatus shown in FIG. 3 is realized by program processing by the computer 2.

  In FIG. 2, the passing object number measuring device includes a partial image generation unit 10, a foreground detection unit 11, an on-image movement amount detection unit 12, a movement amount coordinate conversion / representative movement amount calculation unit 13, an in-image number calculation unit 14, and It consists of a passing number calculation unit 15. The partial image generation unit 10 generates a partial image obtained by extracting a preset range from a temporally continuous image series. The foreground detection unit 11 detects a foreground area in the extracted partial image, and generates a foreground image representing the foreground area. The on-image movement amount detection unit 12 detects a local movement amount on the image between the current frame and the previous frame of the video in the detected foreground region.

  The movement amount coordinate conversion / representative movement amount calculation unit 13 considers the relationship between the coordinates on the image and the floor coordinates of the real space (hereinafter simply referred to as real space coordinates) from the detected local movement amount on the image. By a method (described later), a movement amount (representative movement amount) as a whole of the object group moving in each passing object number measurement direction is calculated. The number-of-images calculation unit 14 estimates the number of objects (number of images) for the foreground image of the set partial image. The passing number calculation unit 15 calculates the number of passing objects between frames (passing number of persons) in each measurement direction from the representative movement amount and the number of moving objects in each measurement direction.

  In FIG. 3, the “partial image” is output to the partial image generation unit 10 and a “partial image” is output. In the foreground detection unit 11, “partial image” is input and “foreground image” is output. The on-image movement amount detection unit 12 receives “foreground image” and outputs “on-screen movement amount”. In the movement amount coordinate conversion / representative movement amount calculation unit 13, “movement amount on the screen” is input, and “representative movement amount” is output. In the in-image number calculation unit 14, “foreground image” is input, and “in-image number” is output. Then, the passing number calculation unit 15 receives “real space image” and outputs “number of people in the area”.

  FIG. 4 is a flowchart for explaining processing of the passing object number measuring method according to the present embodiment. First, the partial image generation unit 10 generates a partial image obtained by extracting a preset range from the current input image from the camera 1 (step S1). It is clear that the method for determining the extraction range is, for example, a method of setting a range in which a position in the real space where the number of objects are to be counted and a part that can correspond to the input image is extracted, or no object is captured. A method of extracting a portion excluding the portion (wall, fixed equipment, etc.) can be considered. This range may be the entire input image. Although the actual method of applying the extraction process is not limited here, for example, the size is the same as that of the input image, and the pixel value of the region to be extracted is “1” and the other pixel values are “0”. A method of limiting the processing range by masking using a value image is conceivable.

  Next, the foreground detection unit 11 detects a foreground area in the extracted partial image, and generates a foreground image representing the foreground area (step S2). Here, the foreground area represents an area where a moving object exists in the partial image, and the foreground image is “1” for a point that is the foreground in the partial image, and “0” for a point that is not, that is, the background. Is the image. There are various methods for detecting the foreground area (for example, Reference 1: Hitoshi Namibe, Toshikazu Wada, Takashi Matsuyama, “Background Difference Method Robust against Lighting Changes”, Information Processing Society of Japan Research Report (1998-CVIM115) , Vol. 1999, No. 29, pp. 17-24, 1999., Reference 2: Kedar A. Patwardhan, Guillermo Sapiro, Vassilios Morellas, “Robust Foreground Detection in Video Using Pixel Layers”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 30, No. 4, April 2008) is known, but the method is not limited here.

  Next, the on-image movement amount detection unit 12 detects a local movement amount on the image between the current frame and the previous frame of the video in the foreground area detected in step S2 (step S2). S3). That is, a set of coordinates on the frame image and a movement amount [pixel] between frames on the image at the coordinates is detected. The number of points for detecting the amount of movement is arbitrary, and it is not necessary to detect all the pixels in the foreground region. However, the more passing points are detected, the more accurately the number of passing objects can be expected.

  The detection method of the movement amount is not limited here, but it is preferable to select the detection method stably according to the scene conditions and the processing frame rate. For example, SIFT (Reference 3: DG Lowe, “Object recognition from local scale-invariant features”, Proc. Of IEEE International Conference on Computer Vision (ICCV), pp. 1150-1157, 1999.), SURF (Reference 4: H. Bay, A. Ess, T. Tuytelaars, and LV Gool, “Speeded-up robust features (SURF),” Elsevier, pp.346-359, 2008.) Conceivable. If this method is used, reliable acquisition of the local movement amount can be expected even when the frame rate is low or the degree of congestion is high. In addition, when the frame rate is sufficiently high, a method using an optical flow is also effective.

  Next, the movement amount coordinate conversion / representative movement amount calculation unit 13 calculates the coordinates on the image and the floor coordinates of the real space (hereinafter simply referred to as real space coordinates) from the local movement amount on the image detected in step S3. A movement amount (representative movement amount) as a whole of the object group moving in the direction of measuring the number of passing objects is calculated by a method considering the relationship (described later) (step S4).

The coordinate conversion of the movement amount is, for example, by obtaining the coordinates on the image of the start point and the end point of the movement for each detected movement amount, converting the coordinates on the image into real space coordinates, and taking the difference between the real space coordinates. It can be realized. Conversion from the coordinates (x, y) on the image to the real space coordinates (X _h , Y _h ) when the point is located at the height h in the real space is expressed by the following equation (1). (It is derived by setting T _X = T _Y = 0 and Z = h in the expression (12) of Non-Patent Document 2).

FIG. 6 is a conceptual diagram for explaining the relationship between the real space height and the movement amount. f is the focal length of the camera, T _Z is the floor surface to the camera height. R is a rotation matrix, and is a value that depends only on the rotation angle θ around the optical axis of the camera and the depression angle φ from the horizontal direction. At this time, the height h in the real space of the point to be coordinate-converted cannot be specified from the local movement amount on the image.

  FIGS. 7A and 7B are conceptual diagrams for explaining the height h in the real space of the point to be coordinate-transformed by taking a person as an example. As shown in FIG. 7A, when the height h in the real space is H, the movement amount is small, and as shown in FIG. 7B, the height h in the real space is “ In the case of “0”, the movement amount becomes large. Thereby, the real space movement amount that can be considered from the detected local movement amount on the image is not uniquely determined but has a certain width. When calculating the representative movement amount, it is necessary to perform processing based on this fact.

  Considering the above indefiniteness of the real space movement amount, the representative movement amount calculation method described below can be considered. The representative movement amount in each measurement direction is calculated by performing the processing described below for all the passing object number measurement directions.

  FIG. 5 is a flowchart showing a representative movement amount calculation procedure according to the present embodiment. First, for each local movement amount on the image calculated in step S3, the movement direction on real space coordinates is calculated, and the difference between that direction and the passing object number measurement direction of interest is below a certain threshold value. The upper local movement amount is extracted (step S41).

If the difference between the real space coordinates at the start point and the end point of the movement is (ΔX _h , ΔY _h ), the direction of the movement amount on the real space coordinates is, for example, d = Although tan ⁻¹ (ΔY _h / ΔX _h ) can be set, from Equation (1), this value is constant regardless of the value of h, and can be calculated by setting an arbitrary h. In addition, the threshold value of the difference between the directions described above considers how the passing object number measurement direction is set so that one local movement amount is not extracted for a plurality of passing object number measurement directions. Set. For example, when the number of passing objects is measured in four directions of + X, −X, + Y, and −Y, the threshold value can be set to ± 45 °. In addition, a smaller threshold may be set as necessary in consideration of the small change in the moving direction of the object and the influence of the amount of local movement detected in error.

At this time, the movement amount in which the absolute value r _h = (ΔX _h ² + ΔY _h ² ) ^{1/2 of} the real space movement amount is inappropriate as the movement amount of the object may be removed from the extraction target. For example, when the processing frame rate is 5 [fps] and the maximum value allowed for the moving speed of the object is set to 9 [km / h], the maximum allowable real space movement amount between frames is set. The value is 0.5 [m]. In this case, when h = H (in this case, the real space movement amount takes a minimum possible value), the local movement amount whose absolute value of the real space movement amount exceeds 0.5 is excluded from the extraction target. Can do. Thereby, the influence of the amount of local movement detected accidentally can be reduced.

Next, for each local movement amount extracted in step S41, the absolute value R _H , R of the real space movement amount when the height in the real space is h = H and when h = 0. ₀ is calculated (step S42). These values correspond to the case where the characteristic points are derived from the upper end and the lower end of the object, respectively. That is, it represents the minimum value and the maximum value that can be considered as the real space movement amount in the local movement amount.

Next, the passing object number measurement direction component _{RD, H} , _{RD, 0} is calculated for each real space movement amount absolute value _RH , _R0 calculated in step S42 (step S43). This value can be calculated from the relationship R _{D, h} = R _h cos d _dif . Here, d _dif is the difference between the moving direction in the real space and the passing object number measuring direction of interest.

Next, the representative movement amount is calculated by voting processing to the voting space with the movement amount absolute value as an axis (step S44). That is, each local movement amount extracted in step S41 is voted in the range of R _{D, H to} R _{D, 0 in} the movement amount.

  FIG. 8 is a conceptual diagram showing an example (in the case of a person) of voting processing according to the present embodiment. When voting is performed by the above-described method, as shown in FIG. 8, it can be expected that the number of votes has a peak in the actual real space movement amount. Therefore, the absolute value of the movement amount having the largest number of votes is set as the representative movement amount in the passing object number measurement direction of interest. Note that the step size of the coordinates of the voting space is determined according to an error in the allowable representative movement amount. For example, when the allowable error is ± 0.1 [m], the step size of the voting space can be 0.2 [m].

Here, when the processing based on the system according to Equation (1) is performed, R _{D, 0} / R _{D, H} = T _Z / (T _Z −H) holds. Therefore, when the voting space is considered as a logarithmic scale with respect to the movement amount direction, after voting for coordinates corresponding to log (R _{D, H} ) in each movement amount, the length in the movement amount absolute value direction is logT _Z / ( T _Z −H), the voting process can be realized by applying a filter whose coefficient is “1” to the voting space.

  The calculation time required for normal voting processing can be expressed in the order of (number of feature point pairs) x (voting space size), whereas the voting processing order using the logarithmic scale is (voting space size). X (filter length). Here, both the size of the voting space and the length of the filter can be replaced with fine voting space. Therefore, it is desirable to select a voting method that can be processed at higher speed in consideration of the number of sets of feature points and the fineness of the voting space. Thereafter, the process returns to the main routine shown in FIG.

  Next, returning to FIG. 4, the number-of-images calculation unit 14 estimates the number of objects (number of images) in the partial image generated in step S1 (step S5). Although the method for estimating the number of objects is not limited here, for example, when the technique of Non-Patent Document 2 is used, the number of objects in the partial image can be stably estimated even during congestion. Non-Patent Document 2 describes a measurement target limited to a person, but can be applied to an arbitrary rectangular object by replacing the person model with a shape model of an appropriate object.

  At this time, when a plurality of passing object number measurement directions are set, it is necessary to calculate the number of objects moving in each measurement direction. Although the method is not limited here, for example, the local movement amount can be obtained by the following equation (2) that is established when it is assumed that the group moving in any measurement direction is evenly acquired. .

Here, i is an index of the passing object counting direction, N _i is the number of the moving object to the i-th measurement direction. N is the total number of objects in the partial image calculated in the previous stage of this step. M _i is the number of local movements extracted in step S3 with respect to the i-th measurement direction.

  Next, the passing person calculation unit 15 calculates the number of passing objects between frames in each measurement direction from the representative movement amount calculated in step S4 and the number of moving objects in each measurement direction (number of persons in the image) calculated in step S5. Is calculated (step S6). Although the calculation method of the number of passing objects between frames is not limited here, for example, it can be calculated by the following equation (3).

Here, the subscript i indicates that the value relates to the i-th passing object number measurement direction. C is the number of inter-frame passing objects to be obtained. v and S are the real movement area of the area (measurement area) shown in the representative movement amount and the partial image, respectively. _Wi is an index of the real space length of the measurement area in the direction perpendicular to the i-th measurement direction. Here, depending on the shape of the measurement area, the length of the measurement area in the direction may not be uniquely determined. As a _Wi determination method in this case, for example, a method of taking an average value of the area lengths can be considered. If the length of the area is uniquely determined, it is possible to the value and W _i.

  Next, the passing number calculation unit 15 adds the number of passing objects between frames for each measurement direction, thereby obtaining the cumulative number of passing objects in each measurement direction (step S7). The cumulative number of passing objects is the number of objects that have passed through the video in each measurement direction from the frame in which processing is started to the current frame.

  Next, it is determined whether or not the end condition is satisfied (step S8). If the end condition is not satisfied, the process proceeds to the next frame (step S9), returns to step S1, and repeats the above-described processing. . Thereafter, the above-described processing is repeated while moving to the next frame until the end condition is satisfied. When the end condition is satisfied, the process is ended.

  According to the above-described embodiment, the number of passing objects can be measured without tracking individual objects shown in the video. This makes it possible to measure the number of passing objects even in situations where it is difficult to track individual objects.

DESCRIPTION OF SYMBOLS 1 Imaging device 2 Computer 2-1 Input device 2-2 ROM
2-3 CPU
2-4 RAM
2-5 I / F
2-6 External Storage Device 2-7 Recording Medium Drive Device 2-8 Recording Medium 3 Image Display Device 10 Partial Image Generation Unit 11 Foreground Detection Unit 12 On-Image Movement Amount Detection Unit 13 Travel Path Hot Water Coordinate Conversion / Representative Movement Amount Calculation Unit 14 Number of people detection section in image 15 Number of people passing through calculation section

Claims (5)

A method for measuring the number of passing objects for estimating the number of objects passing through a predetermined area in a video,
A partial image generation step in which the partial image generation means generates a partial image from video frames that are continuously input in time series;
A foreground detecting means detecting a foreground region in the generated partial image;
A step of detecting a movement amount on the image in the detected foreground region, a local movement amount on the image;
The moving amount coordinate conversion / representative moving amount calculating means calculates the moving direction on the real space coordinates for each of the detected local moving amounts on the image, and the direction and the passing object number measuring direction of interest are calculated. Extracting a local movement amount on the image whose difference is equal to or less than a predetermined threshold;
A step of calculating a movement amount coordinate conversion / representative movement amount calculation means for each of the extracted local movement amounts, an absolute value of a real space movement amount;
A step of calculating a passing object number measurement direction component for each of the calculated real space movement amount absolute values by a movement amount coordinate conversion / representative movement amount calculation unit;
An object in the real space in the voting space, in which the movement coordinate conversion / representative movement amount calculation means has a coordinate axis of the absolute value of the real space movement amount and is a step size of coordinates set based on the allowable movement amount error. Voting processing is performed on the coordinates of the voting space corresponding to each of the calculated absolute values of the real space movement amount in a range corresponding to the upper end to the lower end of the real space movement amount, and the absolute value of the real space movement amount having the most votes is obtained. Outputting as a representative amount of movement of the entire moving object group;
A step of estimating the number of moving objects in the image, wherein the number of objects in the image calculating means moves in a specific direction shown in the generated partial image;
A passing object number calculation means divides the result obtained by multiplying the representative movement amount, the real space length of the predetermined area by the number of moving objects in the image, and the real space area of the predetermined area by the image. And calculating as the number of objects passing through the predetermined region . The step by the passing object number calculating means further calculates the cumulative passing object number in each measurement direction by cumulatively adding the calculated inter-frame passing object number for each measurement direction. Item 2. The passing object number measuring method according to Item 1 . 3. The step according to claim 1, wherein the step of calculating the number of objects in the image calculates the number of objects moving in each measurement direction when a plurality of passing object number measurement directions are set. Number of passing objects. A passing object number measuring device for estimating the number of objects passing through a predetermined area in a video,
It means for generating a partial image from the image frame input in succession in time series,
It means for detecting a foreground region before Kisei made partial image,
In the foreground area issued before dangerous, it means for detecting a local amount of movement in the image,
For each of the detected local movement amounts on the image, the movement direction on the real space coordinates is calculated, and the difference between the direction and the passing object number measurement direction of interest is below a predetermined threshold value. Means for extracting the quantity;
Means for calculating an absolute value of a real space movement amount for each of the extracted local movement amounts;
Means for calculating the passing object number measurement direction component for each calculated real space movement absolute value;
In the voting space having the coordinate axis of the real space movement amount absolute value and the step size of the coordinates set based on the allowable movement amount error, the range corresponding to the upper end to the lower end of the object in the real space The voting process is performed on the coordinates of the voting space corresponding to each of the calculated absolute values of the real space movement amount, and the real space movement amount absolute value having the largest number of votes is used as the representative movement amount of the object group as a whole. Means for outputting;
It means for estimating the image moving objects number moving in a certain direction that is reflected in the front Kisei made partial image,
A value obtained by multiplying the representative movement amount, the real space length of the predetermined area by the number of moving objects in the image, and the area of the predetermined area in the real space is defined as the predetermined area in the video. And a means for calculating the number of passing objects. A passing object number measuring program for causing a computer to execute the passing object number measuring method according to any one of claims 1 to 3 .

Images