JP2004021375A - Object counting method and object counting device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To rapidly and precisely count the number of objects to be monitored with a simple constitution. <P>SOLUTION: The background image of a monitoring area is specified (Step 112) from an image pickup result of the monitoring area obtained by the image pickup (Step 111) of the monitoring area, and the area of the object to be monitored in the monitoring area is extracted on the basis of a difference image between the image pickup result of the monitoring area and the background image (Step 113). Then, the area occupying ratio of the object area in a window area in each position of the window is computed while moving the window having a size corresponding to an estimated maximum size of the object in the monitoring area (Step 114). The number of the objects in the monitoring area is computed on the basis of the distribution of the area occupying ratio in the monitoring area (Step 115). <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an object counting method and an object counting apparatus, and more particularly, to an object counting method and an object counting apparatus for counting an object such as a person in a square or a vehicle on a road as a monitoring target.
[0002]
[Prior art]
Conventionally, various proposals have been made in order to grasp the current number of people on a square or vehicles on a road. One of them is to take advantage of the fact that these objects are essentially moving objects, take images of a predetermined monitoring area with a video camera or the like at a time interval, and use the difference between frames as the difference between the imaging results. An object is extracted from an image and an attempt is made to grasp the number of objects (hereinafter, referred to as “conventional example 1”).
[0003]
In addition, a technique has been proposed in which individual objects are pattern-recognized from the result of imaging of a predetermined monitoring area by a video camera, and the number of objects in the predetermined monitoring area is grasped (hereinafter, “conventional example 2”). ).
[0004]
[Problems to be solved by the invention]
In the technology of the first conventional example described above, since the object to be monitored is a moving object, when a person stops in a square or when a vehicle on the road is not moving due to traffic congestion, The object to be monitored cannot be counted. Also, even if the monitored object is moving, if the object is moving slowly, it is difficult to accurately count the objects using the difference image between adjacent frames. Therefore, the object cannot be counted quickly and accurately.
[0005]
Further, in the technique of the second conventional example, since an individual object is extracted from an imaging result by pattern recognition, a very large-scale calculation resource is required. Such a system having a large-scale computational resource necessarily has a large-scale and complicated configuration.
[0006]
The present invention has been made under such circumstances, and provides an object counting method and an object counting device that can quickly and accurately count the number of objects to be monitored with a simple configuration. Aim.
[0007]
[Means for Solving the Problems]
The present invention relates to an object counting method for counting the number of objects to be monitored, an imaging step of imaging a monitoring area for monitoring the object, and an image capturing method of the monitoring area based on an imaging result image of the monitoring area. A background image specifying step of specifying a background image that is a captured image in the monitoring area when the object does not exist; and an object area occupied by the object based on a difference image between the captured result image and the background image. Extracting an object region; and moving a window having a size corresponding to the maximum size and the estimated size of the object in the monitoring region, while moving the window in an area within the window at each position of the window. An area occupancy calculating step of calculating an area occupancy of the object area; and a counter for counting the number of the objects in the monitoring area based on the distribution of the area occupancy in the monitoring area. When; it is an object counting method comprising.
[0008]
In this object counting method, first, in the imaging step, a monitoring area for monitoring an object is imaged. Subsequently, in the background image specifying step, a background image that is a captured image in the monitoring area when the object does not exist in the monitoring area is specified based on the captured result image of the monitoring area.
[0009]
Next, in the object area extraction step, a difference image between the specified background image and the actually captured image in the monitoring area is calculated. Then, an object region in the monitoring region is extracted from a distribution of pixels having an absolute value equal to or greater than a predetermined threshold in the calculated difference image.
[0010]
Subsequently, in the area occupancy calculating step, the area occupancy of the object area in the window area at each position of the window is moved while moving a window having a size corresponding to the estimated size of the object in the monitoring area. Is calculated. Here, when the object to be monitored covers a plurality of types having greatly different sizes from each other, for example, a car and a bus, each of the plurality of types of windows corresponding to the size corresponding to each type is provided. Just set it.
[0011]
Next, in a counting step, the number of objects in the monitoring area is counted based on the distribution of the area occupancy in the monitoring area.
[0012]
Therefore, according to the object counting method of the present invention, since the number of objects is counted by extracting the object region from the imaging result for one frame related to the monitoring region, the number of monitoring target objects in the monitoring region is quickly counted. be able to. Further, according to the object counting method of the present invention, the large-scale calculation is not required as in the case of identifying each individual object by pattern recognition, the object area is extracted, and the area occupancy distribution of the object area in the monitoring area is obtained. Since the number of objects in the monitoring area is counted by a relatively simple calculation of calculating the number of objects, the number of objects in the monitoring area can be accurately counted with a simple configuration. That is, according to the object counting method of the present invention, the number of objects to be monitored can be counted quickly and accurately with a simple configuration.
[0013]
In the object counting method according to the present invention, in the counting step, a peak shape of the distribution of the area occupancy may be analyzed to count the number of the objects. For example, when the monitoring area is imaged in a perspective view from an obliquely upward direction, the number of objects can be counted with high resolution even when there are a plurality of objects partially overlapping.
[0014]
Further, in the object counting method according to the present invention, it is possible to select, as the monitoring area, an area in which the variance and change width of the pixel density value in the imaging result are equal to or less than an allowable value when the object does not exist inside the monitoring area. In such a case, the background image can be specified with high accuracy by simple captured image processing.
[0015]
The object counting device of the present invention is an object counting device that counts the number of objects to be monitored, and an imaging unit that captures an image of a monitoring region that monitors the object; based on an imaging result image of the monitoring region, Background image specifying means for specifying a background image that is a captured image in the monitoring area when the object does not exist in the monitoring area; and determining the object based on a difference image between the captured result image and the background image. Object area extracting means for extracting an object area occupied by the object; and moving the window having a size estimated to be the maximum size of the object in the monitoring area while moving the object in an area within the window at each position of the window. Area occupancy calculating means for calculating the area occupancy of the area; and a counter for counting the number of objects in the monitoring area based on the distribution of the area occupancy in the monitoring area. When; it is an object counting device comprising a.
[0016]
In the object counting device, the background image specifying unit specifies a background image that is a captured image in the monitoring area when the object does not exist in the monitoring area, based on an imaging result of the predetermined area captured by the imaging unit. . Subsequently, after the object region extracting means calculates a difference image between the specified background image and the captured image in the actually captured monitoring region, pixels having an absolute value equal to or larger than a predetermined threshold value in the calculated difference image are calculated. , An object region within the monitoring region is extracted.
[0017]
Next, the area occupancy calculating means moves the window having a size corresponding to the size estimated to be the maximum size of the object in the monitoring area while occupying the area of the object area in the window area at each position of the window. Calculate the rate. Then, the counting means counts the number of objects in the monitoring area based on the distribution of the area occupancy in the monitoring area.
[0018]
That is, the object counting device of the present invention counts the number of objects in the monitoring area using the object counting method of the present invention. Therefore, according to the object counting device of the present invention, the number of objects to be monitored can be counted quickly and accurately with a simple configuration.
[0019]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an embodiment of the present invention will be described with reference to FIGS.
[0020]
FIG. 1 is a block diagram illustrating a configuration of an object counting device 100 according to an embodiment. The object counting device 100 is a device that counts the number of people in a predetermined monitoring area in a square.
[0021]
As shown in FIG. 1, the object counting apparatus 100 includes a video camera 10 that captures an image of a predetermined monitoring area WTA (see FIGS. 2A and 2B) in a square, and an imaging result obtained by the video camera 10. And a processing device 20 for performing a process of counting the number of persons who are the monitoring target objects, and a screen display device 43 and a voice display device 45 for displaying the counting result. Further, the object counting device 100 further includes a keyboard device 41 and a mouse 44 for an operator to input an instruction to the processing device 20. The screen display device 43 is also used for displaying input information from the keyboard device 41 and the mouse 44 for confirmation, and for displaying the operation status of the object counting device 100.
[0022]
As shown conceptually in FIG. 2A, the video camera 10 is installed above a square, and observes the monitoring area WTA from diagonally above. As a result, in the visual field of the video camera 10, the persons OBJ1, OBJ2, OBJ3 in the monitoring area WTA and their surroundings can be captured in a perspective image as shown in FIG. I have.
[0023]
In the following description, a vertically upward direction is referred to as a Z direction, a lateral direction of the monitoring area WTA is referred to as an X direction, and a depth direction of the monitoring area WTA is referred to as a Y direction. Also, the captured image (i.e., see FIG. 2 (B)) monitor coordinate system in (X _{M, Y M)} in the strict sense is the intersection angle between the Y-direction and the Y _M direction for each position in the image is different, especially when there is no need to focus on the difference between the Y-direction and the Y _M direction, Y _M direction of the monitor coordinate system, as ones to match the depth direction i.e. Y-direction of the monitoring region WTA, the following description Do.
[0024]
Returning to FIG. 1, the processing device 20 includes (a) a control device 29 that integrally controls the entire object counting device 100, (b) an imaging data collection device 21 that collects imaging result data of the video camera 10, and (c) A) a background image specifying device 22 for specifying a background image in the monitoring area WTA. Further, the processing device 20 calculates (d) an object region extraction device 23 that extracts an object region (human region) in the monitoring region WTA, and (e) calculates an area occupancy of the object region at each position in the monitoring region WTA. (F) a counting device 25 for counting the number of people in the monitoring area WTA based on the distribution of the area occupancy of the object area in the monitoring area WTA.
[0025]
Further, the processing device 20 includes (g) a storage device 30. The storage device 30 has an imaging data storage area 31, a background image storage area 32, an object area information storage area 33, and an area occupancy storage area 34.
[0026]
In FIG. 1, the flow of data is indicated by solid arrows, and the flow of control is indicated by dotted arrows. The operation of each device of the processing device 20 configured as described above will be described later.
[0027]
In the present embodiment, the processing device 20 is configured by combining various devices as described above. However, the processing device 20 is configured as a computer system, and the functions of each device except the storage device 30 are built in the processing device 20. It is also possible to realize this using a program.
[0028]
Hereinafter, the counting operation of the object (person) by the object counting device 100 of the present embodiment will be described along the flowchart shown in FIG. 3 and appropriately referring to other drawings.
[0029]
First, in step 111 of FIG. 3, imaging for counting an object (person) is performed. In this imaging, the imaging data collection device 21 collects the imaging data sent from the video camera 10 for one screen. Then, the imaging data collection device 21 stores the collected imaging data in the imaging data storage area 31.
[0030]
Next, in step 112, the background image in the monitoring area WTA is specified. In specifying such a background image, first, the background image specifying device 22 reads out the current image data from the image data storage area 31. Then, in the monitoring area WTA, the position of the monitoring area WTA in the captured result image is specified with reference to the reference object REF such as a tree that originally does not move. By specifying the position of the monitoring area WTA, the instability of the position of the monitoring area WTA in the captured image due to the shaking of the video camera or the like is eliminated.
[0031]
Subsequently, the background image specifying device 22 specifies a background image in the monitoring area WTA whose position has been specified. When specifying such a background image, in the present embodiment, the background image specifying device 22 divides the monitoring area WTA into a plurality of small areas. Subsequently, for each of the plurality of small areas, whether the variance value is within the allowable error range from the reference pixel density distribution expected as the pixel density distribution in the background image of the monitoring area is determined by the allowable variance value. Is determined based on whether or not the change width is equal to or less than the allowable change width.
[0032]
Then, the background image specifying device 22 specifies a new background image based on the estimated background image up to the previous time and the current internal pixel density distribution for the small area for which a positive determination has been made. In addition, the background image specifying device 22 specifies a new background image based on the estimated background image up to the previous time or a new background image of a nearby small region for a small area for which a positive determination has been made. To specify a new background image.
[0033]
The background image specifying device 22 stores the position information and the background image of the monitoring area WTA specified in the background image storage area 32.
[0034]
Next, in step 113, the object region extraction device 23 extracts an object region (human region). When extracting such an object region, the object region extraction device 23 first reads out the imaging result image from the imaging data storage region 31 and also reads out the position information and the background image of the monitoring region WTA from the background image storage region 33. Subsequently, after calculating the difference between the captured image and the background image, the object region extracting device 23 calculates the difference image by averaging the density values of several pixels in the vicinity.
[0035]
Next, the object region extracting device 23 extracts, as an object region, a region including pixels whose absolute value of the difference is equal to or larger than a predetermined threshold. Then, the object area extraction device 23 sets each pixel value of the area determined as the object area in the monitoring area WTA to “1 (or 0)”, and sets each pixel value of the area determined as the background image area. Is stored in the object area information storage area 33.
[0036]
Next, in step 114, the area occupancy calculating device 24 calculates the area occupancy of the object area at each position of the monitoring area WTA. In calculating the area occupancy, the area occupancy calculation device 24 first reads a binarized image from the object area information storage area 33. Subsequently, as shown in FIG. 4, the area occupancy calculation device 24 sets a rectangular window WIN corresponding to the maximum size of the area occupied by the person who is the counting target object in the monitoring area WTA and the estimated size. While moving in the monitoring area WTA, the occupancy of the area occupied by the object area in the area within the window WIN at each position is calculated. The area occupancy calculating device 24 stores the area occupancy at each position of the monitoring area WTA thus calculated in the area occupancy storage area 34.
[0037]
In the present embodiment, since the captured image perspective image, depending on the position in the depth direction (Y _M direction), while adjusting the size of the window WIN, occupancy of the area occupied by the object area in the window WIN region Is calculated.
[0038]
Next, in step 115, the counting device 25 counts the number of objects (persons) in the monitoring area WTA based on the calculated area occupancy distribution. In such counting, the counting device 25 first reads the area occupancy at each position of the monitoring area WTA from the area occupancy storage area 34, and obtains the area occupancy distribution in the monitoring area WTA. FIG. 5 shows an example of the area occupancy distribution thus obtained.
[0039]
Subsequently, the counting device 25 extracts the peak of the obtained area occupancy distribution. Then, the counting device 25 analyzes the shape of each of the extracted peaks, and determines whether or not each peak is caused by a person being at the peak position. Then, for each peak determined to have occurred due to the presence of a person, the number of persons at that peak position is determined. For example, in the case of a peak having a typical peak width and a peak height as shown in FIG. 6A and having a monotonous shape when viewed from a normal human physique, a person is positioned at the peak position. It is determined that there is one person.
[0040]
In addition, as shown by the solid line in FIG. 6B, in the case of a peak having a wide peak width and a high peak height but a monotonous shape, it is determined that one person with a large body is located at the peak position. I do. Further, as shown by a broken line in FIG. 6B, in the case of a peak having a narrow peak width and a low peak height but having a monotonous shape, it is determined that there is one small body at the peak position. . That is, when the peak width is within the predetermined peak width range, the peak height is within the predetermined peak height range, and the peak has a monotonous shape, it is determined that one person exists at the peak position. .
[0041]
Also, as shown in FIG. 6 (C), when there is a peak shape of two peaks, it is determined that there are two persons at the peak position. In the case where the peak shape has two or more peaks, it is determined that there are people by the number of peaks.
[0042]
Thus, the total number of persons and the position of each person in monitoring area WTA are detected.
[0043]
The peak shapes of the area occupancy distributions shown in FIGS. 6A to 6C are obtained by performing the smoothing process performed in the present embodiment. In analyzing the shape of each peak, a simple template matching process can be performed.
[0044]
The above detection result is notified to the control device 29. The control device 29 instructs the screen display device 43 and the audio display device 45 to display the received detection result. Thereby, the total number of persons and the distribution of the presence density in the monitoring area WTA are provided to the operator through vision or hearing.
[0045]
Next, in step 116, the control device 29 determines whether or not the end of the counting operation is instructed from the keyboard device 41 or the mouse 44. Here, if a negative determination is made, the process proceeds to step 111, and the counting of the target object (person) is continued in the same manner as described above. On the other hand, if a positive determination is made in step 116, the counting operation of the target object ends. Thus, the counting operation of the target object is repeatedly performed until the end of the counting operation is instructed from the keyboard device 41 or the mouse 44.
[0046]
As described above, according to the object counting device 100 of the present embodiment, the area occupancy distribution of the object area in the monitoring area WTA is obtained based on the imaging results of the video camera 10 each time, and the area occupancy distribution is obtained. , The number of objects to be counted in the monitoring area WTA is counted. Therefore, the number of objects can be counted quickly and accurately with a simple configuration.
[0047]
Further, since the number of objects in the monitoring area WTA is counted by analyzing the peak shape in the area occupancy distribution, the number of objects (persons) can be counted with high resolution.
[0048]
In the above embodiment, the monitoring area is imaged from obliquely above by the video camera 10. However, if a wide-angle imaging device is used, the monitoring area can be imaged from vertically above the monitoring area. In such a case, it is possible to eliminate the trouble of coordinate conversion due to the perspective image.
[0049]
Further, in the above embodiment, the total number of persons and the position of each person in the monitoring area WTA are detected in each frame image obtained by imaging with the video camera 10, but each frame image in each frame image is further detected. A so-called tracking function can also be realized by determining the time change of the position of each person or specific person based on the position of the person or specific person.
[0050]
Further, in the present embodiment, the monitoring target is a person in the open space, but the present invention can be applied to a vehicle on a road, a vehicle in a parking lot, and the like. Further, the present invention can be applied regardless of whether it is outdoors.
[0051]
【The invention's effect】
As described above in detail, according to the object counting method and the object counting apparatus of the present invention, the number of objects to be monitored can be counted quickly and accurately with a simple configuration.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a schematic configuration of an object counting device according to an embodiment of the present invention.
2A is a diagram for explaining an arrangement of the video camera in FIG. 1, and FIG. 2B is a diagram for explaining an image pickup result by the video camera in FIG. 1; .
FIG. 3 is a flowchart illustrating a counting process performed by the object counting device of FIG. 1;
FIG. 4 is a diagram for explaining movement of a window in a monitoring area.
FIG. 5 is a diagram illustrating an example of an area occupancy distribution of an object area in a monitoring area.
FIGS. 6A to 6C are diagrams for explaining examples of peak shapes in an area occupancy distribution; FIG.
[Explanation of symbols]
10 video camera (imaging means), 22 background image specifying device (background image specifying device), 23 object region extracting device (object region extracting device), 24 area occupancy calculating device (area occupying calculating device) 25 ... Counter (counter).

Claims (4)

An object counting method for counting the number of objects to be monitored,
An imaging step of imaging a monitoring area for monitoring the object;
A background image specifying step of specifying a background image that is a captured image in the monitoring area when the object does not exist in the monitoring area, based on the captured result image of the monitoring area;
An object region extracting step of extracting an object region occupied by the object based on a difference image between the image of the captured image and the background image;
While moving a window having a size corresponding to the maximum size and the estimated size of the object in the monitoring region, calculate the area occupancy of the object region in a region within the window at each position of the window. Calculating an area occupancy ratio;
A counting step of counting the number of the objects in the monitoring area based on the distribution of the area occupancy in the monitoring area. 2. The object counting method according to claim 1, wherein in the counting step, the number of the objects is counted by analyzing a peak shape of the distribution of the area occupancy. 3. The object according to claim 1, wherein the monitoring region is a region where a variance and a variation width of a pixel density value in an imaging result are equal to or less than an allowable value when the object does not exist inside the monitoring region. 4. Counting method. An object counting device for counting the number of objects to be monitored,
Imaging means for imaging a monitoring area for monitoring the object;
Background image specifying means for specifying a background image that is a captured image in the monitoring area when the object does not exist in the monitoring area, based on the captured result image of the monitoring area;
Object region extracting means for extracting an object region occupied by the object based on a difference image between the captured result image and the background image;
Area occupancy calculating means for calculating an area occupancy of the object area in a region within the window at each position of the window while moving a window having a size estimated to be the maximum size of the object in the monitoring area. When;
Counting means for counting the number of objects in the monitoring area based on the distribution of the area occupancy in the monitoring area.

Images