JP5279459B2 - Congestion detector - Google Patents

Description

  The present invention relates to an apparatus for detecting a crowded state of people at a place where people gather, such as an escalator entrance / exit and an elevator hall.

  In Patent Document 1 below, a change area in a monocular camera image is extracted as an outline image of only a person, and the density (density) of the person is calculated from the area on the image, and the amount of movement of the change area on the image It is described that the moving speed is calculated from the above. Note that the changed area is subjected to differential processing between an unattended background image and an input image that have been taken in advance, and the changed area is extracted.

  When a background object such as an escalator step is moving, or when the background brightness is apparently changed due to a change in sunlight, the above method may extract them as a change area. In this case, since the area that is not a person is included in the change area, an error occurs in the measurement value of the density or movement speed of the person, and the congestion state cannot be detected correctly.

  In addition, the two-dimensional image to be photographed is the one in the three-dimensional real space mapped to a two-dimensional plane with radial non-parallel rays, so the apparent measurement value on the two-dimensional image is It cannot be said that the physical state in the real space is detected as it is, and the congestion state can be detected correctly.

  In Patent Document 2 below, the height in the three-dimensional space is calculated for each image block constituting the image of the object captured by the camera by stereo image processing using the image of the monitoring space photographed by the stereo camera, It describes that the degree of congestion is determined based on the total area of image blocks whose calculated height is within a predetermined range. Furthermore, it is also described that the area composed of the image blocks extracted in this way is projected three-dimensionally on a virtual plane parallel to the floor surface, and the degree of congestion is determined from the size of the projected area. In either case, the degree of congestion is determined from information in a two-dimensional plane, and the degree of congestion is not determined directly from information in a three-dimensional space.

JP 2007-55727 A JP 2001-34883 A

  An object of the present invention is to provide a congestion detection device capable of more accurately grasping the degree of congestion.

  The above-mentioned purpose is to determine a three-dimensional shape of a group of persons in a monitoring space by applying a stereo image processing to a plurality of imaging devices that capture a predetermined monitoring space and images obtained by the plurality of imaging devices. The present invention is achieved by a congestion detection device including a stereo image processing unit that performs a congestion degree determination unit that determines a congestion degree based on the determined three-dimensional shape.

  The aforementioned object is to determine a plurality of vectors representing the movement of a person group in a monitoring space based on a plurality of imaging devices that photograph a predetermined monitoring space and images obtained by the plurality of imaging devices at different times. And means for calculating at least one of the magnitude of the movement of the entire human population in the monitoring space and the degree of variation of the movement from the plurality of determined vectors, and the calculated magnitude of the movement of the entire human population, and This can also be achieved by a congestion detection device including a congestion degree determination unit that determines a congestion degree based on the degree of variation in motion.

  Since the degree of congestion is determined directly from the information in the three-dimensional space obtained by using a plurality of imaging devices, the degree of congestion can be determined more accurately.

  In addition, by appropriately defining the monitoring space, a changing background such as an escalator step can be easily excluded from the detection target.

  FIG. 1 is a diagram showing an example of camera arrangement and monitoring space setting in a congestion detection apparatus according to an embodiment of the present invention. In the illustrated example, the monitoring space 16 is set in a range that does not include the handrail 12 and the floor surface 14 of the escalator 10 at the entrance of the escalator 10. Two cameras 18 and 20 are installed above the monitoring space 16 to photograph a group of people in the monitoring space 16.

  FIG. 2 shows a configuration of a congestion detection apparatus according to an embodiment of the present invention, and FIG. 3 shows a schematic flowchart of processing in the computer 22 in FIG. In FIG. 2, images taken by the two cameras 18 and 20 are captured by the computer 22 (step 1000 in FIG. 3). The computer 22 applies stereo image processing to the captured image, that is, calculates the distance from the cameras 18 and 20 based on the principle of triangulation, and determines the three-dimensional shape of the person group in the monitoring space 16 ( Step 1002).

  FIG. 4 is a diagram for explaining the principle of obtaining three-dimensional coordinates based on the principle of triangulation. As shown in FIG. 4, one point P (X, Y, Z) of the subject is photographed with two cameras. An image pl (xl, yl) is imaged on the imaging surface of the left camera, and pr (xr, yr) is imaged on the imaging surface of the right imaging device. The position of the point P (X, Y, Z) is obtained by searching for a corresponding point between the two imaging devices. At this time, if the focal length of both imaging devices is F and the distance between both imaging devices is B, the three-dimensional coordinates X, Y, Z of the point P can be obtained by the following equations.

X = B (xl + xr) / 2d
Y = B (yl + yr) / 2d
Z = BF / d
Where d = xl−xr

  The broken line 26 in FIG. 4 is called an epipolar line, and there is a corresponding point pr of pl on this line, so the corresponding point is searched for on this line. In the search, for example, a point having the smallest sum of luminance differences between corresponding pixels in a region in the vicinity of each point is set as the corresponding point.

  Since the three-dimensional shape of the person group occupying the monitoring space 16 is determined as a group of the three-dimensional coordinates determined in this way, the volume of the person group is determined therefrom. The volume occupancy is determined by dividing the volume of the person group by the volume of the monitoring space 16 (step 1004). Although it is not necessary to separate a plurality of persons individually, it is possible to calculate the number of persons in the monitoring space by separating the persons individually. Further, the weight of the person group occupying the monitoring space 16 can be determined by multiplying the volume by a predetermined density. Since the portion on the back side when viewed from the camera is not reflected on the camera, three-dimensional coordinates cannot be obtained. However, in the example shown in FIG. 1, the cameras 18 and 20 are arranged above the monitoring space 16 (for example, embedded in the ceiling). Therefore, sufficient information can be obtained for determining the degree of congestion.

  Since the data of the three-dimensional shape is obtained, instead of calculating the volume occupation ratio, the area of the cross section by the virtual plane parallel to the floor surface 14 and at a predetermined height from the floor surface 14 is calculated, The occupation ratio may be obtained by dividing by the bottom area of the monitoring space. Even in this case, a more accurate value can be obtained than the conventional method obtained from a non-parallel light beam mapped to a two-dimensional plane. Although two cameras 18 and 20 are sufficient, three or more cameras may be used to increase the accuracy of association.

  Next, the three-dimensional flow of each point of the person group is calculated from a plurality of stereo images along the time series (step 1006). Specifically, a three-dimensional flow can be obtained by associating each point constituting the three-dimensional shape determined as described above with one at a different time. In addition, “Jin Seki, Masatoshi Okutomi,“ Simultaneous estimation of motion and depth of dynamic scenes using stereo video ”, Image Recognition and Understanding Symposium MIRU 2006, pp. 352-357, includes three-dimensional coordinates and flow. Is calculated at the same time, and the calculation of the three-dimensional shape (step 1002) and the calculation of the three-dimensional flow (step 1006) may be performed simultaneously. Even when it is performed separately, the calculation of the volume occupation ratio in step 1004 may be performed after the average speed calculation in step 1008.

  As shown schematically in FIG. 5, the three-dimensional flow starts from each point constituting a reference three-dimensional shape, or a representative point determined at a fixed interval, and the corresponding point in the three-dimensional shape at a subsequent time. It is expressed by a plurality of vectors with ending at. In step 1008 (FIG. 3), the average value of the lengths of these vectors (or the length of the average vector) is set as the average speed of the person group in the monitoring space 16.

  In step 1010, the congestion state is determined from the occupation rate obtained in step 1004 and the average speed obtained in step 1008. For example, when the occupation ratio is equal to or higher than a predetermined value and the average speed is equal to or lower than the predetermined value, it is regarded as being congested because there are many people and the movement is slow. Although the determination is most simply made in two stages of a congested state and a normal state, the determination level may be set in multiple stages as shown in FIG. The multi-stage setting method is not limited to that shown in FIG. 6 and can be arbitrarily set in a two-dimensional coordinate system of occupation ratio and average speed. The degree of congestion can also be expressed numerically.

  According to the present embodiment, by obtaining the three-dimensional shape in the predetermined monitoring space by stereo vision, the background and the person can be accurately separated even when the background is fluctuating, As a result, it is possible to correctly calculate the occupation ratio and the moving speed without being affected by the background. In addition, since the three-dimensional shape is obtained, the occupation rate based on the volume can be obtained for the monitoring space, and the physical movement in the real space is directly obtained by the three-dimensional flow. Can be measured correctly.

  Instead of determining the degree of congestion based on the combination of the occupation rate and the average speed as described above, the degree of congestion can also be determined based on only the occupation rate or only the average speed. Also, instead of or together with calculating the average velocity from a plurality of vectors representing the movement of a person group, the degree of variation of the vector representing the movement is calculated, and based on this, or in combination with the occupation ratio, etc. The degree can also be determined.

  The degree of variation of the velocity vector is calculated as follows, for example.

Here, λ ₁ represents the variation in the direction in which the variation is the largest in the monitoring space (the direction of the axis of the first principal component), and this is the degree of variation in the velocity vector. A large variation in the velocity vector means that it is crowded and the flow of people is not smooth, and a small variation in the velocity vector means that the flow of people is not crowded and the flow of people is smooth. Therefore, the greater the degree of variation, the higher the degree of congestion.

  Instead of using a three-dimensional vector, you can use a two-dimensional vector with only a horizontal component, that is, a projection of a three-dimensional vector on a horizontal surface in the vertical direction. good.

  The determination of the degree of congestion may be performed on the weight calculated by multiplying the volume of the person group by the density instead of the above or on the number of persons calculated by separately separating the persons included in the person group. You may go on a basis.

  The determination result of the degree of congestion in step 1010 (FIG. 3) is displayed on the display 24 (FIG. 2). If it is determined that the traffic is congested, an alarm is output or transmitted to another device (step 1012).

  It is also possible to divide the monitoring space 10 (FIG. 1) into a plurality of partial spaces, determine the degree of congestion in each partial space as described above, and determine the congestion bias based on the result.

It is a figure which shows an example of the setting of the monitoring space. It is a block diagram which shows the structure of the congestion detection apparatus which concerns on one Embodiment of this invention. It is a flowchart of the process in the computer 22 of FIG. It is a figure explaining the principle of a stereo image process. It is a figure explaining the some vector showing the motion of a person group. It is a figure explaining determination of the congestion degree based on an occupation rate and an average speed.

Claims (4)

A plurality of imaging devices for photographing a predetermined monitoring space;
Stereo image processing means for determining a three-dimensional shape of a group of persons in the surveillance space by applying stereo image processing to images obtained by a plurality of imaging devices;
A congestion degree determining means for determining a congestion degree based on the determined three-dimensional shape ,
The stereo image processing means is configured to represent points of a reference three-dimensional shape representing a movement of a group of persons in the monitoring space based on images obtained by the plurality of imaging devices at different times, or at regular intervals. And further determining a plurality of vectors starting from the representative point determined in step 3 and ending at the corresponding point in the three-dimensional shape at the subsequent time,
Means for calculating at least one of the magnitude of motion and the degree of motion variation of the entire human population in the monitoring space from the determined vectors;
The congestion degree determining means is a congestion detection device for determining a congestion degree based on the three-dimensional shape and at least one of the calculated magnitude of movement of the entire person group and the degree of variation in movement . A plurality of imaging devices for photographing a predetermined monitoring space;
Based on images obtained by multiple imaging devices at different times, each point that constitutes a reference three-dimensional shape that represents the movement of a group of people in the surveillance space , or a representative point that is set at regular intervals And a means for determining a plurality of vectors whose end points are corresponding points in the three-dimensional shape at a subsequent time
Means for calculating at least one of the magnitude of motion and the degree of motion variation of the entire human population in the monitoring space from the determined plurality of vectors;
A congestion detection device comprising: a congestion degree determining means for determining a congestion degree based on at least one of the calculated magnitude of movement of the entire person group and the degree of variation in movement. 3. The congestion detection apparatus according to claim 1, further comprising means for outputting an alarm when the congestion level determined by the congestion level determination unit exceeds a predetermined threshold. The monitoring space is divided into a plurality of partial spaces, and the congestion degree determining means determines the congestion degree for each individual partial space, and further reduces the congestion bias based on the congestion degree determined for each individual partial space. The congestion detection device according to any one of claims 1 to 3 .

Images