WO2015159412A1

WO2015159412A1 - Moving-body detection method and moving-body detection system

Info

Publication number: WO2015159412A1
Application number: PCT/JP2014/060980
Authority: WO
Inventors: 浩一後迫
Original assignee: 株式会社日立システムズ
Priority date: 2014-04-17
Filing date: 2014-04-17
Publication date: 2015-10-22

Abstract

When determining whether or not a moving body has crossed a plurality of virtual straight detection lines drawn on image data such as surveillance images, only one-dimensional movement such as left-right movement is identified, and it is difficult to also identify two-dimensional movement. As such, in this invention, each frame of image data captured by an imaging means is partitioned into a plurality of image blocks, a plurality of concentric circular virtual lines are superimposed on said image data, and at prescribed intervals, each image block that contains a virtual line in at least part of said image block is scanned a prescribed number of times. By analyzing the path of a moving body, said path being obtained from scanned image blocks that exhibit color changes in excess of a threshold, a determination as to whether said moving body is passing through or dwelling is made.

Description

Moving object detection method and moving object detection system

The present invention relates to a method and system for analyzing and detecting the movement of a moving object using image data acquired from imaging means such as a monitoring camera.

Conventionally, in surveillance systems using surveillance cameras installed in commercial facilities such as supermarkets and shopping centers, transportation facilities such as stations and airports, and amusement facilities such as theme parks, people and objects in the captured surveillance images are used. A method has been used in which the movement of the mobile phone (hereinafter sometimes referred to as “moving body”) is detected and notified to the observer.

In this case, as a notification method based on motion detection, only the presence / absence of the movement of the moving body was used, and the position and moving direction of the moving body in the monitor image screen could not be determined. For this reason, for example, it has not been possible to use the monitoring image of the monitoring camera for purposes such as determining whether or not a person has entered the prohibited entry area and whether or not to enter or leave a specific venue.

Recently, one or more detection lines are virtually drawn on the monitoring image, and it is checked whether the moving object has crossed this line. A mechanism that can determine the moving direction of a moving body has been provided (for example, Patent Document 1).

JP 2010-9134 A

A plurality of virtually linear detection lines are drawn on the image data such as the monitoring image, and whether or not the moving object crosses each detection line is determined by the time timing difference (time lapse) across the plurality of detection lines. ). However, in this case, as the moving direction of the moving body, for example, as shown in FIG. 4, only a one-dimensional direction such as from left to right (or from right to left) is determined. It is difficult to distinguish.
Further, in order to detect the moving direction and stay of the moving body, for example, the determination is made by detecting temporal changes such as color tone in all image blocks of one frame of the screen. However, in this case, since all image blocks must be scanned a predetermined number of times, it takes time to scan and is not efficient.

The features of the present invention are as follows. The image data captured by the imaging means is divided into a plurality of image blocks for each frame, a plurality of concentric virtual lines are superimposed on the image data consisting of the image blocks, and the virtual data is formed on at least a part of the image block. All image blocks including lines are scanned a predetermined number of times every predetermined time. By analyzing the trajectory of the moving body obtained from the image block exhibiting a color change exceeding a predetermined threshold in the scanned image block, it is determined whether the moving body is a passing object or staying.

According to the present invention, by using image data such as a monitoring image, the moving direction of a person or an object as a moving object and the flow line including the moving direction can be quickly and easily two-dimensionally (three-dimensionally as necessary). Can be detected. As a result, security measures (for example, detection of entry into prohibited areas) and safety protection measures (for example, determination of whether the vehicle is moving along the route and reverse running detection), as well as marketing (for example, the density of human flow) There is an effect that it can be applied in the direction of information grasping) and traffic control (for example, whether there is a stagnation or traffic jam).

FIG. 1 is a diagram illustrating a system configuration in which functional blocks of a moving body detection system according to an embodiment are illustrated. FIG. 2 is a flowchart illustrating a process of analyzing the movement of the moving object from the monitoring image data according to the embodiment. FIG. 3 is a schematic diagram in the case where the flow line of the moving body is detected by concentric virtual lines. FIG. 4 is a schematic diagram in the case of detecting a flow line of a moving body using a linear virtual line. FIG. 5 is a diagram illustrating a case where concentric virtual lines are square lines. FIG. 6 is a diagram illustrating a case where concentric virtual lines are circular lines.

First, a method for analyzing and detecting the moving direction of the moving body or the presence or absence of staying according to the present invention will be described.
The feature of the present invention is that a plurality of concentric virtual images are included in image data photographed by an imaging means such as a surveillance camera in order to determine the moving direction of the moving body two-dimensionally (three-dimensionally as necessary). The line is drawn and image data is scanned along the virtual line. The details of the processing flow will be described later, but the outline will be described below.

The image data photographed by the imaging means such as a monitoring camera is divided into appropriate block sizes, and a plurality of concentric virtual lines are drawn from the center of the image data for the divided block image data. Then, each image block obtained by dividing the image data is scanned for a plurality of concentric virtual lines, and a change in color tone such as color and brightness of the scanned image block is detected. Or the presence or absence of a stay is determined.

For example, if a change in color tone that is equal to or greater than a predetermined threshold is continuously detected in the same or adjacent concentric virtual lines in the same or adjacent image block, it is determined that the moving body is staying. be able to.
Alternatively, when a change in color tone of a predetermined threshold value or more is detected from the outer concentric virtual line toward the inner concentric virtual line, the moving body is in a moving state and each of the color tone changes. The moving direction can be determined from the block number of the image block.

Furthermore, by counting the number of times an image block has passed, it is possible to discriminate not only a simple or linear movement but also a moving body having a complicated movement such as a zigzag.
Further, regarding the time required for scanning, by narrowing the image blocks to be scanned to concentric virtual lines, the scanning time can be shortened and the processing efficiency can be improved.

As the closed curved line to be scanned, a concentric quadrangular line as shown in FIG. 5 can be assumed instead of a concentric circle as in the present invention, but the distance in the oblique direction becomes longer and the distance from the center depending on the direction Since these are not equidistant, the calculation process becomes complicated in calculating the moving direction and the like, which is not a good idea.

In FIG. 3, a plurality of concentric virtual lines are expressed in an elliptical shape because it is assumed that the ground surface is imaged at an angle by an overhead surveillance camera installed near the ceiling or near the ceiling. is there. According to the surveillance camera facing directly below from the overhead such as the ceiling, it becomes a concentric circle as shown in FIG. It shows the place where it becomes elliptical.

Next, the processing content that is a feature of the present invention will be described.
First, image data acquired by an imaging unit such as a monitoring camera is divided into m1 × m2 image blocks. Here, m1 and m2 are parameters for determining the block size of the image data in accordance with the size of the moving object. As the values of m1 and m2 are larger, the block size is smaller and a small moving object can be detected.

Next, virtual lines defined as C1, C2, C3... Cn are drawn from the center of the image as a plurality of concentric lines for the image data divided into blocks. Here, n is a parameter indicating the accuracy for detecting the moving direction of the moving body. The larger the n, the more accurately the moving direction of the moving body can be grasped. FIG. 3 shows concentric imaginary lines C1, C2, and C3 in the case of n = 3.

Under the above concentric virtual lines and block size, each image block of image data is P times as a predetermined number of times per predetermined unit time along the concentric circles C1, C2, C3... Cn. to scan. Here, P is a parameter indicating the detection accuracy according to the moving speed of the moving body. It becomes possible to detect even a moving body whose moving speed is faster as P is larger.

When there is a change in color tone such as color or brightness in the scanned image block, it is determined whether or not the degree of change exceeds a threshold value T. The threshold T can be detected even when the value is smaller even when the change in color tone is small, and it is necessary to consider the surrounding environment (for example, lighting and solar radiation) and to set an appropriate value according to the change. Sex may come out. It is also possible to use a predetermined color tone area ratio in the image block as a threshold value. For example, if it is 100%, it is not detected that the entire image block is not changed, and if it is 50%, it is detected that the area of more than half of the image block is changed.

FIG. 1 is a diagram showing functional blocks when applied to a monitoring apparatus using two monitoring cameras as an embodiment of a moving body detection system according to the present invention. Assuming a three-dimensional space composed of XYZ axes as the monitoring images taken by the two monitoring cameras A10 and B20, for example, when monitoring images in the XZ plane direction and monitoring images in the YZ plane direction, or For example, a monitoring image of the XY plane (that is, the ground surface) taken from the upper side of the Z-axis vertically downward and a monitoring image of the vertical plane direction from the X-axis or Y-axis are assumed.

The functional blocks (11 to 19) for processing the monitoring image data captured by the monitoring camera A10 and the functional blocks (21 to 29) for processing the monitoring image data captured by the monitoring camera B20 are both the same in the monitoring apparatus 1. 1 (block surrounded by a two-dot chain line in FIG. 1), this will be described with reference to a functional block diagram on the monitoring camera A10 side.

Image data captured by the monitoring camera A10 (B20) is stored as a monitoring image in the monitoring image recording device 13 (23) by the image data recording unit 12 (22) via the image data input unit 11 (21). Are input to the image block dividing unit 14 (24). Although the monitoring

image recording devices

13 and 23 are illustrated on the monitoring camera A side and the B side in FIG. 1, respectively, the monitoring

image recording devices

13 and 23 may be one monitoring image recording device on the A side and the B side.

In the image block dividing unit 14 (24), image data for one frame of the screen is divided into m1 × m2 image blocks. Here, as the image data to be input, the image data input unit 11 (21) directly captures as described above, and is temporarily stored in the monitoring image recording device 13 (23) via the image data recording unit 12 (22). It is possible to arbitrarily adopt a case of capturing at an appropriate timing from the state.

As shown in FIG. 3, the virtual line creation unit 15 (25) applies n concentric virtual images composed of C1, C2, C3,..., Cn to the image data divided into blocks in one frame of the screen. Create a line.

The moving body scanning unit 16 (26) scans all the image blocks that approximate the concentric virtual lines with respect to the image data obtained by superimposing the created (n) concentric virtual lines. Here, the image block approximating the concentric virtual line includes the concentric virtual line in a part of the image block (in other words, the concentric virtual line is slightly applied in the image block). ) Refers to the image block.

The moving object detection unit 17 (27) changes color tone such as color and brightness from the scanned image block through a predetermined number of times (P times) of scanning every predetermined time (S time) by the moving object scanning unit 16 (26). Detects an image block that exceeds a predetermined threshold (T). Thereby, the movement of a person or an object is detected as a moving body.

The detection block storage unit 18 (28) generates the occurrence time, the image block number, and the image for all the image blocks in which the color change such as color and brightness detected by the moving object detection unit 17 (27) exceeds the threshold (T). Necessary data such as concentric virtual line numbers closest to the block are stored. In addition, although the detection block memory |

storage parts

18 and 28 are illustrated in the monitoring camera A side and B side in FIG. 1, respectively, it does not interfere at all as one detection block memory | storage part in A side and B side.

The trajectory analysis unit 19 (29) detects temporal changes (transitions) in color tone between image blocks from all the image blocks detected / stored by the detection block storage unit 18 (28). As a result, information on the trajectory such as the moving direction of the moving body is analyzed.

The correlation analysis determination unit 31 analyzes both the trajectories in a wide area and three-dimensionally in consideration of the correlation between both image data based on the analysis result (trajectory) input from both the

trajectory analysis units

19 and 29. Then, the movement of the moving body (flow line including the moving direction), the presence / absence of staying, and the like are determined comprehensively. When this correlation is taken into consideration, necessary correction processing such as alignment of the coordinates of the respective image data and alignment of the scales is performed in accordance with the relative positional relationship between the two monitoring cameras.

Based on the analysis result from the correlation analysis determination unit 31, the analysis result classification unit 32 classifies the movement of the moving object (two-dimensional or three-dimensional movement direction including the height direction, stay in a specific range, etc.). .

When the above constituent requirements are roughly classified, the image data input unit 11 (21) to the moving object scanning unit 16 (26) execute processing relating to the image block and can be said to be a block processing unit. Further, the moving object detection unit 17 (27) to the determination result classification unit 32 determine the movement of the moving object by obtaining and analyzing the trajectory of the moving object, and can be regarded as an analysis determination processing unit.

The notification unit 33 notifies the monitoring display terminal 2 of necessary information regarding movement based on the classified analysis results. The analysis information is displayed on the screen of the monitoring display terminal 2 using characters and graphics. Moreover, it is also possible to notify by voice together with display as necessary.

In FIG. 1, the monitoring image recording device 13 (23) and the detection block storage unit 18 (28) are shown as separate recording media. However, it is of course possible to use one recording medium.
In FIG. 1, two monitoring cameras are used. However, even if one monitoring camera is used, an analysis of a moving line or a stay of a moving body in a two-dimensional plane such as the ground surface or a floor surface, or 2 It is possible to analyze the movement of the moving body up and down in the vertical dimension.

In addition, when three or more monitoring cameras are installed, one or two of them can be selected in combination as appropriate, and the flow line of the moving body can be analyzed as described above. . Furthermore, it is also possible to analyze the flow line and stay of the moving body in more detail or with high accuracy by using all three or more surveillance cameras. However, in that case, necessary correction processing such as alignment of coordinates of each image data and alignment of scales is performed in accordance with the relative positional relationship between three or more monitoring cameras.

Next, a processing flow for analyzing the movement of the moving body (flow line including the moving direction, staying, etc.) from the image data such as the monitoring image by each functional block of FIG. 1 will be described.
FIG. 2 is a diagram illustrating a flowchart of processing for analyzing the movement of a moving object from image data captured by an imaging unit such as a monitoring camera.

In step S <b> 101, the image data input unit 11 (21) acquires image data for one frame of a screen captured by an imaging unit such as a surveillance camera.
In step S102, the image block dividing unit 14 (24) divides the acquired image data of one frame of the screen into m1 × m2 image blocks (m1 and m2 are positive integers and variable parameters indicating detection accuracy). ), And assign coordinates to each image block. At that time, the block size (the size of m1 and m2) can be made variable in a form proportional to the size (size) of the moving body. That is, if the size (size) of the moving body is large, the block size can also be increased (m1 and m2 are decreased, but the balance with detection accuracy is taken into consideration).

In step S103, the virtual line creation unit 15 (25) determines an integer multiple of the radius r from the center of the image data of one frame of the screen (r × 1, r × 2, r × 3,... R × n, where n is a positive value. N concentric virtual lines are created every r × n ≦ m1 / 2 and r × n ≦ m2 / 2). At that time, the number n can be varied in proportion to the detection accuracy of the trajectory of the moving body.
In step S <b> 104, the virtual line creation unit 15 (25) stores all the coordinates of a block that approximates each of these n concentric virtual lines (hereinafter referred to as “concentric line approximate block”).

In step S105, the moving body scanning unit 16 (26) scans all the concentric line approximate blocks.
In step S106, the moving body scanning unit 16 (26) adds 1 to the number of scans (+1).

In step S107, the moving object detection unit 17 (27) determines whether or not a change in color tone such as color and brightness has exceeded a threshold value T from the previous scan in the concentric line approximate block scanned this time.
If the image block indicates a change exceeding the threshold T (Yes), the moving object detection unit 17 (27) stores the coordinates of the image block in the detection block storage unit 18 (28) in step S108. To do. For an image block that does not exceed the threshold T (No), the process proceeds to step S111 (wait for S seconds).

Next, in step S109 subsequent to step S108, the moving object detection unit 17 (27) stores that the image block that shows a change exceeding the threshold T in the current scan has changed beyond the threshold T in the previous scan. It is determined whether the image block is close to the coordinates of the image block. If it is determined that it is not an adjacent image block (No), in step S110, the coordinates of the image block stored in step S108 are deleted, and the process proceeds to step S111 (weight of S seconds). This determination step is provided in order to determine and exclude a change in color tone of an image block that is not close as being erroneously detected due to noise or the like.

If it is determined in step S109 that the image blocks are close (Yes), the process proceeds to step S111 as it is, and the moving body scanning unit 16 (26) waits for S seconds. Here, S seconds indicates a predetermined time interval of the scanning process.
After the S second wait, in step S112, the moving body scanning unit 16 (26) determines whether or not the predetermined number of scans has been performed. If the execution has not yet been performed P times (No), the process returns to step S105, and the process up to step S111 (the above-described scan process) is repeated.

When the P scanning processes are completed (Yes), in step S113, the trajectory analyzing unit 19 (29) converts the trajectory of the moving object for the P scans from the coordinates of all the stored image blocks. Further, as shown in FIG. 1, when there are two systems of image data captured from the monitoring camera, the processing from step S101 to step S113 is executed in each system. Then, the trajectory data of the moving body obtained from each system is sent to the subsequent step S114.

In step S114, the correlation analysis determination unit 31 analyzes the trajectory of the moving body converted into data, and determines whether the trajectory is a straight line or a curved line that changes in a specific direction with continuity. At this time, if the data trajectory is taken from two systems, necessary correction processing such as alignment based on the mutual coordinate relationship and scale alignment is performed, analyzed in three dimensions, and the above determination is made. Do.

If the above condition is satisfied (Yes), in step S115, the analysis result classification unit 32 determines that the moving body is a passing object, and calculates the moving direction. In subsequent step S116, the notification unit 33 notifies the monitoring display terminal 2 of the moving body as a passing object together with its moving direction.

On the other hand, if the above condition is not satisfied (No), in step S117, the analysis result classification unit 32 determines that the moving body is a staying object staying in an area in the photographed screen. In subsequent step S118, the notification unit 33 notifies the monitoring display terminal 2 of the moving body as a staying object.

As the display mode of the monitoring display terminal 2 that has received the above notification, the moving direction of the moving object can be quickly displayed by superimposing the displayed image data (monitoring image) with a virtual arrow or the like. It can be connected to grasp. In addition, various display patterns can be provided, such as counting the number of moving objects for each time period and displaying them by a statistical graph or the like.

In addition, when a jurisdiction area to be monitored includes a dangerous area, an entry prohibition area, or the like, a scene in which movement or reverse movement of a person or an object within the area is detected is assumed. At that time, on the monitoring display terminal 2 or the like, a warning message is displayed together with the monitoring image (highlighted display, blinking display, etc.), or a warning sound is uttered (sounded) in accordance with the monitoring image. To do.
Although the surveillance camera has been mainly described as the imaging means, the present invention is not limited to this. For example, a general imaging means such as a portable type or a handy type video camera can be used as a matter of course. It is.

DESCRIPTION OF SYMBOLS 1 ... Monitoring apparatus 2 ... Display terminal 10 for monitoring, 20 ... Surveillance camera (A, B)
DESCRIPTION OF

SYMBOLS

11, 21 ... Image

data input part

12, 22 ... Image

data recording part

13, 23 ... Surveillance

image recording device

14, 24 ... Image

block division part

15, 25 ... Virtual

line creation part

16, 26 ... Moving

body scanning unit

17, 27 ... Moving

body detection unit

18, 28 ... Detection

block storage unit

19, 29 ... Trajectory analysis unit 31 ... Correlation analysis determination unit 32 ... Analysis Result classification unit 33 ... notification unit

Claims

A first step of dividing image data captured by at least one of a plurality of imaging means into a plurality of image blocks for each frame;
A second step of superimposing a plurality of concentric virtual lines on the image data comprising the image block;
A third scan that scans all of the image blocks including the virtual line in at least a part of the image block a predetermined number of times every predetermined time and stores an image block exhibiting a color tone change exceeding a predetermined threshold among the scanned image blocks And the steps
A fourth step of obtaining a trajectory of the moving object in the image data from the stored image block;
And a fifth step of determining whether the moving body is a passing object or staying by analyzing the trajectory.
It is a moving body detection method of Claim 1, Comprising:
The fifth step determines that the moving body is a passing object when the trajectory of the moving body is a linear shape or a curved shape that changes in a specific direction with continuity, and otherwise, A moving body detection method, characterized by determining that a moving body is stagnant.
It is a moving body detection method of Claim 1 or 2,
When the image data is captured by two or more of the imaging means,
The first to fourth steps are executed for each image data,
In the fifth step, the movement including the three-dimensional movement of the moving body is analyzed by analyzing the locus obtained for the respective image data according to the correlation between the respective image data. A moving body detection method comprising: determining whether a body is a passing object or staying.
It is a moving body detection method in any one of Claim 1 to 3,
The image block in the first step changes the block size in proportion to the size of the moving object,
The virtual line in the second step is characterized in that the number of virtual lines is changed in proportion to the detection accuracy of the trajectory of the mobile body.
It is a moving body detection method in any one of Claim 1 to 4, Comprising:
A moving body line detection method comprising: a sixth step of displaying the result determined in the fifth step over the image data or notifying the sound in accordance with the display of the image data. .
A plurality of imaging means;
Recording analysis means for recording and analyzing image data taken by at least one of the imaging means;
Output means having a display and notification function,
The record analysis means includes
The image data is divided into a plurality of image blocks for each frame, a plurality of concentric virtual lines are superimposed on the image data including the image blocks, and the virtual lines are included in at least a part of the image block. A block processing unit that scans all the image blocks a predetermined number of times every predetermined time;
An image block exhibiting a color change exceeding a predetermined threshold is stored in the scanned image block, a trajectory of the moving object in the image data is obtained from the stored image block, and the movement is analyzed by analyzing the trajectory. An analysis determination processing unit for determining whether the body is a passing object or staying, and
A notification unit that notifies the image data and the determination result to the outside;
The output means includes
The moving object detection system, wherein the image data from the notification unit and the determination result are displayed in an overlapping manner, or the determination result is notified by voice in accordance with the display of the image data.
It is a moving body detection system of Claim 6, Comprising:
The block processing unit executes the process for each of the image data photographed by two or more imaging units,
The analysis determination processing unit analyzes the trajectory of the moving object obtained for each of the image data in accordance with the correlation between the respective image data, so that the three-dimensional movement of the moving object is also detected. In addition, it is determined whether the moving body is a passing object or stays.