JPH04273689A - Monitoring device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a monitoring device for detecting the presence or absence of a suspicious person or object within a monitoring area using image input means such as a television camera.

0002

[Prior Art] In the conventional monitoring device,
As described in No. 6988, by tracking the movement path of an object detected by image changes, image changes that do not move or image changes whose movement path is different from the intrusion path are removed as noise, and an intruder can be detected. Ta.

0003

[Problem to be Solved by the Invention] The above conventional technology usually has the following problems:
It is assumed that there are no people, and normal people will also come,
They did not take into consideration the possibility that suspicious persons would come. Furthermore, since image changes that do not move are removed as noise, there is a problem in that suspicious objects cannot be detected.

[0004] The purpose of the present invention is to treat people who are always normal,
It is an object of the present invention to provide a monitoring device that detects suspicious persons and suspicious objects in places where suspicious persons also come, such as places where automatic teller machines are installed, guidance centers, stations, and various other public facilities and conditions.

[0005]

[Means for Solving the Problems] In order to achieve the above object, the present invention provides an image input means for capturing an image of a monitoring area, and an image captured by the image input means.
An image processing means that performs various image processing to extract an object, and extracts the feature amount of the object, and the feature amount obtained by the image processing means is used to obtain the position information of the object, and the position information is Object tracking means for tracking an object, and information on the movement route and time of existence of the object obtained by the object tracking means to determine whether the object is a normal person, a suspicious person, a normal object, or a suspicious object. The structure includes an object determining means for determining whether the object is an object, and an output means for outputting the result of this determination.

[0006]

[Operation] According to the monitoring device according to the present invention, it operates to determine whether a normal person, a suspicious person, a suspicious object, etc. are detected by using the movement path, shape feature amount, and shape change rate of the object detected by image change. Therefore, even in locations and conditions where normal people, suspicious people, normal objects, suspicious objects, etc. coexist, suspicious persons and suspicious objects can be detected with high accuracy.

[0007]

[Embodiment] An embodiment of the present invention will be described below with reference to FIG.

The monitoring device according to this embodiment includes cameras 101a, 101b, 101 that take in images of the vicinity (inside and outside) of the installation location of an automatic teller machine 20 having an entrance 19.
It has a monitoring device main body 90 that processes the images captured in c and 101d and detects suspicious persons and objects.

The monitoring device main body 90 includes an image processing section 100 that extracts feature quantities of objects from input images, controls the entire device, processes the processing results of the image processing section 100, tracks objects, makes judgments, and issues alarms and images. CPU 10 controls the output of
9. Memory 110 for storing monitoring results, abnormality (suspicious person,
It consists of an alarm output section 111 that outputs an alarm in response to a command from the CPU 109 when a suspicious object (suspicious object) occurs, an image transmission section 112 that sends a monitoring image to a central monitoring center, and a timer 113 that measures time.

The image processing unit 100 includes a camera switch 102
, A/D converter 103, image memory 104, inter-image calculation circuit 105, binarization circuit 106, labeling circuit 107
, a feature extraction circuit 108.

The operation will be explained below.

The image memory 104 has, for example, 256×25
K 6-pixel gray scale memories G1 to Gk are provided, and j binary image memories B1 to Bj are provided to store binary images as required.

Based on a command from the CPU 112, the image processing section 100 takes in image signals taken by the cameras 101a to 101d, and selects one of the inputs from the four cameras using the camera switch 102. , A/D
The converter 103 converts the data into, for example, 128-gradation density data, and stores it in the image memory 104.

Furthermore, the image processing section 100 includes a CPU 1
12, the data in the image memory 103 is used to perform processing such as inter-image calculation, binarization, labeling, feature amount extraction, etc. in the inter-image calculation circuit 105 and the binarization circuit 1, respectively.
06, processing is performed by the labeling circuit 107, feature amount extraction circuit 108, etc.

Subsequently, the CPU 109 performs object tracking 6 and object determination 7, which will be described later, and instructs the alarm output unit 111 to output an alarm if there is an abnormality.

[0016] In this mechanism, in order to avoid camera blind spots, an example of installing a plurality of cameras including a camera 101a for obtaining a planar image, a camera 101b for obtaining a frontal image, and a camera 101c for obtaining a side image is installed. Although an example of installing a camera consisting of the indoor cameras 101a to 101c and the outdoor camera 101d has been shown, other system configuration examples include only the indoor cameras (101a to 101c), only the outdoor camera 101d, or even only the outdoor camera 101d. , camera input from only the planar image camera 101a is considered, and when only one camera is used, the camera switch 102 is not required.

Next, an outline of a method for extracting objects and detecting suspicious persons and objects using a background image will be explained.

FIG. 2 shows a conceptual diagram of this object extraction process. First, the image processing unit 100 obtains a difference image 13 between the input image 11 and the background image 12, binarizes this using a predetermined threshold value, creates a binary image 14, and then labels each object by labeling. 15. Then, for each object, feature quantities such as area, center of gravity coordinates, length and width of circumscribed rectangle, aspect ratio, perimeter length, etc. are extracted. Next, the CPU 109
An object having an area within a certain range is stored in the memory 110 as a candidate for a suspicious person or suspicious object together with the feature amount,
Using this information, the object is tracked 16 and it is determined whether the object is a suspicious person, a normal person, or a suspicious object 17. If the result of the judgment is that there is a suspicious person or suspicious object, the alarm output section 1
11, and outputs the information to the central monitoring center 18. Note that 20 out of 11 input images
2 represents an automatic teller machine, 21 represents a person, and the input image 11 and the background image 12 represent images of the automatic teller machine installation location taken from above.

FIG. 3 illustrates a series of movements of the monitored object. (a) in FIG. 3 shows the background image 12. (b in Figure 3)
) to (h) show the input image 11 at each time. It is assumed that these images are taken from above by the camera 101a. First, object 1 appeared from the entrance 19 and stood in front of the automatic teller machine 20 ((b)), then object 3 entered and placed object 2 ((c)). Next, object 1 leaves ((d)), object 3 moves to the left while holding object 2 ((e)), and then moves upward and stops ((f).
)). Then, object 3 places object 2 on the spot ((g)
), exit. Object 2 exists in the same place ((h))
. Note that in FIG. 3, only objects 1 to 3 indicated by diagonal lines are obtained as a binary image 14 by the image processing unit 100.

FIG. 4 shows object numbers corresponding to each time (b) to (h) in FIG. 40, Various feature amounts 41, Object No. at the previous point in time. 42, and a state determination result 43. Figure 4
In the column of the feature quantity 41, a circle indicates that the various feature quantities 41 have been determined, and a minus mark in each column indicates that there is no corresponding data. The data shown in FIG. 4 is obtained by obtaining a binary image 14 at each time as shown in FIG.
4, perform labeling processing to obtain various feature quantities 41 (area, center of gravity coordinates, length and width of circumscribed rectangle, aspect ratio, perimeter, etc.) of each label image, and also label each label at the previous point in time. This can be obtained by checking the overlap between the label images or determining the correspondence by taking into consideration the distance between the label images and the direction of movement. In addition, each object 4 in FIG.
The determination result 43 for the state 0 is determined by comparison with the object 42 at the previous point in time, and if a new object occurs, it is determined as "appearance" or "separation"; if the object disappears, it is determined as "disappearance" or "merging"; Even if the position of is changed, if it exists at both times, it will be a "match". Here, the discrimination between "appearance" and "separation" and "disappearance" and "merging" is based on the position and size of the target object, the rate of change in size, shape information, and the position and size of nearby objects at the previous point in time. This is done by comparing the rate of change in size, shape information, and the position of the entrance/exit. For example, the object moving to the left at time (e) in Figure 3 is known to be object 3, and based on the rate of change in size, it is determined that object 2 has "merged" with object 3. be done. On the other hand, between time (c) and time (d) and between time (g) and time (h),
Similarly to the above, it is considered that Object 1 and Object 3 have "disappeared" from the information on the position and size of each object and the position of the entrance/exit. Similarly, between times (f) and (g), it is determined that object 2 is "separated" from object 3. Note that, like 3' in FIG. 4(e), n' indicates that the object has merged with object n or separated from object n.

[0021] If the object that has "appeared" or "separated" does not change its position for a certain period of time as described above, it is determined that a suspicious object has been placed or that a suspicious object exists17, and the monitoring center An alarm is output from the alarm output unit 111 to 18.

FIG. 5 shows the movement path of each object determined from the position information of each object at each time in FIG. Reference numeral 51 represents the movement path of object 1, 52 represents the movement path of object 2, and 53 represents the movement path of object 3. Using these movement route information, for example, a normal area 24 and a suspicious area 25 are set in advance, and it is possible to check whether the movement path of each object is within the normal area 24 or passing through the suspicious area 25, and whether the object is in the same location. It is also possible to determine whether the object is normal 1, suspicious object 2, or suspicious person 3 based on presence time information (whether the object remains stationary in the same place for a certain period of time or more).

The above explanation is based on images captured from above using one camera 101a, but due to the positional relationship between the camera and the object, a blind spot occurs with respect to the camera, and some objects are Something that cannot be seen happens. One method to solve this problem is to use multiple cameras, for example, three cameras, one above (camera 101a), one in front (camera 101a),
b), image from the side (camera 101c) ((in Fig. 6)
a): Planar image, (b) in Figure 6: Front image, (c) in Figure 6: Side image), and from these three images, accurately determine the presence of each object and each of the above-mentioned features. You can ask for it. In (a) of FIG. 6, the size of object 3 is the same as that of object 2.
6 is constant regardless of the presence or absence of the object 26, but in (b) in FIG. 6, it can be seen that the object 3 is larger by the amount of the object 26. This change in size causes "merging" and "
It becomes possible to accurately judge "separation".

FIG. 7 shows fighting objects 27 and 28.
This shows an example where . In this example, time (a),
The shape change rate of objects 27 and 28 between (b) and (c) (
(at least one of the following: change in area, change in perimeter, change in length and width of circumscribed rectangle, change in aspect ratio of circumscribed rectangle) or rate of change in position coordinates is normal movement. It can be determined that there is abnormal movement such as a fight because it is large compared to .

Another embodiment for discovering suspicious objects is shown in FIG.
Shown below. FIG. 8 shows a case where a weighing scale 21 is installed at the entrance/exit 19 of the automatic teller machine installation location in contrast to FIG. 1. The weight of the moving object n is determined by using the weight scale 21 as a feature quantity of the object n, and from the difference (W1-W2) between the weight W1 of the object n when it enters the room and the weight W2 of the object n when it leaves the room, the weight of the moving object n is calculated. Determine whether a normal person, a suspicious person, and whether a suspicious object has occurred. It is assumed that tracking of the object n from entering the room to leaving the room is performed in the same manner as described above. The determination is made as follows based on the sign of the difference in weight (W1-W2) of the object n. However, the measurement error of the weighing scale 21 is taken into consideration.

(W1-W2)>0: A suspicious object has occurred, and the object n is a suspicious person.

(W1-W2)=0: No suspicious object has occurred, and the object n is a normal person.

(W1-W2)<0: A stolen item has occurred, and the object n is a suspicious person.

[0029]

[Effects of the Invention] According to the present invention, since a normal person, a suspicious person, a normal object, a suspicious object, etc. are determined using the movement path, shape feature amount, and shape change rate of an object detected by image change, Suspicious persons and objects can be detected with high accuracy even in locations and conditions where suspicious persons, normal objects, suspicious objects, etc. coexist.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a monitoring device that is an example of the present invention.

FIG. 2 is an explanatory diagram showing the flow of processing for monitoring suspicious persons and objects.

FIG. 3 is an explanatory diagram of the movement of an object in a monitoring target area.

FIG. 4 is an explanatory diagram of a table that stores feature amounts and states of objects.

FIG. 5 is a diagram showing a moving path of an object.

FIG. 6 is an explanatory diagram of a plane image, a front image, and a side view obtained by three cameras.

FIG. 7 is an explanatory diagram of state detection based on shape change rate.

FIG. 8 is a diagram showing another embodiment of the present invention using a gravimeter.

[Explanation of symbols]

100...Image processing section, 101...Camera, 109...CPU
, 111... Alarm output section, 112... Image transmission section, 11
3...Timer.

Claims (6)

[Claims] Claim 1: Image input means for capturing an image of a monitoring area;
image processing means for extracting an object by performing various image processing on the image captured by the image input means and extracting the feature amount of the object; An object tracking means that obtains the position information of the object and tracks the object using the position information, and the movement route information and the same location time information of the object obtained by the object tracking means determine whether the object is a normal person. 1. A monitoring device comprising an object determining means for determining whether an object is a suspicious person, a normal object, or a suspicious object, and an output means for outputting the determination result. 2. The monitoring device according to claim 1, wherein said image input means includes a plurality of cameras. 3. The monitoring device according to claim 2, wherein a plurality of cameras are arranged so as to obtain a plane image from above, a front image from the front, and a side image from the side. . 4. The monitoring device according to claim 1, wherein the object determining means determines whether the object is in a normal state or an abnormal state using a rate of change in shape of the object. 5. In the monitoring device according to claim 4, the rate of shape change is determined by determining the amount of change in the circumference of the object, the amount of change in area, the amount of change in length and width of a circumscribed rectangle, and the position of the center of gravity of the object per certain period of time. A monitoring device characterized in that at least one of the amounts of change in coordinates is determined. 6. The monitoring device according to claim 1, further comprising weight measuring means for measuring the weight of an object at an entrance/exit of the area to be monitored, in addition to the above-mentioned components, when the object enters/exits the area. A monitoring device characterized by determining the presence or absence of a suspicious person or suspicious object based on the difference in weight.