JPH05284501A - Intrusion monitor picture device - Google Patents

Abstract

PURPOSE:To provide the intrusion monitor picture device which detects an intrusion and reduced erroneous reports and can monitor the intrusion in a wide range. CONSTITUTION:Feature data of the area, the movement direction, and the movement speed of an object acquired in a taken-in monitor area is extracted by a detection object feature extracting part 8. With respect to this extracting method, the processing picture and the background are differentiated, and the labeling processing is performed with the part, where components of '1' of respective binarized pictures are connected, as one cluster to obtain areas and centroids. These obtained centroids are compared to extract the movement speed in accordance with the movement direction of the object, the number of traversed picture elements, and time intervals of picture taking-in. Extracted feature data is compared with intruder feature data, which is preliminarily stored in a set data memory part 9, by a feature data comparing part, 12; and if they coincide with each other, the object is detected as an intruder to raise an alarm.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an intrusion monitoring image device for automatically and highly accurately detecting an intruder (object) to be detected from a surveillance area image captured by a surveillance camera.

[0002]

2. Description of the Related Art FIG. 5 shows the configuration of a conventional intrusion monitoring image device. In the figure, 1 is a surveillance camera, 2
Is an input part of the surveillance area image captured from the surveillance camera 1,
Reference numeral 3 is a background memory unit for storing a background image, 4 is a difference processing unit for performing a difference between a processed image newly captured from the monitoring camera 1 and the background image stored in the background memory unit 3, and 5 is a difference processing unit. A binarization unit that binarizes the difference result,
Reference numeral 6 denotes an intrusion determination unit that determines the presence or absence of an intruder (object) based on the result of binarization, and 7 denotes an intrusion detection output unit that outputs an alarm when the intrusion determination unit 6 detects an intruder (object). ..

Next, the operation will be described. Surveillance camera 1
To input the monitoring area image to the input unit 2. The image captured at a certain time is stored in the background memory unit 3 as a background image, and a new monitoring area image is captured from the monitoring camera 1. The difference processing unit 4 calculates the difference between this input image and the background image stored in the background memory unit 3. The binarization unit 5 binarizes the difference result with a preset threshold value. The intrusion determination unit 6 outputs an alarm from the intrusion detection output unit 7 when an abnormality occurs when the number of pixels whose binarization result is 1 is greater than or equal to a certain size.

[0004]

Since the conventional intrusion detection image device is constructed as described above, all those which cause a certain density change on the screen are detected as abnormal, and the leaves shake. It also issued warnings for many false alarms such as birds, shadows of buildings, and paper scraps. Further, since the difference image is also binarized according to a threshold value to be set in advance, many false alarms due to the influence of noise components have occurred. In addition, it will not be used outdoors when the environment changes drastically because it will give false alarms about changes in concentration due to light, such as sudden changes in sunshine, lighting of lights, reflection of sunlight due to water pools, incidence of headlights, etc. It is possible, and there is a problem that it is possible to perform reliable detection only on limited objects such as when the door in front of you is open, the window glass is broken, or people pass by, etc. There was a point.

The present invention has been made to solve the above problems, and reduces false alarms due to swaying of leaves, birds, shadows of buildings, paper scraps, etc. The purpose is to improve the detection accuracy.

Another object of the present invention is to provide a highly accurate intrusion monitoring image device in which false alarms due to noise components are reduced and false alarms due to abrupt density changes due to incident light are also reduced.

Further, the present invention comprises one surveillance camera,
It is an object of the present invention to obtain an intrusion monitoring image device that enables wide area intrusion monitoring.

[0008]

SUMMARY OF THE INVENTION An intrusion monitoring image device according to the present invention is a feature of an intruder (object) to be detected in order to accurately detect the intruder (object) to be detected and reduce false alarms. A setting data memory unit that stores data, a detection target feature extraction unit that extracts captured target feature data, and a feature data comparison unit that compares / determines the set data with the extracted feature data Is.

Further, the intrusion monitoring image device according to the present invention is
An automatic binarization processing unit for automatically determining an optimum binarization threshold, and a density range of a detection target are set in order to detect the detection target with high accuracy, and noise components and light outside the range are set. It is provided with a binarization execution range setting unit for removing the influence of the incidence of light.

Further, according to the present invention, in order to monitor the intrusion of a wide area, the camera turntable for turning the surveillance camera 1 and the angle information of the camera in order to know the surveillance area expected by the surveillance camera 1 as the angle information. It has a section.

[0011]

In the intrusion detection according to the present invention, the binarized object of the difference image is subjected to labeling processing to obtain the area and the center of gravity of the object. Further, the feature data is extracted by a method of tracking it and obtaining the moving direction and moving speed of the object. By comparing the actually extracted characteristic data with the preset characteristic data in the characteristic data comparison unit, an alarm is output when the characteristic data is matched.

The intrusion detection according to the present invention is automatic.
In the binarization processing unit, an optimal binarization threshold value is obtained by using the binarization threshold value determination method based on the average number of neighbors for the difference image. Further, the binarization execution range setting unit sets the range for binarization and removes the influence of the component outside the range, so that the influence of noise components and the incidence of light can be reduced.

Further, in the intrusion detection according to the present invention, in order to associate the characteristic data of the intruder (object) to be detected with a plurality of monitoring areas, a certain monitoring area is detected. The angle information of the surveillance camera is read, and the characteristic data set in the surveillance area is stored in the setting data memory in association with the angle information. Then, when the surveillance camera turns to that angle, the intrusion detection is executed based on the corresponding setting data, so that it becomes possible to monitor the invasion over a wide area with one surveillance camera.

[0014]

EXAMPLES Example 1. An embodiment of the present invention will be described below with reference to the drawings. In FIG. 1, 8 is a detection target extraction unit for extracting the area (size) of the detection target, the moving speed, and the moving direction, 9 is a setting data memory unit for setting the characteristic data of the intruder (object) to be detected, and 12 is It is a characteristic data comparison unit for comparing the set characteristic data and the characteristic data of the target actually captured.

Next, the operation will be described. In FIG. 1, the detection target feature extraction unit 8 performs a labeling process on the connected component that has become 1 as a result of the binarization unit 5 as a block, and obtains the number of pixels as a planarity (size), Furthermore, the center of gravity of the mass is sought. Further, a new processed image is taken in after a fixed time, and the binary result after the background difference is similarly subjected to the labeling process to obtain the area and the center of gravity. At this time, blocks having substantially the same area in the two processed images are associated with each other as the same object. The direction in which the target has moved is extracted from the positional relationship of the centers of gravity of the respective blocks, and the moving speed of the target is extracted from the time interval at which the processed image is captured and the number of pixels in which the centers of gravity of the blocks have moved. 3 of the area (size), moving direction, and moving speed of the intruder (object) to be detected in the setting data memory unit 9
One feature data is stored and the feature data comparison unit 12 is stored.
The data extracted in step 3 is compared with the set data, and if they match, the intrusion detection output section 7 outputs an alarm.
By this method, only an object having a certain area (size) and moving in a certain direction at a certain speed can be detected. Therefore, noise components having different areas (sizes), small animals, birds having different speeds, etc. are not detected, and false alarms are reduced.

Example 2. Next, another embodiment of the present invention will be described with reference to the drawings. In the figure, 10 is an automatic binarization processing unit for automatically determining an optimum binarization threshold value by using a binarization threshold value determination method based on the average number of neighbors for the result of the difference processing unit 4, 11 Is designed to accurately detect the detection target and remove the effects of noise components and the incidence of light.
A binarization execution range setting unit that sets a density range in which binarization is performed.

FIG. 3 shows an example of an image obtained by binarizing the image after the difference by the automatic binarization processing unit 10 and the binarization execution range setting unit 11. FIG. 3A shows an arbitrary difference image. FIG. 3B shows a histogram of pixels having each density value in the image. FIG. 3C shows an example in which the binarization execution range is set to the density range of the detection target so that only the detection target can be detected well, and the automatic binarization processing unit 10 is further illustrated.
It shows an example of the optimum binarization threshold value obtained by. FIG. 3 (d) is the same as the one shown in FIG. 3 (c).
This is an image in which binarization is executed in the binarization execution range and components outside the range are removed. FIG. 3 (e) is an image obtained by further performing optimal binarization processing on the image of FIG. 3 (d).

Next, the operation will be described. First, the method of determining the average adjacent threshold value in the automatic binarization processing unit 10 will be described. When binarizing a grayscale image including a bright object (high density value) against the background while raising the threshold value,
The arrangement state of 0 and 1 in each binary image gradually changes. At this time, focusing on the component of 1, it can be intuitively considered that the threshold value at which the connected component is extracted as a “collected” region is an appropriate value. Here, the “collected” area refers to a connected component with few jagged edges and holes inside the area. That is, it can be said that a shape having a low ratio of boundary points to internal points is a well-organized shape.

Therefore, as the evaluation function expressing the goodness of the unity, the length of the boundary line is one candidate, but in order to reflect different boundary patterns in the evaluation function, 2 × 3 × 3 is used.
In the value pattern, when the center is 1, the number of 1s in the 8 neighborhoods is called the number of adjacencies. Then, the adjacency number of the internal points is 8, and the adjacency number of the boundary points takes a value from 0 to 7 depending on the shape. Therefore, if the sum of the numbers of adjacencies for one point is calculated and this value is divided by the total number of points to be defined as the average number of adjacencies, the higher the ratio of internal points, the larger the average number of adjacencies.

In a binary image with respect to a certain threshold T, the number of 1 points whose adjacency number is i (i = 0 to 8) is a
_{Expressed as i} (T), the sum A (T) of the numbers of adjacencies and the area S
(T) becomes equation (1) and equation (2), respectively, and the average number of adjacent neighbors R (T) is represented by equation (3).

[0021]

[Equation 1]

When binarizing while changing the threshold value T, a binary image of all 1s is obtained at the threshold value 0. As the threshold value is increased, holes and the like are generated in the binary image due to the influence of background noise, the ratio of boundary points is increased, and R (T) is decreased. As the influence of this background noise disappears, the target is extracted as a “lumped” region, and R (T) increases. After that, since the ratio of the internal points occupying the area becomes low, R (T) decreases again. At this time, R (T) has a minimum value and a maximum value, and reaches a maximum value at a threshold value at which the target is most gathered, so that point is determined as an appropriate threshold value. The optimum binarization threshold value is obtained by this method.

Next, the drawings will be described. In the example of the difference image in FIG. 3A, there are light incidents by the intruder A ₀ , the noise N _0, and the headlight L ₀ , and each exists as a cluster of concentrated densities. FIG. 3B is a result of taking a histogram for the density value, and a peak is generated for each component due to a difference in each density. Thus the binarization execution range, valid for component L ₁ of the intruder, by setting a good concentration ranges convenient to remove components L ₁ by and incident noise component N ₁ and light, unnecessary The influence of the components can be eliminated. FIG. 3C shows an example of the binarization execution range, and an image in which the area outside the range is removed is shown in FIG.
2 is a removed image outside the binarization execution range. Furthermore, FIG.
An example of the optimum binarization threshold value determined by the automatic binarization processing unit 10 in FIG. 2 is shown in (c). The optimum threshold binary image of FIG. 3E shows the result obtained by performing binarization with the threshold S in the set binarization execution range Z. In this way, unnecessary components can be removed, and an intruder (object) can be detected accurately.

Example 3. Further, another embodiment of the present invention will be described with reference to the drawings. In FIG. 4, 13 is a rotary base of the surveillance camera, and 14 is an angle information section for obtaining angle information of the surveillance camera.

Next, the operation will be described. In FIG. 4, the camera turntable 13 turns the surveillance camera 1. Then, when the surveillance camera 1 takes in a surveillance area, the surveillance area is known as camera angle information by the camera angle information section 14, and the angle information and the set characteristic data are stored in the setting data memory section 9 in association with each other. .. By performing this for a plurality of monitoring areas, characteristic data suitable for each monitoring area can be set, and wide-area intrusion monitoring can be performed.

[0026]

As described above, according to the present invention, the characteristic data of the target captured in the monitoring area is extracted and compared with the characteristic data of the intruder (object) to be detected which is set in advance. As a result, the accuracy of intrusion monitoring can be improved.

Further, according to the present invention, since the binarization execution range can be set and the optimum binarization threshold value is used, the influence of unnecessary components can be removed, and highly accurate intrusion detection can be performed. it can.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an intrusion monitoring image device according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram of an intrusion monitoring image device according to a second embodiment of the present invention.

FIG. 3 is an example of a processed image according to a second embodiment of the present invention.

FIG. 4 is a configuration diagram of an intrusion monitoring image device according to a third embodiment of the present invention.

FIG. 5 is a configuration diagram of a conventional intrusion monitoring image device.

[Explanation of symbols]

 8 Detection Target Feature Extraction Section 9 Setting Data Memory Section 10 Automatic Binarization Processing Section 11 Binarization Execution Range Setting Section 12 Feature Data Comparison Section 13 Camera Rotating Platform 14 Camera Angle Information Section

Claims (2)

[Claims] 1. A surveillance camera, a background memory section for storing a background image, a difference processing section for performing a difference between a surveillance area image from the surveillance camera and a background image in the background memory, and an intruder (object) to be detected. ) Has the feature data,
A setting data memory unit that sets the area (size), moving speed, and moving direction, a binarizing unit that binarizes the difference result of the difference processing unit, and an output of the binarizing unit are input, The detection target feature extraction unit that extracts the feature data of the target captured in the surveillance area captured from the surveillance camera is compared with the output of the setting data memory unit and the output of the detection target extraction unit, and when the two match. An intrusion monitoring image device including a feature data comparison unit that determines an intruder (object). 2. In the binarization process for detecting an intruding object, in order to automatically determine an optimum binarization threshold value,
Automatic 2 using the binarization threshold decision method by the average number of neighbors
The binarization processing unit and the density difference result are set to the density range in which the binarization is performed as the density range of the detection target, and the effects of noise components and light incident outside the range are removed to intrude. The intrusion monitoring image device according to claim 1, further comprising a binarization execution range setting unit for improving the accuracy of person (object) detection.