JP2003087773A - Monitoring device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a monitoring device which uses a monitoring camera image and which more precisely detects the abnormality of the object of monitoring. SOLUTION: The monitoring device obtains a differential image from an image having the time difference of an image obtained from the monitoring camera and detects abnormality from the feature quantity of the differential image. The device is provided with a means 23 dividing the image obtained from the monitoring camera into mesh shapes, a means 24 generating the differential image where differences are given for the respective meshes of the image having the time difference, which are divided into the mesh shapes, and means 25, 26 and 27 judging abnormality from the feature quantity of the differential image.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention monitors, for example, smoke and flames generated in an abnormal state of various equipment installed in a building, or abnormal operation of a movable part of equipment such as a belt connected to a motor. The present invention relates to a monitoring device that detects and notifies from a camera image.

[0002]

2. Description of the Related Art In a conventional device of this type, for example, in Japanese Unexamined Patent Publication No. 9-322153, in automatic monitoring of a device based on an image from a surveillance camera, a differential image is obtained from an image having a time difference, or a time difference is obtained. It has been proposed that a difference image is obtained from an input image which is a certain image and a background image, geometrical features such as an area and a center of gravity are obtained from the difference image, and an abnormality is detected accordingly.

[0003]

However, in these monitoring devices, when the difference from the background image is small, an abnormality may not be detected, or the image may be mistakenly recognized due to a sudden change in the image. It could not be said that it was a monitoring device with a certain nature.

Generally, in automatic monitoring by a monitoring camera, a plurality of devices are often monitored, and a device having a short distance from the monitoring camera and a device having a long distance may have different sensitivity for abnormality detection. For this reason, there is a problem that the monitoring results may differ. Further, since there is a large number of blind spots in a long-distance device and it is sometimes required to monitor abnormality with high sensitivity, there is not a sufficient device capable of coping with this.

On the other hand, in this type of monitoring device, when a change occurs from a stationary state, it is considered as an abnormality, and in a monitoring area where there is always a change, it is impossible to monitor without setting it as a monitoring area, and in practice monitoring is performed. I couldn't.

The present invention has been made in order to solve the above problems, and it is an object of the present invention to provide a monitoring device that more accurately detects an abnormality in a device to be monitored in a monitoring device using a monitoring camera image. To aim.

[0007]

In view of the above object, the present invention provides a monitoring apparatus for obtaining a difference image from images having a time difference between images obtained from a monitoring camera and detecting an abnormality from a feature amount of the difference image. And a means for dividing an image obtained from the surveillance camera into a mesh shape, a means for generating a difference image for each mesh of the time-divided images divided in the mesh shape, and a difference image And a means for making an abnormality determination based on the characteristic amount of.

Further, the image dividing means includes a background image generating means for temporarily holding an image from the surveillance camera to generate a background image for monitoring, which is an image having a time difference from the current image. Both the background image and the current image are divided into meshes each having a predetermined number of pixels, and the difference image generating means obtains the difference image from the difference between the background image and the current image for each mesh.

Further, the number of pixels and the gradation of the image inputted from the surveillance camera can be set, and the image inputting means provided with the image dividing means for inputting the image with the preset number of pixels for each mesh, Background image generating means for generating a background image for monitoring, which is an image having a time lag with respect to the current image by temporarily holding the input image, and the difference image generating means includes the background image and the current image. It is characterized in that a difference image is obtained from the difference of each mesh for each image.

Further, it is characterized by further comprising means for detecting a positional shift of the surveillance camera from a change in brightness of image data of a boundary line of the mesh of the difference image.

Further, the camera position shift detecting means is characterized in that the camera position shift is judged from the luminance change rate or the change duration time.

Further, the abnormality determining means determines that the abnormality is detected when the difference in the difference image cannot be detected in the area of the movable portion to be monitored.

Further, it is characterized in that the area around the movable part to be monitored is monitored with high sensitivity.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described below in accordance with embodiments. FIG. 1 is a diagram showing the concept of a monitoring device according to the present invention. In the present invention, the analog image obtained by the surveillance camera CA is converted into the digital image D by the A / D converter.
P, and each of these digital images DP is divided into meshes. The current image D, which is an image with a time difference
A difference image DP3, which is obtained by calculating a difference for each mesh between P2 and the background image DP1 for monitoring, is generated, and abnormality determination is performed based on the difference image DP3.

FIG. 2 shows the structure of a monitoring device according to an embodiment of the present invention. In FIG. 2, reference numeral 10 is a surveillance camera for imaging the surveillance area, 11 is an image processing device for processing an image signal input from the surveillance camera 10 to detect an abnormal state,
An alarm device 13 reports an abnormality occurrence in response to an alarm signal from the image processing device 11.

In the image processing apparatus 11, 20 is an image input unit for A / D converting the image signal from the monitoring camera 10 and sending it to the frame memory 30, and 21 is color information for the image signal of the frame memory 30. , A background image updating unit that updates the background image 32 with the image data 31, and an image 23 that obtains a divided image 33 obtained by dividing the image data 31 and the background image 32 into a mesh shape. It is a division unit. Here, the background image 32 is a reference for a normal state having a time difference from the current image (image data 31) obtained by temporarily storing the input image and partially updating the input image with a new image. The image dividing unit 23 divides the image data 31 and the background image 32 into meshes.

Reference numeral 24 is a difference image generation unit for generating a difference image 34 showing a difference for each mesh of the divided images 33 of the image data 31 and the background image 32, and 25 is a binary value obtained by performing a labeling process on the difference image 34. A binary image generation unit that obtains the binarized image 35, a feature amount generation unit that calculates a feature amount 36 such as the area of the changed portion, the center of gravity, and the center point from the binarized image 35, and 27 indicates an abnormal state from the feature amount 36. An abnormality determining unit that determines whether or not there is an alarm, 28 is an alarm output unit that outputs an alarm signal to the alarm device 13 when the abnormality determining unit 27 determines that an abnormality has occurred, and 41 is the position of the monitoring camera 10 from the difference image 34. A camera position shift detection unit 42 for detecting a shift is a memory for temporary storage used for this camera position shift detection.

The image processing apparatus 11 basically comprises a computer that operates according to a program.

Next, the operation will be described. Surveillance camera 1
The image signal of the monitoring area obtained by imaging 0 for every predetermined time is stored in the frame memory 30, the color information is removed by the noise removing unit 21 to reduce the calculation amount, and the median filter processing or the like is applied. By doing so, fine noise on the image is removed to obtain the image data 31. The background image updating unit 22 partially updates the background image 32 with the image data 31.

Then, the image division unit 23 divides the image data 31 and the background image 32 into meshes each having a predetermined number of pixels to generate a divided image 33. This image division unit 23
May be provided at any stage from when the image signal is input from the monitoring camera 10 to when the difference image generating unit 24 generates the difference image 34. For example, the image signal of the monitoring area obtained by capturing images with the monitoring camera 10 at every predetermined time may be input for each mesh of the number of pixels preset by H / W,
In this case, the image division unit 23 is provided in the surveillance camera 10 or the image input unit 20 as shown by the broken line. Further, in this case, the subsequent processing may be performed in mesh units in each processing. Further, the surveillance camera 1 is provided with an image dividing means 23 that allows the number of pixels and the gradation of an image input from the surveillance camera to be set, and that inputs an image with a preset number of pixels for each mesh.
0 or the image input unit 20 may be used. When there are multiple surveillance objects and these surveillance distances are different, the sensitivity for long-distance surveillance is reduced when processing is performed with the same number of pixels, but this is because the area per pixel is different. In order to make the distance closer to that of the object, the number of pixels in the mesh in the long-distance monitoring area is increased as compared with that of the short-distance object, and the sensitivity of the short distance and the distance can be made closer. Furthermore, even if the monitored object is in the shadow of another object, that is, in a so-called blind spot, and it is not possible to directly monitor the place where an abnormality is expected, the smoke and flame will spread,
Increase the sensitivity around the monitored object (set a large number of pixels)
Inevitably, abnormality monitoring can be performed.

In the difference image generator 24, a divided image 3 of the image data 31 and the background image 32, which are divided into meshes, is formed.
The difference image 34 in which the difference for each mesh of 3 is obtained is generated. Further, for the difference image 34, the binary image generation unit 25
Then, image calculation such as labeling processing is performed to obtain a binarized image 35. In this case, there is also a monitoring environment in which the processing of the noise removing unit 21 becomes unnecessary if the label that does not meet a certain reference value is deleted. Further, by reducing the number of labels, it is possible to reduce the calculation load of the feature amount 36 described below.

In the feature quantity generation unit 26, the area, center point, and barycentric position of the portion of the binarized image 35 that has changed with respect to each label, and further, the contour length and the vertical length at the barycentric position. , A geometrical feature amount 36 such as the horizontal length of the position of the center of gravity is calculated.

The abnormality determining section 27 determines whether or not there is an abnormality in the monitored object based on the calculated feature amount. When the abnormality is recognized, an alarm signal is generated from the alarm output unit 28,
The alarm device 13 outputs an alarm.

In addition, the camera position shift detecting section 41 cuts out an image div for each boundary line of the mesh from the difference image 34, compares the data with the prestored installation data, that is, the position shift of the surveillance camera, that is, An abnormality in the installation position condition is detected, and an alarm is issued via the alarm output unit 28. In the camera position shift detection unit 41, the monitoring camera 10
Brightness information for one vertical or horizontal line of the boundary line of the mesh set in advance at the time of installation of the monitor or setting of the monitoring area is stored in the memory 42, and the difference images 34 input one after another are stored.
The change in the installation position or direction of the monitoring camera 10 is detected by comparing the brightness information for one line in the vertical or horizontal direction at the same position.

Further, in order to eliminate the possibility of misidentification, FIG.
The image data is cut out from each of the boundary lines of the mesh or the preset measurement line from the difference image 34 obtained during a certain time as shown in FIG. By detecting the position shift of the monitoring camera 10 from the change rate or the change duration time,
Prevents false positives.

Further, for example, as shown in FIG. 4, an area A of a movable portion which is constantly moving, such as a belt B which is hung between gears G on both sides.
In the case of R1, it is necessary to make a determination opposite to the above determination. For example, in the case of a movable part, it is normal for the abnormality determination unit 27 to always generate a feature amount having a value of a certain value or more (that is, a difference is detected in a difference image) in the case of a normal operation, If the feature amount is less than a certain value, it is determined to be abnormal.

On the other hand, in the case of such a belt or the like, an area AR2 which does not normally change around the area AR1 of the movable portion.
In addition, changes in the features appear due to slack in the belt or partial breakage. Therefore, for example, in the abnormality determining unit 27, an area AR that does not normally change around the area AR1 where the movable portion exists.
2 is detected with high sensitivity, for example, by lowering the reference value for the determination, and damage to the belt or the like is detected. In addition to increasing the sensitivity, in addition to changing the setting of the number of pixels in the mesh, for example, in the difference image generating unit 24 to the feature amount generating unit 26, the mesh in the image dividing unit 23 is made finer. It can also be implemented by changing the criteria in various judgments.

Furthermore, if a series of processes from the noise removing unit 21 to the feature amount generating unit 26 is performed for each mesh, the abnormality is reported when an abnormality is detected in a certain mesh, and the image is detected. It is possible to reduce the calculation amount of processing.

[0029]

As described above, according to the present invention, there is provided a surveillance device which obtains a difference image from images obtained from a surveillance camera with a time difference and detects an abnormality from the feature amount of the difference image. A means for dividing the image obtained from the surveillance camera into a mesh shape, a means for generating a difference image for each mesh of the time-divided images divided into the mesh shape, and a feature amount of the difference image Since the monitoring device is characterized by including means for determining abnormality from
Since a judgment is made for each mesh obtained by dividing the image, a more accurate monitoring device can be provided.

Further, there is provided background image generation means for temporarily holding an image from the surveillance camera to generate a background image for monitoring which is an image having a time difference from the current image, and the image division means is provided with the background image generation means. Since the background image and the current image are both divided into meshes each having a predetermined number of pixels, and the difference image generating means obtains the difference image from the difference of each mesh between the background image and the current image. Since a judgment is made for each mesh obtained by dividing, it is possible to provide a more accurate monitoring device.

Further, the number of pixels and the gradation of the image inputted from the surveillance camera can be set, and the image inputting means provided with the image dividing means for inputting the image with the preset number of pixels for each mesh, Background image generating means for generating a background image for monitoring, which is an image having a time lag with respect to the current image by temporarily holding the input image, and the difference image generating means includes the background image and the current image. Since the difference image is obtained from the difference between each image and each mesh, the determination is made for each mesh obtained by dividing the image based on the image having appropriate sensitivity, so that a more accurate monitoring device can be provided.

Further, since a means for detecting the positional deviation of the surveillance camera from the brightness change of the image data of the boundary line of the mesh of the difference image is further provided, by recognizing the positional deviation of the surveillance camera, it is possible to intentionally It is possible to recognize the movement of the surveillance camera in the installation direction, etc., and to prevent misidentification of abnormalities due to this.

Further, since the camera position deviation detecting means determines the camera position deviation from the luminance change rate or the change duration time, the position deviation can be accurately detected.

Further, since the abnormality determining means determines the abnormality when the difference in the difference image cannot be detected in the area of the movable portion to be monitored, the abnormality of the movable portion can be monitored.

Further, since the area around the movable portion to be monitored is monitored with high sensitivity, the abnormality of the movable portion can be monitored more reliably.

[Brief description of drawings]

FIG. 1 is a diagram showing a concept of a monitoring device according to the present invention.

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

FIG. 3 is a diagram for explaining detection of positional deviation of the surveillance camera according to the present invention.

FIG. 4 is a diagram for explaining abnormality detection of a movable part to be monitored according to the present invention.

[Explanation of symbols]

10 surveillance camera, 11 image processing device, 13 alarm device, 20 image input unit, 21 noise removing unit, 22 background image updating unit, 23 image dividing unit, 24 difference image generating unit, 25 binary image generating unit, 26 feature amount Generator, 27
Abnormality determination unit, 28 alarm output unit, 30 frame memory, 31 image data, 32 background image, 33 divided image, 34 difference image, 35 binarized image, 36 feature amount, 41 camera position deviation detection unit, 42 memory.

Continued front page    F-term (reference) 5C054 AA01 CA04 CC02 CH01 EA05                       FC00 FC01 FC15 FF06 HA01                       HA18                 5C087 AA02 AA03 AA08 BB02 BB74                       DD08 DD21 DD22 DD23 DD27                       DD33 EE08 FF01 FF02 FF04                       FF19 GG02 GG06 GG18

Claims (7)

[Claims] 1. A monitoring device which obtains a difference image from images having a time difference between the images obtained from the surveillance camera and detects an abnormality from a feature amount of the difference image, wherein the image obtained from the surveillance camera is mesh-shaped. And a means for generating a difference image for each mesh of time-divided images divided into meshes, and a means for making an abnormality determination from the feature amount of the difference image. A monitoring device characterized by the above. 2. A background image generation unit for temporarily holding an image from a surveillance camera to generate a background image for monitoring, which is an image with a time difference from the current image, and the image division unit is provided with the background image generation unit. Both the background image and the current image are divided into meshes each having a predetermined number of pixels, and the difference image generation means obtains a difference image from a difference between the background image and the current image for each mesh. Item 1
The monitoring device described in 1. 3. An image input unit having an image dividing unit capable of setting the number of pixels and gradation of an image input from a surveillance camera, and inputting an image with a preset number of pixels for each mesh, and Background image generating means for generating a background image for monitoring, which is an image having a time lag with respect to the current image by temporarily holding the input image, and the difference image generating means includes the background image and the current image. The monitoring device according to claim 1, wherein a difference image is obtained from a difference between each image and each image. 4. The device according to claim 1, further comprising means for detecting a positional shift of the surveillance camera from a change in brightness of image data of a boundary line of the mesh of the difference image. Monitoring equipment. 5. The monitoring device according to claim 4, wherein the camera position shift detecting means determines the camera position shift based on a luminance change rate or a change duration time. 6. The monitoring device according to claim 1, wherein the abnormality determining means determines an abnormality when a difference in the difference image cannot be detected in a region of a movable portion to be monitored. . 7. The monitoring device according to claim 1, wherein a region around the movable part to be monitored is monitored with high sensitivity.