JPS62136991A - Abnormality monitoring device - Google Patents

Abstract

PURPOSE:To eliminate complexity or inacuracy in the setting of a detection area by calculating a texture feature quantity at every micro area within a monitoring area, and setting automatically an area, the property of which is identified as different from other part in the texture feature quantity, on a detection area memory. CONSTITUTION:The patterns of power spectrum, such a a tree, a fence, a concrete wall, a ground, the sky, and the surface of a water, etc., are registered in advance at a detection area automatic setting means 7 as texture feature quantities, and the pattern of the power spectrum obtained at a texture arithmetic means 6 is compared and checked with each of registered patterns, and it is identified to which pattern is the micro area corresponding, and a data regarding a risk degree corresponding to an identified factor is corresponded at every picture element, then being registered at a detection area memory 3. For example, the area where the tree is present is set as the area of risk degree '0', the concrete wall the ground within a site as of risk degree '1', the fence as of risk degree '2', or the ground at a distant vie as of risk degree '0' because it is out of th4 site respectively. In such a way, a detection capacity against an invader can be improved.

Description

Detailed Description of the Invention (Technical Field) The present invention relates to an image recognition type abnormality monitoring device that detects the presence or absence of an abnormality in an area to be monitored using image input means such as a television camera, and mainly relates to an abnormality monitoring device for detecting an abnormality in an area to be monitored. In addition to crime prevention such as theft, it is used for fire detection and accident prevention due to abnormal occurrences in factories.

(Background Art) FIG. 4 is a block diagram showing the basic configuration of a conventional abnormality monitoring device proposed by the present inventors. The image signal of the monitoring area captured by the imaging device of the image human power means 1 is A/
After being D-converted, it is sent to the image processing means 2. The image processing means 2 performs pixel-to-pixel subtraction between a pre-stored reference image and the current image, and converts the image into an image in which only pixels that have changed in brightness have values. In the abnormality determination means 4, the presence or absence of an abnormality is determined by the inference section 41 from the information obtained by the image processing means 2 and based on the knowledge base 42 stored in advance. The output means 5 has a function of displaying the status when an abnormality determination is made by the abnormality determination means 4.

By the way, when trying to monitor an abnormality in a certain place using the above-mentioned abnormality monitoring device, the entire monitoring area does not have the same degree of danger; In this case, it is necessary to set the risk level high when a brightness change occurs.On the other hand, in areas that inherently include brightness changes such as trees and water surfaces, the risk level when a brightness change occurs needs to be set low. There is. Taking this point into consideration, in the conventional example shown in Fig. 5, the monitoring area is divided into multiple detection areas, and the degree of danger when a brightness change occurs within each detection area can be individually set. It's Nishide. The detection area setting means 31 performs this setting, and it is possible to set a detection area of any shape using a pointing device such as a light pen or a graphic tablet while looking at the reference screen. The shape and risk of each detection area are stored in the detection area memory 3.

In such a conventional example, the setting of the detection area was manually performed by a human, so there was a problem that the setting work was extremely complicated.

In addition, setting errors can occur unless the operator is a certain level of skill, and there is room for omissions or mistakes in settings, especially in monitoring areas that require the setting of a large number of detection areas. was strongly desired.

(Object of the invention) The present invention has been made in view of the above points, and
The purpose of this is to identify a detection area that has different characteristics from other parts by identifying the texture of the image obtained from the image input means, and to automatically set the detection area. To provide an abnormality monitoring device that eliminates the complexity and inaccuracy of settings.

(Disclosure of the Invention) As shown in Figure @1, the abnormality monitoring device according to the present invention includes an image input means 1 that images a monitoring area and quantizes an image signal; an image processing means 2 that compares the image with a reference image and obtains the information necessary for abnormality determination; a detection area memory 3 that sets a detection area with different properties from other parts within the monitoring area; Based on the knowledge for abnormality determination, an abnormality determination means 4 determines the presence or absence of an abnormality from the information obtained by the image processing means 2 and the stored contents of the detection area memory 3, and outputs the determination result. In the abnormality monitoring device including an output means 5, a texture operator stage 6 calculates a texture feature amount for each minute region within the monitoring region, and a texture characteristic amount obtained by the texture calculation means 6 determines the characteristics of other parts. Detection area automatic setting means 7 is provided for automatically setting an area identified as having a different value in the detection area memo +73.

That is, in the present invention, in order to automatically set the detection area, a texture calculation method that is well known in the technical field of image processing and image recognition is used. Sky,
Water surfaces, etc. are identified and a separate risk level is set for each.

Features for textured liquids generally include statistical values such as the average value and standard deviation of the first and second derivatives of images;
In addition to simultaneous concentration occurrence probability, concentration transition probability, and run occurrence frequency, autocorrelation functions and power spectra are used.

Preferred embodiments of the present invention will be described below with reference to the accompanying drawings. FIG. 1 is a block diagram of an abnormality monitoring device according to an embodiment of the present invention. This embodiment differs from the conventional example shown in FIG. 5 in that texture calculation means 6 and detection area automatic setting means 7 are added. The texture calculating means 6 includes a power spectrum calculating means for calculating the power spectrum of the image as the texture vf@ quantity, and is configured to obtain the power spectrum for each minute area within the monitoring area. In the detection area automatic setting means 7, power spectrum patterns of trees, fences, concrete walls, ground, sky, water surface, etc. are registered in advance as texture special weights, and the power spectrum patterns obtained by the texture calculation means 6 are registered in advance. is compared with each of these registered patterns to identify whether the micro area corresponds to a tree, fence, concrete wall, ground, sky, water surface, etc., and to determine the degree of danger according to the identified element. Data is registered in the detection area memory 3 in association with each pixel.

For example, in the monitoring area shown in Figure 2, if the power spectrum is determined for each minute area, (a) in Figure 3 is obtained.
A pattern as illustrated in ~(c) is obtained. In FIG. 3, the graph on the left shows the power spectrum in the horizontal direction of the screen (referred to as the X direction), where the horizontal axis represents the frequency and the vertical axis represents the power 1Fx12 for that frequency component.

Also, the graph on the right side is in the vertical direction of the screen (assumed to be the Y direction)
The horizontal axis is the frequency f, and the vertical axis is the power IFsI for that frequency component.
Represents +2. Figure (a) shows the power spectrum of an area with little change, such as a concrete wall or the ground. Figure (b) shows the power spectrum of an area that includes shaking, such as a tree or water surface. Figure (C) shows the power spectrum of an area that includes shaking, such as a fence. This is a power spectrum of a region having a grid pattern with regular intervals. Figure 3, (b)
As shown in , the power spectrum of trees, etc. has high values in the high frequency region in both the X and Y directions. Also, Figure 3 (C
), the power spectrum of 7ence has a high value in the high frequency region only in the X direction where the lattice pattern exists.

In the scene shown in FIG. 2, trees within the monitoring area are likely to cause malfunctions when monitoring for abnormal occurrences such as intruders. That is, it is possible that the brightness changes within the screen due to the sway of the trees due to the wind, resulting in a detection signal similar to that generated when an intruder is present. Therefore, in order to prevent malfunctions, the area including such trees is set to have a lower degree of danger than the surrounding areas, or is made into a dead area where no detection signal is generated. Also, conversely, 7
The border areas of premises such as buildings will be set at a high level of risk as a warning, and the ability to detect intruders will be improved.

FIG. 2 shows how the risk level data for each detection area is automatically set in this way.

In the example shown in the figure, the area where trees exist is set to danger level (0), the concrete wall and the ground within the site are set to danger level (1), and the fence part is set to danger level (2). It is set. Furthermore, the ground that is more distant than the fence is set to a risk level (0) because it is an area outside the premises.

Next, the actual abnormality monitoring operation after the detection area is set will be briefly explained. Imaging device 1 of image input means 1
1, an image of the monitoring area is input to the image processing means 2. In this embodiment, the image processing means 2 is an image memo 1721.
．． 22 and an image processing section 23. Input image memory 2j
The current image obtained by the imaging device 11 is input to the image capturing device 11 . The reference image memory 22 includes an image capturing device? 111, an image of the monitoring area when there is no abnormality is input in advance as a reference image. The image processing unit 23 includes an input image memory 21
and the image in the reference image memory 22 by pixel-to-pixel subtraction to extract changes in the image. The abnormality determination means 4 determines in which detection area the abnormality has occurred based on the amount of change in this image and the contents stored in the detection area memory 3 that have been set in advance. The output means 5 issues an alarm based on the output from the abnormality determination means 4.

At this time, it is possible to issue a more appropriate warning by not only issuing a warning for the occurrence of an abnormality, but also by changing the warning level depending on the degree of danger. The alarm level setting memory 51 determines which danger level and alarm level to issue in advance.
If the alarm is generated while referring to the memory 51, the alarm can be generated as desired by the supervisor.

Note that although the automatic setting of the detection area is performed when the present device is installed, the settings may be updated at regular intervals even after installation.

Further, in the above embodiment, the detection head-mounted setting means 31 for manual setting is provided in the conventional manner, and the contents of the automatic setting can be appropriately corrected by manual fabrication.

(Effects of the Invention) As described above, in the present invention, the detection area can be almost automatically set using the identification result of the texture calculation means, so the effort required for setting the detection area can be greatly reduced. There is also an effect that the detection area can be set in a detailed and always optimal state even in a monitoring area that includes a complex scene.

[Brief explanation of drawings]

Fig. 1 is a block diagram of an abnormality monitoring device according to an embodiment of the present invention, Fig. 2 is a diagram showing an example of a reference image in the above embodiment, and Fig. 3 is an operation of the texture calculation means in the above embodiment. The explanatory diagram, FIG. 4, is a block diagram showing the basic configuration of a conventional example, and FIG. 5 is a block diagram of another conventional example. 1 is an image input means, 2 is an image processing means, 3 is a detection area memory, 4 is an abnormality determination means, 5 is an output means, 6 is a texture calculation means, and 7 is a detection area automatic setting means. First drawing image input means Image processing means Figure 3 f ff
f
f
fFigure 4/' Figure 5

Claims (1)

[Claims] (1) An image input means that images the monitoring area and quantizes the image signal, an image processing means that compares the image obtained by the image input means with a reference image and obtains information necessary for abnormality determination, and the monitoring area Based on the detection area memory that sets a detection area with different properties from other parts within the interior, and the knowledge for abnormality determination stored in advance, the information obtained by the above image processing means and the detection area memory are In an abnormality monitoring device including an abnormality determining means for determining the presence or absence of an abnormality based on stored contents, and an output means for outputting the determination result, a texture calculating means for calculating a texture feature amount for each minute area within a monitoring area; Detection area automatic setting means for automatically setting in the detection area memory an area identified to have different properties from other parts based on the texture feature amount obtained by the texture calculation means. Device.