JP2003189294A - Image monitoring device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To detect soiling adhered to the lens surface of a camera with good accuracy in an image monitoring device. <P>SOLUTION: S110: the degree of an edge is calculated in complying with a variation in a brightness value against the peripheral pixels of a monitoring image for each pixel; S125: an N/M, of the number of pixels M exceeding a usual threshold value TH1 to the number of pixels N exceeding a high threshold value TH2 (TH2>TH1) is calculated to assign a high threshold value edge ratio A; decreasing the contrast of an image decreases the ratio A. The degree of its decrease is higher in the case soiling the lens surface or the like than the case decreasing illumination. Thus, S140: establishing an appropriate determination reference value TH4 to determine the presence of soiling adhesion in the case showing A is TH4 or less. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image monitoring apparatus for monitoring an area to be monitored based on an optical image, and more particularly to detecting stains or the like attached to a lens of a camera or the like for acquiring an image or a monitoring window. , Technology that makes it easy to deal with it.

[0002]

2. Description of the Related Art Conventionally, there is an image monitoring apparatus in which a camera is installed toward an area to be monitored and the appearance of an intruder is detected based on an optical image acquired by the camera. There are various monitoring target areas in which the cameras of the image monitoring apparatus are installed. Therefore, the lens surface of the camera and the surface of the monitoring window provided in the camera storage case may be contaminated indoors with, for example, a cigarette tar, or outdoors with dust and mud due to wind and rain, for example. is there. Such stains can blur or smear an image, making it unclear and obstructing surveillance based on the image.

As a conventional technique for detecting such dirt on the basis of an image acquired by a camera, Japanese Patent Laid-Open No. 200-200200
There is a technique disclosed in Japanese Patent Laid-Open No. 1-119614. In the related art, when an image becomes unclear due to stains, the steepness of change in pixel value (referred to as edge amount) in a portion (so-called edge) indicating an outer edge of an object such as a boundary between objects in the image decreases. By utilizing this fact, the dirt on the lens surface or the like is detected based on the fact that the edge amount becomes equal to or less than the predetermined value.

[0004]

However, in this conventional technique, since the stain amount is determined and detected by the decrease in the edge amount of the entire image, for example, when the monitoring target area is dark. There is a problem that it may be determined that stains are attached when the contrast of the entire image is low.

The present invention has been made in order to solve the above problems, and it is an object of the present invention to provide an image monitoring apparatus which is hardly affected by the environment in which a camera is installed and which can detect a monitoring obstacle such as dirt. To aim.

[0006]

The inventor of the present application has conducted research to realize an image monitoring apparatus that solves the above problems,
As a result of the experiment, if dirt adheres to the lens surface of the camera, the reduction of the area where the edge amount is large in the image is relatively large, while the reduction of the area where the edge amount is small is relatively small. On the other hand, it has been found that, due to illumination fluctuations, a portion with a normal edge amount and a portion with a larger edge amount change at the same degree. The present invention utilizes this knowledge.

An image monitoring apparatus according to the present invention is based on an image pickup means for generating an image corresponding to an optical image of a monitoring target area, and for each pixel of the image, based on a difference in pixel value between the pixel and a peripheral pixel of the pixel. A pixel value change strength calculating means for calculating a pixel value change strength; a first reference value obtained by summing the pixel value change strengths of the respective pixels having the pixel value change strength exceeding a first threshold value; Second greater than the threshold of
A second reference value obtained by summing the pixel value change intensities of the pixels having the pixel value change intensities exceeding the threshold value of, and determining the relative magnitude of the second reference value with respect to the first reference value. Comparing means for calculating a corresponding comparative evaluation value,
An obstacle detection unit that detects a semi-transparent monitoring obstacle in the optical path between the monitoring target region and the imaging unit based on the comparison evaluation value being smaller than a predetermined determination reference value. is there.

Another image monitoring apparatus according to the present invention is an image pickup means for generating an image corresponding to an optical image of a monitoring target area,
For each pixel of the image, a pixel value change strength calculating means for calculating a pixel value change strength based on a difference in pixel value with a peripheral pixel of the pixel, and the pixel value change strength exceeding a first threshold value. A first reference value, which is the number of pixels, and a second reference value, which is the number of pixels having the pixel value change intensity exceeding a second threshold value that is larger than the first threshold value, are obtained, and the first reference value is calculated. Comparing means for calculating a comparative evaluation value according to the relative size of the second reference value with respect to the monitoring target area and the imaging means based on the comparative evaluation value being smaller than a predetermined determination reference value. And an obstacle detection means for detecting a semi-transparent surveillance obstacle in the optical path between the and.

According to these aspects of the present invention, the pixel value change intensity represents the edge amount. A first threshold value that is relatively low and a second threshold value that is relatively high are set, and this is the number of pixels whose pixel value change intensity exceeds the first threshold value or the total value of the pixel change intensities of those pixels. The above knowledge is applied to the first reference value and the second reference value that is the number of pixels whose pixel value change strength exceeds the second threshold value or the total value of the pixel change strengths of those pixels. Thus, a semi-transparent monitoring obstacle such as dirt on the lens that causes blurring or haze of an image is detected. That is,
If the monitoring obstacle to be detected is present in the optical path between the monitoring target area and the image pickup device, the second reference value decreases relatively greatly, whereas the decrease of the first reference value relatively small. On the other hand, when the illumination is dark, both the second reference value and the first reference value decrease uniformly. As described above, in the case where there is a semi-transparent monitoring obstacle and the case where the illumination is dark or the like due to a contrast reduction factor, the second reference value with respect to the first reference value is used.
The relative size of the reference value is different. Therefore, an appropriate judgment reference value is set for the comparative evaluation value corresponding to this relative size, and when the image is blurred due to the presence of a monitoring obstacle such as dirt on the lens, or when the illumination is dark. It is possible to determine that the contrast is reduced due to a factor that is not caused by the monitoring obstacle.

A preferred aspect of the present invention is the image monitoring apparatus, wherein the comparison evaluation value is a ratio of the second reference value to the first reference value.

The semi-transparent surveillance obstacle may be cigarette smoke or the like existing in the optical path, but in a preferred aspect of the present invention, the obstacle detecting means is arranged on the optical path as the surveillance obstacle. The image monitoring device detects a stain attached to the transparent member or the reflecting mirror.

Empirically through experiments, the first threshold value corresponds to the contour of the detection target object located in the monitoring target area in the image acquired in the absence of the monitoring obstacle. It is preferable that the second threshold value has a value corresponding to twice the first threshold value, and the second threshold value has a value corresponding to the pixel value change intensity.

In the image monitoring apparatus according to the present invention, the obstacle detecting means determines that the dirt is attached when the comparison evaluation value is smaller than the determination reference value for a predetermined time. If the comparative evaluation value decreases due to a monitoring obstacle such as dirt adhering to the lens or the like, that state continues. For such a persistent monitoring obstacle, it is necessary to dispatch a coping person from the monitoring center or the like to perform the removal work. On the other hand, it is not necessary to dare to remove transient materials such as smoke and fog. According to the present invention, a persistent monitoring obstacle is detected, and, for example, a notification is sent to a monitoring center or the like based on the detection result.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION Next, an intruder monitoring apparatus according to an embodiment of the present invention will be described with reference to the drawings.

[First Embodiment] FIG. 1 is a schematic block diagram of an intruder monitoring apparatus according to the present invention. This device
It is composed of a camera 2 capable of capturing a surveillance target area and an image processing device 4 for processing a surveillance image obtained from the camera 2.

In addition to the intruder detection processing unit 10 which determines whether or not there is a shadow in the monitored area based on the monitoring image, the image processing apparatus 4 also includes optical components such as the lens of the camera 2 based on the monitoring image. It also has a stain detection processing unit 12 for detecting stains such as tars and dusts of cigarettes attached to the monitoring window.

The intruder detection processing section 10 detects an intruder based on various image recognition techniques that have been conventionally used, and notifies the inspector of the intruder detection. On the other hand, when the dirt detection processing unit 12 also detects dirt, the dirt detection processing unit 12 reports it to an observer. The intruder detection processing unit 10 and the dirt detection processing unit 12 are, for example, central processing units (CPU: Central Processing Unit).
Unit), and each of these processing units can be realized as a program executed on this CPU. Hereinafter, the dirt detection processing unit 12, which is a characteristic part of the present device, will be described in detail.

FIG. 2 is a processing flow chart showing the processing of the dirt detection processing section 12. 3 to 5 are explanatory diagrams for explaining the processing contents of the stain detection processing unit 12, and FIG. 3 is a schematic diagram of the monitoring target area. Further, FIGS. 4 and 5 are schematic graphs (FIGS. 4A and 5A) showing the state of fluctuation of the luminance value along the scanning line L shown in FIG. 3 and the luminance value, respectively. 5 is a schematic graph (FIGS. 4B and 5B) showing a variation of the spatial variation along the scanning line L. FIG. Figure 4
FIG. 5 is a diagram showing a state where dirt is not attached to the lens surface of the camera 2, while FIG. 5 is a diagram showing a state where dirt is attached and the contrast is lowered. In addition, in FIGS. 4A and 5A, the horizontal axis represents the arrangement of pixels in the scanning line direction, and the vertical axis represents the luminance value. In FIG. 4B and FIG. 5B, the horizontal axis represents the arrangement of pixels in the scanning line direction, the vertical axis represents the brightness value change amount, and the absolute value of this brightness value change amount is defined as the edge amount. To be done.

The dirt detection logic by the dirt detection processing unit 12 is started and executed at regular intervals (S100).
The dirt detection processing unit 12 acquires the image data input from the camera 2 (S105) and calculates the edge amount of each pixel (S110). The edge amount at a certain pixel is obtained based on the amount of change in the brightness value near the pixel.
For example, a Sobel filter can be used to calculate the edge amount.

Two thresholds TH1 and TH2, which are experimentally determined for the edge amount, are set in the dirt detection processing unit 12. Here, TH1 <TH2, and the first threshold TH1 is called a normal threshold and the second threshold TH2 is called a high threshold. For example, TH1 corresponds to the amount of change in the brightness value that makes the contour of the intruder appearing in the monitoring target area clear in the image acquired when the camera 2 is not contaminated. On the other hand, TH2 is usually around twice as much as TH1, for example 1.5-
It is preferable to set the value within the range of three times.

The stain detection processing unit 12 determines the number M of pixels having an edge amount exceeding TH1 among the pixels forming the image.
(Normally referred to as the threshold edge number) and the number N of pixels having an edge amount exceeding TH2 (referred to as the high threshold edge number) are counted and obtained (S115, S120).
By the way, the pixel counted as the high threshold edge number corresponds to a pixel having a large spatial luminance change, that is, a pixel having a strong contrast.

When M and N are obtained, a high threshold edge ratio A defined by the ratio N / M is calculated as a comparative evaluation value (S125).

When the number of threshold edges M is 0, a division by zero occurs in the calculation of the ratio N / M. Further, when M is very small, the error of A due to the statistical variation of M and N becomes large, which makes it inappropriate to use for the determination described later. Therefore, the calculation process S125 of the ratio N / M is performed.
Is performed only when M is larger than a predetermined threshold TH3, and M
Is less than TH3, the ratio N / M is not calculated (S1
30). Such a very small value of M generally means that the brightness of the entire image is poor. This is because (1) the area to be monitored originally has little change in brightness, (2) the lens surface of the camera is heavily soiled, and nothing is visible. He said that surveillance was hindered. Among these, the factors (2) and (3) are inconvenient because the intruder cannot be detected. Therefore, M is TH3
In the following cases, the high threshold edge ratio A is given a value such as 0, which is detected as an abnormality in the determination process described later (S135). The threshold TH3 can be empirically determined based on experiments and the like.

The high threshold edge ratio A represents the ratio of pixels having a high brightness change among the pixels having a brightness change, and takes a higher value for an image having a better contrast. For example, this is specifically understood by comparing FIG. 4 (b), which corresponds to the image without stains, with FIG. 5 (b), which corresponds to the image with stains. It That is, as shown in FIG. 4B, in a high-contrast image, a large amount of change in luminance value whose absolute value exceeds TH2 occurs in a relatively small number of limited pixels near the contour of the subject. On the other hand, as shown in FIG. 5B, in a low-contrast image that is blurred due to stains, the change in the brightness value corresponding to the contour of the subject is dispersed in a relatively large number of pixels. Therefore, when dirt adheres, the number N of pixels exceeding the high threshold value TH2 decreases. On the other hand, the number of pixels having an edge amount smaller than TH2 increases, which acts to increase M. Therefore, the high threshold edge ratio A when dirt is attached is usually lower than the high threshold edge ratio A when dirt is not attached.

By the way, when the contrast decreases due to the decrease in illuminance, the number N of pixels exceeding the high threshold value TH2 decreases, but basically the increase in the number of pixels causing a brightness value change corresponding to the contour does not occur. As a result, the increase rate of the number of threshold edges M is not likely to be as large as that in the case of adhesion of dirt. Therefore, the decrease in the high threshold edge ratio A in this case is smaller than that in the case of stain adhesion, and even if the same in terms of contrast reduction, in the case of stain adhesion and the illuminance decrease based on the high threshold edge ratio A. It is possible to distinguish between and.

The judgment reference value TH4 is set based on experiments or the like so as to enable discrimination between the case where the stain is attached and the case where the illuminance is decreased. If the high threshold edge ratio A is larger than TH4 (S140), it is determined that no dirt is attached and the dirt count timer is cleared to zero (S1).
45). On the other hand, when the high threshold edge ratio A is equal to or less than TH4, it means that the dirt adheres or the above-mentioned M is very small. In this case, the dirt count timer is counted up (S150). . By the way, the dirt count timer is cleared to zero when the apparatus is started up, but is not cleared to zero when the dirt detection logic that is repeated at regular intervals ends.

The dirt count timer can be configured to count according to a predetermined clock even at the interval of the dirt detection logic unless it is cleared to zero, or to be incremented by 1, for example, in each dirt detection logic. You can also In any case, the value of the dirt count timer is a predetermined threshold value T seconds (for example, 60 seconds).
If the value is equal to or more than (seconds) (S155),
Assuming that the decrease of the high threshold edge ratio A to the threshold TH4 or less is not temporary due to the influence of ambient light or the like, but is an abnormal state having persistence such as dirt adhering to the lens surface,
An alert is sent to the observer (S160). On the other hand, when the value of the dirt count timer is less than T seconds, the dirt detection logic this time does not determine that there is an abnormality (S16).
5) Then, the dirt detection logic is ended.

In the above configuration, the high threshold edge ratio A is defined as the ratio of the number N of pixels exceeding the high threshold TH2 to the number M of pixels exceeding the normal threshold TH1, but the edge amount at the pixels exceeding the normal threshold TH1 is defined. And the integrated value SUM2 of the edge amount at the pixels exceeding the high threshold value TH2 are respectively processed S11.
5, calculated in S120, using them, processing S125
In step S140, A = SUM2 / SUM1 may be calculated, and the determination in step S140 may be performed using this A as a comparative evaluation value.

[Embodiment 2] The intruder monitoring apparatus according to the second embodiment has the same block configuration as that of the apparatus according to the first embodiment shown in FIG. 1, and the processing content is also shown in FIG. 1
There are many points in common with the processing contents of the apparatus of the above embodiment.
Therefore, hereinafter, the same components and processing steps as those in the first embodiment are designated by the same reference numerals to simplify the description.

FIGS. 6 and 7 are process flow charts showing the process of the dirt detection processing unit 12 in the intruder monitoring apparatus of the second embodiment, and FIG. 6 shows the processing contents related to the setting of the reference value for the dirt adhesion determination. FIG. FIG. 7 is a flowchart showing the processing contents of the dirt detection logic.

First, referring to FIG. 6, the setting of the reference value for the dirt adhesion determination will be described. For example, the dirt detection processing unit 1
2 starts the stain determination process, and the process S described below is performed.
After performing the determination reference value parameter setting process of 205 to S235 to determine the determination reference value parameter B used to set the reference value for the stain adhesion determination, the stain detection logic S240 is started to be repeated.

The determination reference value parameter setting processing is basically the same as the determination processing S105 to S135 of the high threshold edge ratio A based on the image in the state where no dirt is attached. That is, the dirt detection processing unit 12 acquires image data from the camera 2 at an appropriate timing (S205),
Edge amount calculation processing S210 is performed on this image,
The normal threshold edge number M and the high threshold edge number N are calculated (S21
5, S220), a higher threshold edge ratio N / M is calculated, and this is used as the determination reference value parameter B (S22).
5). If M is less than or equal to the threshold TH3 (S23
0) and B = 0 are set (S235).

When the judgment reference value parameter B is determined,
After that, the dirt detection logic S240 is repeated at regular intervals. Next, the dirt detection logic will be described with reference to FIG. The stain detection logic of this apparatus is different from the stain detection logic of the first embodiment in that the process S12 is different.
5, the dirt adhesion determination processing S300 based on the high threshold edge ratio A obtained in S135. That is, the process S30
The determination reference value compared with the high threshold edge ratio A at 0 is different from the processing S140 in that it is determined by multiplying the determination reference value parameter B by a predetermined scale factor R.

Although the threshold value TH4 used as the determination reference value in the determination processing S140 of the first embodiment does not correspond to each monitoring target area, this determination processing S3
The product (R · B) used as the determination reference value in 00 is determined by the parameter B in accordance with each monitoring target area. Therefore, by adjusting R, the sensitivity of dirt detection is improved while suppressing erroneous detection. It becomes easy to do.

Here, the judgment reference value parameter setting process is performed as an initial process at the time of starting the dirt detection processing unit 12, but it may be carried out at an appropriate timing thereafter (a state where dirt is not attached). May be. Especially, when the background image obtained from the camera 2 is changed in the absence of a moving body such as an intruder due to the layout change of the monitoring target area or the like, the determination reference value parameter setting process is performed again and the parameter B is updated. Preferably.

[0036]

According to the image monitoring apparatus of the present invention, a monitoring obstacle such as dirt on the lens surface of a camera can be detected with high accuracy.

[Brief description of drawings]

FIG. 1 is a schematic block configuration diagram of an intruder monitoring device according to the present invention.

FIG. 2 is a processing flowchart showing processing of a stain detection processing unit according to the first embodiment.

FIG. 3 is a schematic diagram of an image of a monitored area.

FIG. 4 is a graph showing an example of a brightness value and a brightness value change amount in a state where dirt is not attached to a lens surface or the like of a camera.

FIG. 5 is a graph showing an example of a brightness value and a brightness value change amount in a state where dirt is attached to a lens surface or the like of a camera.

FIG. 6 is a flowchart showing the contents of processing relating to setting of a reference value for dirt adhesion determination.

FIG. 7 is a process flow diagram showing a process of a stain detection processing unit according to the second embodiment.

[Explanation of symbols]

2 camera, 4 image processing device, 10 intruder detection processing unit, 12 stain detection processing unit.

Claims (6)

[Claims] 1. An image pickup unit that generates an image corresponding to an optical image of a monitoring target region, and for each pixel of the image, a pixel value change intensity is calculated based on a difference in pixel value between a peripheral pixel of the pixel. A pixel value change intensity calculation means, a first reference value obtained by summing the pixel value change intensities of the pixels having the pixel value change intensity exceeding a first threshold value, and a second reference value larger than the first threshold value. A second reference value obtained by summing the pixel value change intensities of the respective pixels having the pixel value change intensities exceeding a threshold value is calculated, and the first reference value is calculated.
Comparing means for calculating a comparative evaluation value according to the relative size of the second reference value with respect to the reference value; and the monitoring target area and the monitoring target area based on the comparison evaluation value being smaller than a predetermined judgment reference value. An image monitoring apparatus comprising: an obstacle detection unit that detects a semi-transparent monitoring obstacle in an optical path between the image pickup unit and the image pickup unit. 2. An image pickup means for generating an image corresponding to an optical image of a monitoring target area, and for each pixel of the image, a pixel value change strength is calculated based on a difference in pixel value between a peripheral pixel of the pixel. Pixel value change strength calculation means, a first reference value that is the number of pixels having the pixel value change strength that exceeds a first threshold value, and the pixel value change that exceeds a second threshold value that is greater than the first threshold value. Comparison means for obtaining a second reference value, which is the number of pixels having intensity, and calculating a comparison evaluation value according to the relative size of the second reference value with respect to the first reference value; An obstacle detection unit that detects a semi-transparent monitoring obstacle in the optical path between the monitoring target region and the imaging unit based on a value being smaller than a predetermined determination reference value; Image monitoring device. 3. The image monitoring device according to claim 1, wherein the comparison evaluation value is a ratio of the second reference value to the first reference value. 4. The image monitoring apparatus according to claim 1, wherein the obstacle detecting means is attached to a transparent member or a reflecting mirror arranged on the optical path as the monitoring obstacle. An image monitoring device characterized by detecting stains. 5. The image monitoring device according to claim 1, wherein the first threshold is the image acquired in a state where the monitoring obstacle does not exist, and the first target is the monitoring target. A value according to the pixel value change intensity that occurs corresponding to the contour of the detection target object located in the region, wherein the second threshold value is a value corresponding to twice the first threshold value, An image monitoring device characterized by. 6. The image monitoring apparatus according to claim 4, wherein the obstacle detection unit determines that the stain is attached when the state where the comparative evaluation value is smaller than the determination reference value continues for a predetermined time. Characteristic image monitoring device.