JPH0737057A - Monitoring device - Google Patents

Abstract

PURPOSE:To improve the detecting accuracy of obstacles in monitoring an object passing through a monitoring field. CONSTITUTION:The contour component of a present contour image generated by an image C generating part 300 is eliminated in a passing area image generated by an image A generating part 100 by means of the mask image of a reference contour generated by an image B generating part 200. Thus, an image D is generated as a contour image of an obstacle generated by an image D generating part 400. The past contour component is eliminated out of the present contour component in an object passing area, at the same time, the noise influences are eliminated in the areas excluding the object passing area. Since the presence of an obstacle is decided by the presence or absence of an obstacle within the area of the object passing through a monitoring field, the obstacle detecting accuracy can be improved.

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 monitoring the safety of an object passing on orbit, for example.

[0002]

2. Description of the Related Art As conventional monitoring devices, there are a photoelectric type using infrared rays and laser light and a loop coil type.

In the case of the photoelectric type, as shown in FIG. 1, for example, a pair of light transmitters 2A and 3A and light receivers 2B and 3B are arranged on both sides of the track 1 along the track 1.
The light transmitter 4A is arranged on one side of the track 1, and the light receiver 4B is arranged on the other side of the track 1.

As a result, on both sides of the orbit 1, the light emitted from the light transmitters 2A and 3A is received by the light receivers 2B and 3B. If there is no obstruction, both sides of the orbit 1 are safe. To be judged. Further, the light emitted from the light transmitter 4A is received by the light receiver 4B, and if there is nothing that blocks this light, it is determined that the trajectory 1 is safe.

On the other hand, in the case of the loop coil type, for example, as shown in FIG.
A and 5B are buried, and these loop coils 5A and
These inductances change when the metal body approaches 5B, and thus the obstacles are detected by detecting them.

[0006]

However, in the above-mentioned conventional monitoring device, in the case of the photoelectric type, the light transmitters 2A, 3A, 4A and the light receivers 2B, 3B, 4B malfunction due to snowfall or the like, An accurate obstacle could not be detected.

Therefore, in the section where snow is removed by the Russell car, the light transmitter and the light receiver are buried and cannot be used, and the lens surface becomes dirty, so cleaning work is required several times a year. There is. In addition, the infrared beam has large weaknesses against snowfall and fog, and the laser type has weak points such as weakness in optical axis deviation between the transmitter and the receiver.

On the other hand, in the case of the loop coil type, the change amount of the signal due to the metal of the automobile is small and it is easily affected by the temperature and humidity. Moreover, it is necessary to pave the loop coil on the concrete after burying it, which raises a problem that the installation cost becomes high.

The present invention has been made in consideration of such a situation, and generates an area of an object passing through a surveillance visual field,
An object of the present invention is to provide a monitoring device with high detection accuracy by determining whether or not an obstacle is present in the area.

[0010]

In order to achieve the above object, the present invention provides a monitoring device for monitoring an object passing on a fixed orbit, based on the moving locus of the object passing through the monitoring field of view. Passing area generating means for generating a passing area of the object, and a contour obtained by removing a past contour component considered to have a small change from the contour component of the current image in the passing area generated by the passing area generating means A contour component extracting means for extracting a component and a variation of the contour component extracted by the contour component extracting means are detected based on changes in the sizes of the top several contour components existing in the image, and an obstacle is detected. And an obstacle determining means for determining whether or not

[0011]

The monitoring apparatus of the present invention generates a region of an object which passes through the surveillance field of view, removes the past contour component from the current contour component within the region, and removes the top several components existing in the image. Since it is determined whether or not the obstacle is an obstacle based on the change in the size of the contour component, the influence of the noise component outside the passage area is removed, and thus the obstacle detection accuracy is improved.

Further, since the obstacles are judged based on the changes in the past contour components based on the changes in the sizes of the upper several pieces, it is possible to easily judge whether or not the trajectory through which the object passes is secured. Therefore, it is possible to easily and reliably monitor the object passing through the monitoring field of view.

Furthermore, since the passage area of the object is automatically generated, the restriction on the installation location of the image capturing device (camera) is greatly alleviated. Furthermore, since the passage area of the object is generated for each passage of the object, even if the installation location of the image capturing device (camera) is displaced or the environment changes, the degree of tracking of these changes is high. Is also increased.

[0014]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 3 shows a system configuration according to an embodiment in which the monitoring device of the present invention is applied to a monitoring device for monitoring a railroad crossing where a train passes, and an image A which is a passing area image is displayed. Image A generation unit 10 for generating
0, an image B generation unit 200 for generating a reference contour mask image, an image C generation unit 300 for generating an image C that is the current contour image, and an image D that is a contour image of an obstacle.
Image D generation unit 400 and initial setting unit 50 for generating
0 and an alarm generation unit 600 are provided.

Next, the operation of the monitoring device having such a configuration will be described with reference to FIGS. First, FIG. 6 (a)
Generation of the image A shown in will be described. That is, when the image signal from the camera 10 is taken into the image memory 101 of the image A generation unit 100, the absolute value calculation unit 103 immediately before the luminance component of the image signal of the image memory 101 and the image signal taken into the image memory 101. The absolute value of the difference from the luminance component of the image signal of the image memory 102 that has captured the image signal of is obtained. Here, the input image from the camera 10 is stored as a digital image having a horizontal direction 512, a vertical direction 432, and a density of 8 bits (0 to 255).
The binarized image has the same number of pixels and either zero density or a fixed value other than that.

When the calculation for obtaining such an absolute value is completed, the current image signal stored in the image memory 101 is stored in the image memory 102 to be ready for the next calculation.

The image calculated by the absolute value calculation unit 103 is added to the previous value by the integration unit 110. Further, when the addition is performed by the integrating unit 110, the number is added by the counter 109.

Further, the absolute values calculated by the absolute value calculation unit 103 are averaged by the average value generation unit 104 to obtain the average value of the entire image. This value is added by the integration unit 106, and the number of times of addition is counted by the counter 105.

The set of average values of the entire image thus obtained is averaged by the average value generation unit 107 to obtain a moving average of the density average values at the time of averaging. The comparison unit 108 uses the average value generation unit 104.
And the output of the average value generation unit 107 are compared.

It is considered that the density average value hardly changes when the object is not on the orbit, but when the object is passing on the orbit, the image input from the camera 10 is obtained. Is expected to have a large variation in the area through which the object passes, and a small variation in the other areas.

Therefore, the average value of the absolute values of the differences between the images is within a certain small level when the object is not passing, and is considerably higher than that when the object is passing, until the passing is completed. It is thought that this state will continue.

When the comparing section 108 determines that such a large level continues twice or more, it outputs a passing object detection signal. This passing object detection signal is used by the integration unit 11
0 and the restart signal of the counter 109 are used, and the integration of the absolute value of the difference including the passing object is performed again. As a result, only the image of the absolute value of the difference when the object is passing is integrated.

When the passage of the object is completed, the comparison unit 1
The output of the passing object detection signal from 08 disappears. At this time, it is expected that the image of the integration unit 110 is almost divided into those in which the value of the passing area is large and the value of the other area is small. The image of this integration unit 110 is converted into a binarization unit 1
11 is 2 with the number of additions of the counter 109 as a parameter
Image A is generated by performing the binarization.

When the passing object is short, the passing area may not be properly generated. Therefore, when the passing area image generating section 112 generates this image each time the object passes,
The update unit 113 selects an image having a larger passing area than the passing area images so far.

Next, the case of generating the images B and C will be described. That is, first, in the image B generation unit 200 for generating the mask image of the reference contour, the image C is added to the current image signal stored in the image memory 101.
The Sobel operator in the generation unit 300 acts on the Sobel operator to enhance the contour component, and the histogram generation unit 304 generates a histogram of the density values of the image. After the histogram is generated, the threshold value calculation unit 305 calculates a value at which the upper 25% density exists as a threshold value.

Here, the Sobel operation in the Sobel operation unit 301 is used to emphasize the contour component of the image, and the following operation is performed for each pixel of the image.

That is, as shown in FIG. 5, when pixels A to I are arranged, the value of the position of the central pixel E is represented by | A + 2 × D + G−C−2 × F−I | + | A + 2 × B +
It is calculated by C−G−2 × H−I |. The first half of the equation corresponds to the differential in the horizontal direction, and the second half corresponds to the differential in the vertical direction. Since it also includes the averaging function, it has a characteristic of being resistant to noise.

Using this value, the contour emphasized image is binarized by the binarizing unit 3
By binarizing in 02, only the contour component of the image signal is extracted. By contracting the binarized data by the data contraction unit 303 with respect to the output of the binarization unit 302, a minute contour component is removed, thereby reducing adverse effects caused by rainfall, snow removal, small dust, and the like. .

Here, further explaining the data contraction in the data contraction unit 303, when the pixel at the position E in FIG. 5 is at the logic level 1, the pixel at the logic level 0 in any of B, D, F, and H. , The logic level of the pixel at the position E is set to 0. As a result, the boundary pixels of the contour component are offset.

Unlike the image A described above, this image is integrated by the integration unit 202 only when no object is passing through. That is, when the passing signal (PS) is supplied to the integrating unit 202 and the passing object detection signal from the comparing unit 108 is not transmitted, the binarized image is added by the integrating unit 202. At the same time, the number of additions is counted by the counter 205. Where counter 2
1/2 of the value of 05 is set as the threshold value, and 2 is set based on this threshold value.
The binarizing unit 203 binarizes the output image of the integrating unit 202.
This binarized one becomes the reference contour when there is no passing object. The inversion unit 206 inverts the output of the binarization unit 203 to generate a mask image of the reference wheel portion.

Here, the value of the counter 205 is, for example, 5
In the following cases, since the accuracy of the reference wheel portion is still low, the updating section 204 supplied with the above-mentioned pass signal (PS) uses the previous reference contour to maintain the accuracy. On the other hand, when the value of the counter 205 exceeds 5, the value of the binarizing unit 203 is copied to the updating unit 204 in order to maintain the accuracy next time.

Next, the case of generating the image D will be described. That is, first, in the image D generation unit 400 for generating the image D which is the contour image of the obstacle, the obstacle image generation unit 403 generates the obstacle image based on the output of the logical product unit 207.

The data expansion unit 404 performs data expansion on the obstacle image generated by the obstacle image generation unit 403 to supplement the cut of the image contour.
The determination unit 405 compares the image size of the obstacle output from the data expansion unit 404 with the system-specific value of 4,000 to determine the presence or absence of the obstacle.
When it is determined that there is an obstacle, the alarm generation unit 600 shown in FIG. 3 issues an alarm. At this time, the integration suppression signal is sent from the determination unit 405 so that the binarized image including the obstacle is not added to the integration unit 202, and the integration suppression signal is also sent to the counter 205. As a result, the counting operation of the counter 205 is stopped.

Next, the operation of the initial setting section 500 will be described. That is, the initial setting unit 500 determines whether or not the monitoring environment at the time of initial setting of the monitoring operation is prepared, and also checks whether or not the following two conditions are satisfied at the same time.

As one of the conditions, the average value of the current entire image is obtained by the average value generation unit 506, and this average value is compared with the average value of the above absolute value image. It is compared whether or not the value of (average absolute value of difference) / (average of current image) is smaller than the system-specific value of 0.2. here,
If there is no moving object and the monitoring environment is stable, the average of the absolute values of the differences will be small, but the values themselves will fluctuate depending on the current average value of the concentrations, so the values themselves will be standardized.

As another condition, since the system is monitoring the orbit in order to detect an obstacle at the time of initial setting,
If there are no obstacles, the contour of the trajectory or the supporting portion of this trajectory occupies a large proportion of the overall contour, and several contours of approximately the same accuracy are detected at the same time. On the other hand, when an obstacle is present on the orbit from the beginning, it is expected that the contour components are intricately entangled with each other and jump out to cause some large contours.

In order to distinguish between these, the labeling unit 501 performs labeling for each connected component of the current image that has been contracted (that has contours connected vertically or horizontally), and the sorting unit 502. The size of the area is obtained for each label number, and the size is sorted. After being sorted, the sample extracting unit 503 extracts the top 10 samples for each monitoring operation.

Each of them is stored as a value of 10 in the accumulator 505, and the number of times of storage is stored in the counter 504. Here, the integrating unit 505 calculates (10 × 1
0) has a limited storage area, and the storage area is cyclically used by using the value of the counter 504 as a ring counter. The value of the counter 504 is 4
If the following is true, the comparing unit 507 compares whether or not the higher ranks 1 and 2 of the set of 10 values in the accumulating unit 505 are significantly larger than the other eight. If the size is extremely large, it is determined that an obstacle exists, and the value is always output.

If it is extremely large, the counter 50
By using the value of 4, the average value generation unit 506 obtains the average value in units of 10 so far. The magnitudes of these 10 values and the current 10 values are compared with each other, and if they are smaller than the system-specific value, it is determined that there is no obstacle.

When the logical product of the above two judgment results (whether they are both satisfied or not) is obtained by the logical product 401, this becomes a start signal for newly generating the passing image area. The timer 402 measures the time from the passage signal to the activation signal. If the activation signal does not occur within a certain range, it means that there is a moving object or an obstacle on the orbit. A judgment is made and an alarm is issued.

As described above, according to this embodiment, the passing area on the railroad crossing which is the passing area of the object is generated, the past contour component is removed from the present contour component in the area, and the outside of the passing area is generated. Since the influence of the noise of is removed, the accuracy of obstacle detection can be improved. Further, since the change in the contour component in the past is determined by the change in the size of the top several contours, it is equivalent to simply determining whether or not the trajectory through which the object passes is secured. It contributes to ensuring the safety of traffic.

Further, since the passage area is automatically generated, the flexibility of the installation location of the camera 10 is enhanced. Further, since the passing area and the mask image are generated each time the vehicle passes, the degree of tracking with respect to the positional deviation of the camera 10, changes in the amount of sunshine, weather conditions, etc. can be enhanced.

In this embodiment, the case where the safety of an object passing on the orbit is monitored by using a camera on the ground has been described, but the present invention is not limited to this example, and the camera 1 is placed on the passing object.
0 may be installed. In this case, it is possible to detect the smoothness of the trajectory itself and the physical stability of the object traveling on the smooth trajectory by the same area generation method. Becomes

Further, when the object is made stationary, it is possible to detect that another object is approaching or moving away by the same method of area generation. Moreover, when the number of regions is counted, the number of other objects moving around can be counted.

Further, in the present embodiment, the case where the present invention is applied to the monitoring device for monitoring the level crossing which is the passage area of the train has been described, but the present invention is not limited to this example, and for example, object monitoring on a road. Etc. may be applied.

In this embodiment, the case where the area is set by using the change in density has been described, but the present invention is not limited to this example, and the same area can be set by using the change in saturation.

[0047]

As described above, according to the surveillance device of the present invention, a region of an object passing through the surveillance field of view is generated, and the past contour component is removed from the present contour component in the region, Since it is determined whether or not the obstacle is an obstacle based on the changes in the sizes of the top several contour components existing in the image, the influence of the noise component outside the pass area is removed. The accuracy of obstacle detection is improved.

Further, since the obstacles are judged based on the changes in the past contour components based on the changes in the sizes of the upper several pieces, it is simply judged whether or not the trajectory through which the object passes is secured. Therefore, it is possible to easily and reliably monitor the object passing through the monitoring field of view.

Furthermore, since the passage area of the object is automatically generated, the restriction on the installation location of the image capturing device (camera) is greatly relaxed. Furthermore, since the passage area of the object is generated for each passage of the object, even if the installation location of the image capturing device (camera) is displaced or the environment changes, the degree of tracking of these changes is high. Is also increased. Since the area of the object passing through the surveillance field of view is generated and it is determined whether or not the obstacle exists in the area, the detection accuracy can be improved.

[Brief description of drawings]

FIG. 1 is a diagram showing a conventional photoelectric monitoring device.

FIG. 2 is a diagram illustrating a conventional loop coil type monitoring device.

FIG. 3 is a diagram showing a system configuration according to an embodiment of a monitoring device of the present invention.

FIG. 4 is a diagram showing details of each component in the monitoring device of FIG.

5 is a diagram for explaining the operation of the Sobel operator in the Sobel operation unit of the image C generation unit in FIG.

6 is a diagram showing an image generated by each component in the monitoring device of FIG.

[Explanation of symbols]

 10 camera 100 image A generation unit 200 image B generation unit 300 image C generation unit 400 image D generation unit 500 initial setting unit 600 alarm generator

Continuation of front page (51) Int.Cl. ⁶ Identification number Office reference number FI technical display location G08B 25/00 510 M 7323-5G

Claims (1)

[Claims] 1. A monitoring device for monitoring an object passing on a fixed orbit, and passing area generating means for generating a passing area of the object based on a movement locus of the object passing through the monitoring field of view; Contour component extracting means for extracting a contour component obtained by removing a past contour component which is considered to have a small change from the contour component of the current image in the passage area generated by the area generating means, and the contour component extracting means. An obstacle determining means for detecting a change in the extracted contour component based on changes in the sizes of the top several contour components existing in the image and determining whether or not the obstacle is an obstacle. A monitoring device characterized by being provided.