JP3232502B2 - Fog monitoring system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a fog monitoring system and, more particularly, to a fog monitoring system suitable for detecting the occurrence of fog on a road or the like.

[0002]

2. Description of the Related Art Conventionally, when detecting the occurrence of fog, transmittance meters are installed at two points in a monitoring target area, and the occurrence of fog is detected based on the measured value of each transmittance meter, or a road installation object is detected. Is photographed with a television camera, road-installed objects are identified from images obtained by photographing, and the occurrence of fog is determined based on the degree of identification.

However, in the method using a transmittance meter to monitor the generation of fog, the transmittance meter must be arranged at the monitoring point, and special equipment is required and the installation place is restricted. Sometimes. In addition, in the method of photographing a road installation with a television camera and identifying the image of the road installation obtained by shooting, the burden on the observer is heavy, and fog occurs if the observer does not always look at the monitor screen It cannot be immediately determined whether or not.

Therefore, as described in JP-A-6-308256, there has been proposed a method for detecting whether or not fog has occurred from the luminance of an image obtained by a television camera.

[0005]

In the prior art, since the occurrence of fog is detected based on the luminance of an image obtained by a television camera, the distribution of luminance is shifted depending on the weather and the occurrence of fog is accurately detected. It may not be possible.

An object of the present invention is to provide a fog monitoring system capable of detecting generation of fog with high accuracy.

[0007]

In order to achieve the above object, the present invention provides a photographing means for photographing an object to be monitored placed in an environment where fog is generated to generate a video signal; A template image generating means for converting a video signal into image data to generate a monitoring target video as a template image, and a registered pattern image indicating a reference value when no fog is generated as a pattern image relating to the monitoring target video. Registration pattern storage means, a correlation value calculation means for calculating a correlation value relating to a difference between the template image generated by the template image generation means and the registration pattern stored in the registration pattern storage means, and a calculated value of the correlation value calculation means. Monitoring information output means for outputting monitoring information relating to generation of fog , wherein the monitoring information output means includes
The transmittance around the monitored object is calculated from the correlation value
The transmittance calculating means to be calculated and the transmittance calculating means
Fog around the monitored object when the transmittance is below the set value
Alarm output means for outputting an alarm indicating the occurrence of
The transmittance calculating means is a ramel-heart type neural network.
The correlation value is used as the input neuron value, and the transmission
This is a fog monitoring system in which the rate is used as an output neuron value .

[0008] The present invention also relates to a method of placing the apparatus in an environment where fog is generated.
Imaging means for imaging the monitored object and generating a video signal
And convert the video signal from the shooting means to image data and monitor
Generates the image to be viewed as a template image corresponding to the shade.
Template image generation means to generate
When fog is not generated as a shade pattern image
Registration pattern that stores a registration pattern image indicating the reference value of
Storage means and template image generation means
Stored in the template image and registered pattern storage means
Phase for calculating the correlation value related to the difference from the registered pattern image
Generation of fog from the calculated value of the correlation value calculation means and the correlation value calculation means
Monitoring information output means for outputting monitoring information regarding
In addition, the monitoring information output means is used for calculating the correlation value calculating means.
To calculate the transmittance around the monitoring target from the correlation value
Calculator and transmittance calculated by transmittance calculator are set
Warning that fog has occurred around the monitoring target when the value is below the value
Alarm output means for outputting the transmittance, wherein the transmittance calculation means
Consists of a ramel-heart type neural network
The correlation value is used as the input neuron value, and the transmittance is used as the output neuron.
The fog monitoring system that is
You.

[0009] The present invention is also applicable to an environment where fog is generated.
Imaging means for imaging the monitored object and generating a video signal
And convert the video signal from the shooting means to image data and monitor
A template that generates a video to be viewed as a template image
Rate image generation means and putter related to video to be monitored
Of the reference value when no fog is generated
Registered pattern storage means for storing a recorded pattern image;
Template image generated by template image generation means
And the registered pattern stored in the registered pattern storage means.
A correlation value calculating means for calculating a correlation value relating to the difference;
Monitoring information about fog generation is output from the value calculated by the value calculation means.
Monitoring information output means for outputting the monitoring information output means.
The column indicates the monitoring target based on the correlation value calculated by the correlation value calculation means.
A view distance calculating means for calculating a surrounding view distance, and a view distance
When the visibility distance calculated by the separation calculation means is equal to or less than the set value
An alarm that outputs an alarm that fog has occurred around the monitoring target
Output means, and the view distance calculating means includes
It is composed of a neural network with
Force neuron value, and view distance as output neuron value
Fog monitoring system.

[0010] The present invention is also applicable to an environment where fog is generated.
Imaging means for imaging the monitored object and generating a video signal
And convert the video signal from the shooting means to image data and monitor
Generates the image to be viewed as a template image corresponding to the shade.
Template image generation means to generate
When fog is not generated as a shade pattern image
Registration pattern that stores a registration pattern image indicating the reference value of
Storage means and template image generation means
Stored in the template image and registered pattern storage means
Phase for calculating the correlation value related to the difference from the registered pattern image
Generation of fog from the calculated value of the correlation value calculation means and the correlation value calculation means
Monitoring information output means for outputting monitoring information regarding
In addition, the monitoring information output means is used for calculating the correlation value calculating means.
To calculate the visibility distance around the monitoring target from the correlation value
A distance calculating unit, and a view distance calculated by the view distance calculating unit.
Fog occurred around the monitoring target when the separation was below the set value
Alarm output means for outputting an alarm to the effect that
The separation calculation means is a ramel-heart type neural network
And the correlation value as the input neuron value,
A fog monitoring system that uses
It is a thing.

Further , the present invention is applicable to an environment where fog is generated.
Imaging means for imaging the monitored object and generating a video signal
And convert the video signal from the shooting means to image data and monitor
A template that generates a video to be viewed as a template image
Rate image generation means and putter related to video to be monitored
Of the reference value when no fog is generated
Registered pattern storage means for storing a recorded pattern image;
Template image generated by template image generation means
And the registered pattern stored in the registered pattern storage means.
A correlation value calculating means for calculating a correlation value relating to the difference;
Monitoring information about fog generation is output from the value calculated by the value calculation means.
Surveillance information output means and images taken by the photographing means
Convert signal to image data and calculate average density of image
Average density calculation means and average density calculation means
Day and night discriminating means for discriminating day and night from the average density,
The monitoring information output means outputs the daytime including dusk by the daytime / nighttime determination means.
Information is output on the condition that it is determined to be between
It constitutes a fog monitoring system.

Further , the present invention provides a method for installing the apparatus in an environment where fog is generated.
Imaging means for imaging the monitored object and generating a video signal
And convert the video signal from the shooting means to image data and monitor
Generates the image to be viewed as a template image corresponding to the shade.
Template image generation means to generate
When fog is not generated as a shade pattern image
Registration pattern that stores a registration pattern image indicating the reference value of
Storage means and template image generation means
Stored in the template image and registered pattern storage means
Phase for calculating the correlation value related to the difference from the registered pattern image
Generation of fog from the calculated value of the correlation value calculation means and the correlation value calculation means
Monitoring information output means for outputting monitoring information regarding
Converts the video signal obtained by the means into image data
Average density calculation means for calculating the average density of
Day / Night format to determine day / night from average density by calculation of output means
The monitoring information output means includes a day / night discriminating means.
Condition that it was judged that it was daytime including dusk by step
Fog monitoring system that outputs information to
is there.

In configuring each of the fog monitoring systems,
Can add the following elements:

(1) Monitored object photographed by photographing means
Is set to multiple, and each monitoring is
The registered pattern image corresponding to the target is stored
The correlation value calculation means calculates a correlation value for each monitoring target.
The monitoring information output means uses the correlation value for each monitored
It outputs monitoring information.

(2) For generation of template image generation means
A registered pattern image from a template image
Registered pattern image generating means and registered pattern image generating means
Registration pattern according to the registration pattern image generated by the step
Updates the registered pattern image in the storage means at specified intervals.
With new means.

[0016]

According to the present invention, in order to reliably detect the occurrence of fog even at night, a photographing means for photographing a moving object moving in a photographing target area in an environment where fog occurs to generate a video signal. And image processing means for converting a video signal from the imaging means into image data to sequentially generate an image of a specific part of the moving body in accordance with the movement of the moving body, and photographing the image generated by the image processing means. Determining means for comparing images at different distances from the means to determine whether or not a specific part of the moving body in the distance can be identified, and determining information for outputting determination information on fog generation in accordance with the determination result of the determining means This constitutes a fog monitoring system including output means.

The following elements can be added to the nighttime fog monitoring system.

(1) The determination information output means includes: a transmittance calculation means for comparing the images generated by the image processing means at different distances from the photographing means to calculate a transmittance; and a transmittance calculation means. And an alarm generating means for generating an alarm when it is determined that the specific portion of the distant moving body cannot be identified when the transmittance obtained by the calculation is equal to or less than the set value.

(2) The determination information output means includes: a view distance calculation means for calculating a view distance by comparing images having different distances from the photographing means among the images generated by the image processing means; and a view distance calculation means. And a warning generating means for generating a warning when the visibility distance calculated by the calculation is less than or equal to the set value, and determines that a specific part of the distant moving body cannot be identified.

(3) The image processing means generates an image by illumination of the illumination member of the moving body as an image of a specific portion of the moving body.

(4) Average density calculating means for converting the video signal obtained by the image capturing means into image data to calculate the average density of the image, and day / night discriminating means for discriminating day and night from the average density by the average density calculating means. The determination information output means outputs information on condition that it is determined that it is nighttime by the day / night determination means.

According to the above-described means, in the daytime including dusk, a template image is generated based on the video signal obtained by photographing the monitoring target, and the template image is compared with the registered pattern image. Is obtained. That is, a correlation value indicating how much the two images are different from each other is obtained, and monitoring information on whether or not fog has occurred is output based on the correlation value. For example, when the correlation value is equal to or less than a certain set value, it is determined that fog has occurred and an alarm indicating that fog has occurred is output. Further, when determining the occurrence of fog, since the image is compared as a pattern, the occurrence of fog can be detected with high accuracy while the influence of weather is minimized. Also, when determining the occurrence of fog, the correlation value is input and the learning function of the Ramelheart-type neural network is used to determine the transmittance or the viewing distance according to the correlation value, and from these values it is determined whether or not fog has occurred. Can be determined.

In the nighttime, an image of a specific part of the moving body is sequentially generated in accordance with the movement of the moving body based on a video signal obtained by photographing the moving body moving in the photographing target area. Comparing the images at different distances from the photographing means from the obtained images to determine whether or not a specific part of the distant moving body can be identified, and output determination information regarding fog generation according to the determination result. Like that. In this case, it is possible to determine the occurrence of fog by setting a specific part of the moving body as the lighting member and determining how far the lighting member can be identified in a distant place. For example, by tracking images obtained from vehicle lighting components (headlights, tail lamps) and comparing the distance that can be tracked when fog is not generated with the distance that can be tracked when fog is generated, It can be determined whether or not the lighting member of the vehicle can be identified. The transmittance or the visibility distance is obtained from the traceable distance, and it is possible to determine from these values whether or not fog has occurred.

[0025]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings. In FIG. 1, a fog monitoring system includes a television camera 10, an image processing unit 12, and a fog detection unit 1.
4. The display unit 16 is provided.

The television camera 10 is arranged with a space above the road 18 as a photographing target area 20. The television camera 10 photographs a white line 22 placed in an environment where fog occurs and its surroundings as a monitoring target, and generates a video signal. The video signal is input to the image processing unit 12.

As shown in FIG. 2, the video processing unit 12
/ D converter 24, a plurality of image memories 26, a control unit (CP
U) 28, a template registration unit 30, a correlation operation unit 32,
Vehicle registration unit 34, vehicle tracking unit 38, average concentration extraction unit 38
It is provided with. The A / D converter 24 is the control unit 2
In response to a command from the controller 8, the video signal of the television camera 10 is converted into digital image data and output to the image memory 26. The image memory 26 stores the image data as grayscale image data and also stores the grayscale pattern as a template image. That is, the image memory 26 is configured as a template image generating unit together with the control unit 28.

The area near the white line 22 shown in FIG. 3 is divided into three areas R1, R2, and R3, and the images of the areas R1 to R3 are stored in the image memory 26 as an image of a light and shade pattern to be monitored. Is stored in Among the images of the light and shade patterns in each area, the image of the light and shade pattern when no fog is generated is registered in the template registration unit 30 as a registered pattern image indicating a reference value when no fog is generated. For example, the images of the regions R1 to R3 are registered in the template registration unit 30 as registered pattern images P1, P2, and P3, as shown in FIG. That is, the template registration unit 30 is configured as a registration pattern storage unit. Note that the images P1, P2, and P3 of the registered pattern have a distance of 10 meters from the television camera 10, respectively.
The shading patterns at 5 meters and 20 meters are shown, and each pattern is configured as a pattern of 256 gradations.

The correlation calculation section 32 stores the regions R1, R2,
Shading pattern image relating to R3 and template registration unit 3
It is configured as a correlation value calculation unit that calculates a correlation value (correlation degree or similarity) regarding the difference between the two patterns by comparing with the registered pattern images P1, P2, and P3 registered at 0. When calculating the correlation value using the template of the grayscale pattern image, a normalized correlation process according to the following equation is executed.

[0030]

(Equation 1)

[0031]

(Equation 2)

[0032]

(Equation 3)

[0033]

(Equation 4)

[0034]

(Equation 5)

[0035]

(Equation 6)

When executing the calculation according to the equation (1), the similarity r (u, v) of a certain point (u, v) is obtained by using the density template T for the image f to be processed.
The closer the similarity is to the template T, the closer to the value of 1. Then, in equation (1), f (u + i, v +
j) is the density value of the target image near the (u, v) point, T
(I, j) is the density value at point (i, j) of the template image, and p and q are the x and y sizes of the template image, respectively. In addition, since data relating only to templates that do not exist in the target image in the expression (1) can be calculated in advance, the values of the expressions (2) and (3) can be calculated before the correlation operation. Therefore, in the correlation operation between the template data and the image to be correlated, (4) which belongs to the equation (1)
Equations (5) and (6) are calculated in real time, and the similarity of equation (1) can be finally obtained by using the results of each calculation.

In deriving the equation (1), the following equation can be used.

For example, the equation indicating the correlation is expressed by the following equation (7).

[0039]

(Equation 7)

As for the correlation, if there is a difference in density gradation between f and g, a normal value cannot be extracted, and it is necessary to normalize the density. By normalizing the correlation with respect to the expression (7), the following expression (8) is obtained.

[0041]

(Equation 8)

By transforming equation (8), equation (8) becomes equation (9)
It is expressed by an equation.

[0043]

(Equation 9)

Table 1 below shows specific examples of r 'and the pattern.

[0045]

[Table 1]

In Table 1, when g is a registered pattern, the pattern f1 has a difference in numerical value between adjacent areas opposite to that of the registered pattern g, and the value of r 'is -1. On the other hand, in the patterns f2, f3, and f4, the value of r ′ is 1 because the difference in numerical values between adjacent areas has the same relationship as the registered pattern g.

Using the grayscale template T for the image f to be processed, let the similarity of the point (u, v) be r (u, v), and let the density value of the target image near the point (u, v) be f (u +
i, v + i), the density value of the point (u, v) of the template image is T (u, v), and the x and y sizes of the template image are p and q. Can be derived.

When the correlation value is calculated by the correlation calculation unit 32, as shown in FIG. 4B, the correlation of area R1 is 0.8, the correlation of area R2 is 0.6, and the area R3 is
, The degree of similarity in a region distant from the television camera 10 gradually decreases, such as 0.3. This indicates that fog has been generated from the front of the television camera 10. When fog is generated from behind the television camera 10, the similarity of each of the regions R1 to R3 indicates a low value.

On the other hand, even when the density of the image obtained from each of the regions R1 to R3 is reduced due to a change in the weather or at dusk, the pattern itself does not change when no fog is generated. ), Each region R
The correlation degrees of 1 to R3 are all 0.9 or more.

Next, in consideration of the fact that there is no road illumination or the white line 22 cannot be seen at night and the regions R1 to R3 cannot be set, the vehicle registration unit 34
The vehicle tracking unit 36 detects the occurrence of fog by tracking an image of a vehicle as a moving object. In this case, as shown in FIG. 5, the vehicle registration unit 34 determines that a specific part of the vehicle moving in the vehicle tracking area 40 in the image captured by the television camera 10 is a lighting member, for example, a tail lamp (headlight). The images of L1 and L2 are registered. As shown in FIGS. 5B and 5C, the vehicle tracking unit 36 determines whether the registered images of the tail lamps L1 and L2 can be tracked from the point of 10 meters to the points of 15 meters and 20 meters. The judgment is made. In FIG. 5B, when no fog is generated, the images of the tail lamps L1 and L2 are
This indicates that tracking can be performed up to a distance of 0 meters. When fog occurs, images of the tail lamps L1 and L2 cannot be identified at a distance of 20 meters from the television camera 10 as shown in FIG. Is shown. The relationship between the images of the tail lamps L1 and L2 and the distance can be determined based on the coordinates of the pixels.
That is, as a function of the image processing unit 12 at night, A
The / D converter 24, the image memory 26, the control unit 28, the vehicle registration unit 34, and the vehicle tracking unit 36 convert a video signal into image data and sequentially generate images of the tail lamps L1 and L2 of the vehicle in accordance with the movement of the vehicle. Image processing means for comparing the images of tail lamps L1 and L2 generated by the image processing means with different distances to determine whether or not images of the tail lamps L1 and L2 of the distant vehicle can be identified. Is composed.

The average density extractor 38 receives the image data from the image memory 26, calculates the average density of the image, and determines day or night based on whether the calculated value is darker or lighter than the set value. That is, the average density extracting unit 38 is configured as an average density calculating unit and a day / night determining unit together with the A / D converter 24, the image memory 26, and the control unit 28. Then, the determination result of the average density extraction unit 38 is input to the fog detection unit 14.

When registering a registration pattern image in the template registration unit 30, a registration pattern image is generated from the template image registered in the image memory 26,
The generated registration pattern image can be registered in the template registration unit 30 at a cycle of, for example, 5 minutes or 10 minutes, and the registration pattern image of the template registration unit 30 can be updated at an arbitrary cycle. In this case, the A / D converter 24 and the image memory 2
6. The control unit 28 constitutes a registered pattern image generating means and constitutes an updating means.

The fog detection section 14 constitutes monitoring information output means together with the display section 16 and also constitutes judgment information output means, and responds to the judgment result when the image processing section 12 judges that it is daytime. In order to detect fog in the daytime, an input switching unit 44, a fog detection circuit 48, and a learning mechanism 50 are provided, and the learning mechanism 50 includes an input signal generating mechanism 52, a parameter changing mechanism 54, and learning. A control mechanism 56, a teacher signal generating mechanism 58, and a butting unit 60 are provided. The fog detection unit 14 includes the image processing unit 1
A fog detection circuit 62 is provided to detect fog at night in response to the determination result from 2 to night.

The input switching section 44 is switched to the input signal generation mechanism 52 when learning is performed on the fog detection circuit 48 by the learning mechanism 50, and is switched to the image processing section 12 after the learning process is completed. It has become. The fog detection circuit 48 is formed of, for example, a ramel-heart type forward-looking neural network, and as shown in FIG.
Correlation values X1, X2, X3 of the regions R1, R2, R3 are given as potentials of input neuron values, and 11 output neuron values (Z1 to Z1) are passed through an intermediate neuron group (intermediate layer).
Z11) is output. The eleven output neuron values are respectively assigned to 10% units of transmittance 0 to 100%. In this neural network, a judgment result of a supervisor is given to an input image in advance, and all synapses (signal transmission lines connecting neurons: W11 to W15, V1
01 to V112,..., V501 to V512) are given optimal weights.

For example, when fog occurs, the values of the output neurons of the transmittance assigned to the situation are set to be close to 1.0, and the values of the other output neurons are set to be close to 0.0. I have. As the value of the output neuron, a visibility distance can be assigned in addition to the transmittance.
That is, the fog detection circuit 48 constitutes a transmittance calculating means using the correlation value as the input neuron value and the transmittance as the output neuron value, and the view distance using the correlation value as the input neuron value and the view distance as the output neuron value. It can be configured as calculation means.

The fog detection circuit 62 constitutes judgment information output means together with the display unit 16, and calculates the transmittance by comparing the output of the vehicle tracking unit 36, that is, the images of the tail lamps L1 and L2 with different distances. Along with constituting the transmittance calculating means, it is configured as a view distance calculating means for calculating a view distance by comparing images having different distances among the images of the tail lamps L1 and L2. For example, when determining the transmittance from the tracking result of the vehicle tracking unit 36, when the images of the tail lamps L1 and L2 at a distance of 10 meters from the television camera 10 can be tracked up to 50 meters, the transmittance is set to 100% and the trackable distance is set to 100%. When the distance is less than 50 meters and 45 meters or more, the transmittance can be obtained as 90%. As for the visibility distance, similarly to the transmittance, the visibility distance can be set in association with the trackable distance.

Next, the operation of the fog monitoring system shown in FIG. 1 will be described with reference to the flowcharts shown in FIGS.

First, as a process of the image processing section 12, a video signal relating to the shooting target area 20 shot by the television camera 10 is sequentially converted into image data, and
(Step 100). Imaging target area 20
When the image data in is input, the average density of the image is obtained by the average density extraction unit 38, and it is determined whether it is daytime or not (step 101). When daytime is determined,
The light and shade pattern images of the regions R1 to R3 in the input image are registered in the template registration unit 30 (step 102).
Thereafter, it is determined whether or not the cycle is for determining generation of fog (step 103). Since it is not necessary to perform the fog monitoring cycle in a short cycle such as several seconds, it is determined whether the determination cycle here is a cycle such as 5 minutes or 10 minutes.

Next, the image data relating to the gray-scale pattern image of each of the regions R1 to R3 registered in the template registration unit 30 and the image data stored as the registered pattern images P1, P2, P3 are fetched (step 104). Is calculated according to equation (1) (step 105). Next, a determination is made as to whether or not the correlation value of each of the regions R1 to R3 is 0.90 or more in the update cycle of the image data relating to the grayscale pattern image registered in the template registration unit 30 (step 106). That is, it is determined whether or not to update the image data of the gray-scale pattern image registered in the template registration unit 30 in a 5-minute or 10-minute cycle. If the update cycle is not reached, the process returns to step 103; The image data of the template registration unit 30 is updated according to the shaded pattern image. Further, when the correlation value in each area is 0.90 or more, it is determined that no fog has occurred, and that the situation
It returns to the process of 0.

On the other hand, when it is determined in step 101 that it is nighttime, the vehicle registration unit 34 detects and registers the tail lamps L1 and L2 of the vehicle moving in the vehicle tracking area 40 (step 107). Thereafter, by the processing of the vehicle tracking unit 36, the trackable distance of the tail lamps L1 and L2 is determined, and the process returns to the step 100 (step 108).

After the correlation value or the trackable distance of the vehicle is calculated by the processing of the image processing unit 12, the fog detection unit 14
Move on to processing.

When daytime is determined by the processing of the image processing section 12 (step 200), the correlation value calculated by the correlation calculation section 32 is set in the fog detection circuit 48. That is, the correlation values X1 to X3 of the regions R1 to R3 are given as potentials of the input neuron values, and 11 output neuron values are calculated via the intermediate neuron group (steps 201 and 20).
2). At this time, among the output neurons Z1 to Z11, Z4
Is 1.00 and the values of other output neurons are .0.
If it is 00, the transmittance is determined as 70% or the viewing distance is 50 meters (step 205).

The value of the output neuron Z7 is 1.00.
When the value of the other output neuron is 0.00, the transmission distance is 40% and the visibility distance is 20 meters. Then, it is determined whether fog has occurred based on the calculated transmittance or visibility distance (step 206). For example, when the transmittance is 70%, it is determined that fog has not occurred. When the transmittance is 60% or less, it is determined that fog has occurred. Is displayed (step 207).

On the other hand, when it is determined in step 200 that it is nighttime, a trackable distance by the processing of the vehicle tracking unit 36 is input (step 203), and processing for determining generation of fog is performed based on the input value (step 203). Step 204), calculating the transmittance or the visibility distance according to the trackable distance (step 205). When the trackable distance is 50 meters, for example, it is determined that fog has not occurred (step 206). When the trackable distance is 20 meters, it is determined that fog has occurred and an alarm is displayed on the screen of the display unit 16 (step 206). Step 207). Note that, instead of displaying the alarm on the display unit 16, the alarm may be output as audio.

As described above, in the present embodiment, in the daytime, a gray-scale pattern image of an input image is obtained, and this gray-scale pattern image is compared with a previously registered gray-scale pattern image to obtain a correlation value between the two. Since the presence or absence of fog is determined according to the correlation value, the occurrence of fog can be detected with high accuracy.

In the present embodiment, at night, the tail lamp of the vehicle moving in the vehicle tracking area 40 is registered, the trackable distance of the tail lamp is obtained, and the presence or absence of fog is determined according to the trackable distance. Therefore, the occurrence of fog can be detected accurately even at night.

By using an existing road monitoring television camera as the television camera 10, the cost for constructing the system can be reduced.

[0068]

As described above, according to the present invention,
In the daytime, a pattern image relating to the monitoring target obtained by imaging the imaging target area is generated, and this pattern image is compared with a registered pattern image indicating a reference value to obtain a correlation value between the two. Since it is determined whether or not fog has occurred, the presence or absence of fog can be detected with high accuracy regardless of the surrounding environment such as weather. Also,
At night, the image of a specific part of the moving object that moves within the shooting target area is tracked, and a traceable distance is calculated as to whether or not this image can be identified to a distant point. Since the presence or absence of occurrence is determined, the presence or absence of occurrence of fog can be detected with high accuracy even at night regardless of the surrounding environment such as weather.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram of a fog monitoring system according to an embodiment of the present invention.

FIG. 2 is a block diagram of an image processing unit.

FIG. 3 is a diagram for explaining a method of generating a template image.

FIG. 4 is a diagram for explaining a relationship between a registered pattern and a light and shade pattern based on an input image.

FIG. 5 is a diagram for explaining a vehicle tracking method.

FIG. 6 is a block diagram of a fog detection unit.

FIG. 7 is a block diagram of a fog detection circuit.

FIG. 8 is a configuration diagram of a fog detection circuit using a neural network.

FIG. 9 is a flowchart for explaining the operation of the image processing unit.

FIG. 10 is a flowchart for explaining the operation of the fog detection unit.

[Explanation of symbols]

 Reference Signs List 10 TV camera 12 Image processing unit 14 Fog detection unit 16 Display unit 20 Imaging target area 22 White line 24 A / D converter 26 Image memory 28 Control unit 30 Template registration unit 32 Correlation calculation unit 34 Vehicle registration unit 36 Vehicle tracking unit 38 Average Concentration extraction unit

Continued on the front page (72) Inventor Tadaaki Kitamura 7-1-1, Omika-cho, Hitachi City, Ibaraki Prefecture Inside Hitachi Research Laboratory, Hitachi, Ltd. (72) Inventor Issei Well 5-2-1, Omika-cho, Hitachi City, Ibaraki Prefecture (56) References JP-A-6-308256 (JP, A) JP-A-2-95242 (JP, A) JP-A-7-140003 (JP, A) JP-A-7- 26946 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G01W 1/00 E01F 9/00 G01N 21/17 G06T 7/00 G08B 21/00

Claims (13)

(57) [Claims] 1. A photographing means for photographing an object to be monitored placed in an environment where fog is generated to generate a video signal, and converting a video signal from the photographing means into image data to convert a video to be monitored into a template image. Template image generating means, a registered pattern storing means for storing a registered pattern image indicating a reference value when no fog is generated as a pattern image relating to a video to be monitored, and a template image generated by the template image generating means Correlation value calculation means for calculating a correlation value relating to a difference between the correlation pattern and the registered pattern stored in the registration pattern storage means, and monitoring information output means for outputting monitoring information on fog generation from the calculated value of the correlation value calculation means. The monitoring information output means calculates a correlation value
The transmittance around the monitored object is calculated from the correlation value
The transmittance calculating means to be calculated and the transmittance calculating means
Fog around the monitored object when the transmittance is below the set value
Alarm output means for outputting an alarm indicating the occurrence of
The transmittance calculating means is a ramel-heart type neural network.
The correlation value is used as the input neuron value, and the transmission
A fog monitoring system that uses the rate as the output neuron value . 2. A photographing means for photographing an object to be monitored placed in an environment where fog is generated to generate a video signal, and converting a video signal from the photographing means into image data to make the image of the monitored object dark and light. Template image generating means for generating a corresponding template image, registered pattern storage means for storing a registered pattern image indicating a reference value when fog is not generated as a light and shade pattern image relating to a video to be monitored,
Correlation value calculation means for calculating a correlation value relating to a difference between the template image generated by the template image generation means and the registered pattern image stored in the registered pattern storage means, and monitoring information on fog generation from the calculated value of the correlation value calculation means Monitoring information output means for outputting the monitoring information,
The report output means uses the correlation value calculated by the correlation value calculation means.
A transmittance calculating means for calculating the transmittance around the monitoring target;
When the transmittance calculated by the excess rate calculation means is below the set value
An alarm that outputs an alarm that fog has occurred around the monitoring target
Output means, and the transmittance calculating means includes
And a correlation value input.
Neuron value, and transmittance as output neuron value.
Fog monitoring system that. 3. An object to be monitored placed in an environment where fog is generated.
Shooting means for shooting an image to generate a video signal, and whether the shooting means
These video signals are converted into image data and the video to be monitored is
Template image generation to generate as template image
Means and fog as a pattern image related to the video to be monitored
Registered pattern image showing the reference value when no error occurs
Registration pattern storage means for storing a template image
Template image and registration pattern generated by generation means
Phase related to the difference from the registered pattern stored in the storage means
Correlation value calculating means for calculating the correlation value, and calculation of the correlation value calculating means
Monitoring information that outputs monitoring information about fog generation from the output value
Output means, wherein the monitoring information output means is configured to calculate a correlation value.
The visibility distance around the monitoring target from the correlation value calculated by the output means
Of the view distance calculating means for calculating
Around the monitoring target when the visibility distance is shorter than the set value
Alarm output means for outputting an alarm indicating that fog has occurred.
The visibility distance calculating means includes a Ramelheart-type neural
Input neuron values
And a fog monitoring system that uses the viewing distance as the output neuron value . 4. An object to be monitored placed in an environment where fog is generated.
Shooting means for shooting an image to generate a video signal, and whether the shooting means
These video signals are converted into image data and the video to be monitored is
A template generated as a template image according to shading
Image generation means and shading pattern relating to the video to be monitored
Indicates the reference value when no fog is generated
Registration pattern storage means for storing a registration pattern image;
Template image generated by template image generation means
Image and the registered pattern image stored in the registered pattern storage means.
Correlation value calculation means for calculating a correlation value relating to a difference from an image
And monitoring of fog generation from the value calculated by the correlation value calculation means.
Monitoring information output means for outputting information;
The report output means uses the correlation value calculated by the correlation value calculation means.
View distance calculating means for calculating the view distance around the monitoring target
And the view distance calculated by the view distance calculating means is equal to or less than the set value.
When below, a warning that fog has occurred around the monitoring target is issued.
Warning distance output means, and the view distance calculation means.
Consists of a ramel-heart type neural network
The correlation value is used as the input neuron value, and the viewing distance is used as the output neuron value.
A fog monitoring system that serves as a Duron value . 5. An object to be monitored placed in an environment where fog is generated.
Shooting means for shooting an image to generate a video signal, and whether the shooting means
These video signals are converted into image data and the video to be monitored is
Template image generation to generate as template image
Means and fog as a pattern image related to the video to be monitored
Registered pattern image showing the reference value when no error occurs
Registration pattern storage means for storing a template image
Template image and registration pattern generated by generation means
Phase related to the difference from the registered pattern stored in the storage means
Correlation value calculating means for calculating the correlation value, and calculation of the correlation value calculating means
Monitoring information that outputs monitoring information about fog generation from the output value
The output means and the video signal obtained by the photographing means
Average density calculator that calculates the average density of the image
And the average density calculated from the average density by day and night
Day and night discriminating means for discriminating between the monitoring information output means.
The stage is judged to be daytime including dusk by day / night discriminating means.
A fog monitoring system that outputs information on the condition that it has been performed . 6. An object to be monitored placed in an environment where fog is generated
Shooting means for shooting an image to generate a video signal, and whether the shooting means
These video signals are converted into image data and the video to be monitored is
A template generated as a template image according to shading
Image generation means and shading pattern relating to the video to be monitored
Indicates the reference value when no fog is generated
Registration pattern storage means for storing a registration pattern image;
Template image generated by template image generation means
Image and the registered pattern image stored in the registered pattern storage means.
Correlation value calculation means for calculating a correlation value relating to a difference from an image
And monitoring of fog generation from the value calculated by the correlation value calculation means.
Monitoring information output means for outputting information, and
Video signal into image data and average image density
Average density calculating means to calculate, and calculation of average density calculating means
Day and night discrimination means for discriminating day and night from the average density
The monitoring information output means is provided with
Information is output on condition that it is determined that it is daytime including
Fog monitoring system made by. 7. A plurality of monitoring targets to be photographed by the photographing means are set, a registration pattern image corresponding to each monitoring target is stored in the registration pattern storage means, and a correlation value calculating means relates to each monitoring target. Calculate the correlation value,
The fog monitoring system according to any one of claims 1 to 6, wherein the monitoring information output means outputs monitoring information from a correlation value for each monitoring target. 8. A registered pattern image generating means for generating a registered pattern image from a template image generated by the template image generating means, and specifying a registered pattern image in the registered pattern storing means in accordance with the registered pattern image generated by the registered pattern image generating means. The fog monitoring system according to any one of claims 1 to 7, further comprising updating means for updating the fog in a cycle of (i) . 9. A photographing means for photographing a moving object moving in a photographing target area in an environment in which fog is generated to generate a video signal, and converting a video signal from the photographing means into image data to generate a video signal. An image processing means for sequentially generating an image of a specific part in accordance with the movement of the moving object; and an image generated by the image processing means, comparing images having different distances from the photographing means to specify a moving object in a distant place. A fog monitoring system comprising: a determination unit that determines whether or not the part can be identified; and a determination information output unit that outputs determination information regarding fog generation according to a determination result of the determination unit. 10. A transmissivity calculating means for comparing transmissive images at different distances from the photographing means among the images generated by the image processing means to calculate transmissivity;
Claim and a warning generating means for transmittance by calculation of the transmittance calculating means for generating an alarm as unable to identify the specific site of the mobile in the distance when the set value or less 9
Fog monitoring system as described. 11. A visual distance calculating means for comparing visual distances by comparing images having different distances from a photographing means among images generated by the image processing means, and a visual distance calculating means. The fog monitoring system according to claim 9 , further comprising: an alarm generating unit configured to generate an alarm when the calculated view distance is equal to or less than a set value and that a specific part of the distant moving body cannot be identified. 12. The fog monitoring system according to claim 9 , wherein the image processing means generates an image obtained by illuminating a lighting member of the moving body as an image of a specific portion of the moving body. 13. An average density calculating means for converting an image signal obtained by shooting of an image capturing means into image data and calculating an average density of an image, and a day / night determining means for determining day and night from the average density by calculating the average density calculating means. the provided, determination information output means, according to claim 9, on condition that it is determined to be nighttime by day and night determining means comprising output information, 10,11 also
Is a fog monitoring system according to item 12 .