JPH09282455A - Image monitoring method, image monitoring device, and road monitoring device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the image monitoring device which easily processes a television image of a monitor area and does not require the speediness of hardware and is low in processing cost. SOLUTION: A road image wherein there is no moving body in the monitor area is registered as a template image in a memory 4 in advance. An input image of the same area with the template image is converted from analog to digital and inputted to a memory 3 in order from a camera 1 and the similarity between the template image and input image is calculated by a similarity processing means 7 and stored in a similarity storage part 8. A similarity analyzing process part 9 decides that there is a vehicle when the similarity decreases below a threshold value continuously by more than a specific number of times. Further, the vehicle speed, vehicle length, etc., are easily estimated on the basis of the continuous decrease time of the similarity or similarity variation pattern. Thus, a multifunctional vehicle sensor can be actualized by image processing by only light-shade template matching.

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 intruding object in a surveillance area, and more particularly to a road monitoring apparatus for obtaining the number of vehicles passing through the surveillance area, speed, vehicle length and the like based on vehicle recognition. Further, the present invention relates to an elevator waiting passenger monitoring device, a train platform monitoring device, and a railroad crossing monitoring device by recognizing an object in the monitoring area.

[0002]

2. Description of the Related Art As a vehicle sensor, an existing type sensor represented by an ultrasonic sensor is widely used. This sensor is used to measure the number of vehicles when the vehicle passes under the sensor, and a signal of vehicle presence is obtained.

On the other hand, recently, a system has been developed which measures the number of vehicles by photographing the road with a television camera and image-processing this image. For example, JP-A-7-244
Japanese Laid-Open Patent Publication No. 795 describes a vehicle detection device that periodically marks a road surface, binarizes the density value at the position of the mark, and determines the presence or absence of a vehicle based on whether or not it is the same as the mark.

Since such an image type sensor can be processed planarly, it has an advantage that one camera can measure a plurality of lanes as compared with an ultrasonic sensor. For this reason, it is used to measure the speed of a vehicle to determine normal running / congestion, etc., and to monitor anomalies in traffic flow in tunnels and the like. For example, Japanese Unexamined Patent Publication No. 5-298891 describes an object recognition device that detects a vehicle, tracks the image of the vehicle by light and shade pattern matching, and measures the speed and the number of vehicles with high accuracy.

[0005]

However, in the above-mentioned vehicle detection device, it is necessary to mark the road surface, which is a great difficulty in application. In addition, since it is determined by the binarization process whether the car shields the mark, it is difficult to cope with the change in the brightness of the environment. On the other hand, the above-mentioned object recognition device has a drawback that the high speed is required for the hardware and the processing cost is very large because the vehicle tracking process is required. Moreover, since a two-dimensional background image of the entire measurement area is required, it is difficult to update the background image, and it is difficult to smoothly respond to environmental changes.

In view of the state of the art, the object of the present invention is to
An object of the present invention is to provide an image monitoring method and apparatus with a low processing cost by simple image processing that does not require a road mark or tracking processing. Another object of the present invention is to provide an image monitoring method and apparatus which can update the background image online and have high followability to environmental changes.

Another object of the present invention is to provide at low cost a road monitoring device capable of high-speed detection of a passing vehicle. Another object of the present invention is to provide a multifunctional vehicle measurement sensor that measures the number of passing vehicles, the vehicle speed, and the vehicle length or vehicle type based on the detection of passing vehicles. This sensor is suitable for determining whether or not a traffic flow is abnormal.

The other objects of the present invention will become apparent through the following description.

[0009]

An object of the present invention is to provide an image monitoring method in which an image of a surveillance area is input, and an abnormal state of the surveillance area is detected by performing a density matching process with an image in a normal state. A predetermined area is cut out from the image of the state and stored as a template image, and the similarity between the input image and the template image corresponding to the predetermined area is obtained each time while repeatedly capturing the monitoring area,
This is achieved by determining that the monitoring area is in an abnormal state when the decrease in the degree of similarity continues for a predetermined number of times or more.

The above-described other object is further achieved by updating the template image with the input image when the similarity exceeding the predetermined threshold value continues for a second predetermined number of times or more.

The predetermined area may be a linear area or a linear rectangular area that is sufficiently smaller than the monitoring area, and the similarity may be calculated by a correlation process between the template image and the input image. And

An image monitoring apparatus to which the image monitoring method of the present invention can be applied includes a television camera for photographing the monitoring area,
Equipped with an image processing device that captures and processes camera images,
The image processing device, in the image of the normal state in advance,
A template storage unit for registering an image of a predetermined area as a template image, a similarity calculation processing unit for obtaining a correlation between the template image and an input image corresponding to the predetermined area each time the monitoring area is photographed, and the similarity degree. It is characterized by having a similarity determination processing unit for determining an abnormal state of the monitoring area from a change state that decreases.

Further, the image processing apparatus further comprises template updating means for updating the template image with the input image when the decrease in the similarity does not occur continuously.

A road monitoring apparatus according to one application of the present invention includes a television camera for taking a bird's-eye view of a monitoring area on a road and an image processing apparatus for capturing and processing a camera image, and the image processing apparatus is a vehicle or the like in advance. Among images that do not exist, a template storage unit for registering an image of a predetermined area as a template image, and similarity calculation for obtaining a correlation between the template image and an input image corresponding to the predetermined area each time the monitoring area is photographed. It is characterized by including a processing unit and a similarity determination processing unit that determines the presence of a vehicle when the state in which the similarity is equal to or less than a threshold value continues.

Further, the similarity determination processing unit is characterized by determining that a vehicle is present when the state of decrease in the similarity continues for a predetermined number of times or a predetermined time or more.

Further, when the similarity exceeds the threshold value and then recovers after exceeding the threshold value, the similarity determination processing unit treats this decrease section as one vehicle and the number of passing vehicles. It is characterized by performing measurement of.

Further, the similarity determination processing unit inputs the time of the section in which the similarity continuously decreases, calculates speeds for a plurality of vehicle lengths preliminarily assumed, and calculates the speed among the obtained speeds. It is characterized in that the closest value is selected in comparison with a predetermined vehicle speed in the past, and the vehicle speed and the vehicle length of the vehicle are obtained.

Alternatively, the similarity determination processing unit sets the threshold value as a first threshold value and sets a second threshold value at which the minimum value of the similarity degree is always less than that at a tail lamp position or the like. If there is a minimum value that is set and that is less than or equal to the second threshold value within a section that is continuously lower than the first threshold value, the time before and after the minimum value within the section is set. Based on the above, the vehicle speed and the vehicle length of the vehicle, and further the vehicle type are obtained.

An elevator waiting passenger monitoring device, a train platform monitoring device, a railroad crossing monitoring device, and the like to which the present invention is applied have the same basic configuration as the road monitoring device, respectively, and have an object such as a person or a vehicle entering each monitoring area. Is configured so that it can be monitored.

The operation of the configuration of the present invention will be described. In the present invention, of the background images without intruders in the surveillance area,
A narrow rectangular area equivalent to that normally used for template pattern matching is cut out and used as a template image. In the case of a road, for example, this template image reflects the natural characteristics of the road surface, but no special mark is required. Furthermore, during online monitoring, the input image that is periodically captured may be the same small area as the template image, and the density matching process between the template image and the input image is performed each time the input is performed, and the similarity is calculated. The calculation of the degree of similarity is based on well-known correlation processing or the like, but is performed only once for the grayscale data for the template area. For object recognition in the area, the degree of similarity is stored every time and the number of consecutive drops is determined by a simple algorithm. Therefore, the processing time can be extremely shortened as compared with the conventional pattern matching by the tracking processing of the entire monitoring area. Therefore, the processing cost can be significantly reduced.

Further, in the case of the road monitoring device, not only passing vehicles but also vehicle speed, vehicle length, etc. can be measured by a simple algorithm for analyzing the change state of the similarity, so that it is suitable for traffic flow abnormality monitoring and the like. A high vehicle detection sensor can be provided.

[0022]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the drawings.

[First Embodiment] FIG. 1 is a configuration diagram of a vehicle monitoring system according to a first embodiment of the present invention. The vehicle monitoring device 10 includes an A / D converter 2 for converting a video image of the TV camera 1 into a digital signal, an image memory 3 for storing a digitized video signal, and a template setting means 5 for designating an arbitrary area of the image memory. , A template image memory 4 for storing the image cut out by the template setting means 5 as a template image, a similarity processing means 7 for calculating the similarity between the template image and the input image, and a similarity storage means 8 for sequentially storing the similarity. , A similarity analysis means 9 for analyzing changes in the stored similarity, a timer 6 for managing time, and a CPU for controlling these processing units.

Next, the operation of the vehicle monitoring device will be described. The area to be stored as a template is set in advance by the template setting means 5. The template is a linear area that intersects the road, and is an image that reflects the natural characteristics of the road surface when no vehicle is present.

FIG. 2 shows how to import the template and the input image. As shown in the figure, the camera image 11 of the road surface where the car is not photographed is A / D converted, the designated area 12 is cut out from the grayscale image, and this is stored in the template image memory 4 as the template T. Next, the camera images 13 being monitored online are sequentially captured and A / D converted, and the input image of the vehicle detection window 14 in the same region as the template is cut out from the grayscale image and stored in the image memory 3. The similarity processing means 7 obtains the similarity by one gray-scale matching of the template image and the input image, and stores the similarity in the similarity storage unit 8 in the order of photographing.

FIG. 3 shows the transition of the input image and the change in the degree of similarity according to the measurement order. As shown in the figure, since no vehicle is photographed in the input image f (t), the similarity between the registered template image and the input image is high, and a value close to a value of 1.0 can be obtained. Whenever the vehicle gradually enters the detection window every time a picture is taken at a fixed time interval dt (for example, 33 milliseconds or 66 milliseconds), the input image changes from f (t + dt) to f (t + 4dt). , The degree of similarity decreases. f (t + 5
In the input image dt), the similarity increases again when the vehicle has finished passing through the detection window.

FIG. 4 is a graph showing how the degree of similarity changes. When the degree of similarity falls below a threshold value, for example, 0.8, it is determined that some object exists in the vehicle detection window 14 (processing area) (hereinafter referred to as ON), and this ON state remains for a predetermined time or longer. If it continues, it is judged as a vehicle. The number of passing vehicles can be obtained by counting the number of vehicles determined to be vehicles.

The length of the ON state is proportional to the passage time of the vehicle and changes depending on the vehicle speed and the vehicle length. If the ON time is extremely long, it can be determined that the vehicle is stopped or traveling at a low speed. In this way, by calculating the degree of similarity at regular time intervals, it becomes possible to detect an abnormal state of traffic flow such as traffic jams and accidents as well as the number of passing vehicles.

FIG. 5 shows a flowchart of the vehicle detection processing according to this embodiment. Template setting means 5 in advance
The processing area is set by (S1), the image of the monitored road surface where no vehicle exists is taken in, and the processing area in this image is registered in the template image memory 4 (S2). Next, the on flag is set to 0 (S3), and the infinite loop of online vehicle monitoring on the monitored road surface is activated (S4). First,
An image is input (S41), the current time time is read (S42), the similarity processing unit 7 performs correlation processing between the input image and the registered template to obtain the similarity, and this value is stored in the similarity storage unit 8. Yes (S43).

Next, the similarity analysis processing unit 9 determines whether the calculated similarity is lower than the threshold value (S44). If the similarity is low (Yes), there is something on the road surface, so the state is on (on flag = 1). Is stored, and if the previous time is the ON flag = 0, the current time time is stored in the start time start (S4
41). On the other hand, when the calculated similarity is higher than the threshold value, it is determined whether the previous time is the ON state (S443), and if so (Yes), the current time time is stored as the end time end (S4431), and the end-start is performed. Is a vehicle for a certain time or more, otherwise it is determined to be noise (S44).
32). Since the image input is performed in a constant cycle, it can be determined that the vehicle is a vehicle when the number of measurements from the start time to the end time is a predetermined number or more as shown in the figure.
These operations are the same as C
Each processing unit is controlled by the PU.

By the way, the similarity processing means 7 obtains the gray-scale matching between the image data of the image memory 3 and the image data of the template image memory 4, and executes the following normalized correlation calculation. That is, when the input image f (x, y) is normalized and correlated with the template t (p, q),
The similarity r (u, v) of the input image f (u, v) point is calculated by Equation 1.

[0032]

[Equation 1]

Here, p and q are x of the template image,
It is y size. This normalized correlation processing is well known, and the speed can be further increased by using a dedicated LSI. In addition,
The similarity between the template image and the input image may be obtained by another method.

As described above, according to the vehicle monitoring system of the present embodiment, the existence of the vehicle is determined from the background image of the linear region of the road surface as a template and the side or the state of similarity with the input image from the same region. Since the conventional tracking process is not required, the processing time can be significantly reduced, and real-time processing can be performed without using a high-speed computer. Also,
Since the processing area is the same as the template area and is a narrow area, it is highly sensitive to changes in the image in the processing area, and for example, an image without a tail lamp in a tunnel can be detected, and the measurement accuracy of the number of units is improved.

By using the object detection method based on this similarity, it is possible to detect objects other than the vehicle. For example, when a person smaller than the vehicle, a bicycle, or the like intrudes on the motorway, the degree of decrease in similarity is smaller than that of the vehicle. Therefore, the above threshold value (for example, 0.8) is used as the first threshold value, and the minimum value of the similarity is always less than the second value in the case of vehicle determination.
If the minimum value of the similarity continuously lower than the first threshold value is larger than the second threshold value, it may be determined as an object other than the vehicle. it can.

Alternatively, it can be applied to the measurement of the number of waiting passengers of an elevator as a target other than the road. That is, it is possible to measure the presence or absence of a person by registering a predetermined area near the entrance of an elevator hall where there is no person in the template image and obtaining the similarity using the same area as the input image.
Further, the number of waiting customers can be grasped by dividing the measurement area into a plurality of areas, registering template images for each, obtaining the degree of similarity, and totaling the number of people who pass through and the number of people who are stagnant in the entire measurement area.

[Embodiment 2] Next, as a second embodiment of the present invention, a speed calculation method by vehicle length estimation will be described. When the road is photographed as described above, the length of the similarity ON state changes depending on the vehicle speed and the vehicle length. Therefore, it is possible to obtain the vehicle speed and the vehicle length by estimating the vehicle length.

The vehicle monitoring system of this embodiment has basically the same configuration as that of the above-described embodiment, and the function of the similarity analysis processing unit 9 of FIG. 1 is expanded to add vehicle length / speed processing means. .

FIG. 6 shows processing blocks of the vehicle length / speed processing means. The ON state time input unit 15 and the vehicle length are, for example, 4
Assuming m, 6m, 8m, and 10m, an assumed speed calculation unit 16 that calculates an assumed speed Vi = (vehicle length / ON state time) from the ON state time, and gives an average speed Va from a plurality of vehicle speeds up to a certain time past Average speed calculator 17, each assumed speed Vi
Comparing the average speed Va with the average speed Va, the vehicle speed / vehicle length selection unit 18 selects the assumed speed Vi with the smallest error and the vehicle speed V corresponding thereto. Note that the following speed of the vehicle may be used to select the assumed speed Vi that is closest to the vehicle speed V of the preceding vehicle, instead of the average speed Va.

By the way, in the case of tunnel monitoring, the car is often photographed from behind. Therefore, the ON state time is longer than the actual passing time of the vehicle on the corresponding road surface of the vehicle detection window. As shown in the vehicle detection mechanism of FIG. 7, the vehicle detection line from the TV camera to the detection road surface (corresponding to the vehicle detection window 14) is initially a roof portion (vehicle height H) of the vehicle tip.
And finally near the road surface at the rear of the vehicle. Therefore,
Given that the angle θ between the detection line and the road surface is, the apparent vehicle length appears to be longer than the actual vehicle length L by H / tan (θ),
Therefore, the ON state time becomes longer than the actual passing time.

Therefore, when the vehicle speed is calculated assuming the vehicle length, the vehicle height is also estimated from the vehicle length. For example, when the vehicle length is assumed to be 4 m, the vehicle height may be 1.5 m, and when the vehicle length is assumed to be 6 m, the vehicle height may be 2 m and so on.

According to the present embodiment, past vehicle speeds such as a plurality of previously assumed vehicle lengths and preceding vehicles are used as reference values, so that the vehicle speed and vehicle length can be measured extremely easily and at high speed. Of course, this value is an estimated value, and there is a problem in using it for speed monitoring and the like. However, it can be practically used for tunnel ventilation control by distinguishing between large and small, and traffic flow condition monitoring such as congestion. As a result, a multifunctional vehicle measurement sensor capable of detecting the number of vehicles, vehicle speed, and vehicle length (or vehicle type) can be provided at low cost.

Next, as a modification of this embodiment, a method of obtaining the speed and the vehicle length from the position information of the tail lamp will be described. This embodiment is suitable for vehicle monitoring or ventilation control in a tunnel.

FIG. 8 shows a transition of the degree of similarity when the taillight is turned on. In the figure, (a) is an example of a small vehicle, and (b) is an example of a large vehicle. As for the similarity of passing vehicles, the lighting light is the detection window 1
It becomes the lowest value when passing 4. For this reason,
First, a time A from a time Tstart when the similarity becomes smaller than a first threshold value (for example, 0.8) to a time Tmin when the similarity becomes the minimum similarity, and a similarity from Tmin to a first threshold value or less. The vehicle length L can be roughly estimated by the equation 2 using the time B value up to the last time Tend.

[0045]

(2) L = 4A / B However, in the case of a small vehicle with a vehicle length of 4 m, A≈B, and the larger the size, the longer the time A. It should be noted that if the minimum similarity of the tail lamp unit is set to a second threshold value (for example, 0.2) or less, the detection accuracy is improved.

The processing procedure according to this embodiment is as follows. (1) The degree of similarity is calculated and stored in the degree-of-similarity storage unit 8. (2) Times Tstart and Tend are calculated until the degree of similarity becomes smaller than a predetermined threshold value and increases again. (3) From the time Tstart From the similarities measured up to Tend, the time Tmin with the smallest similarity is obtained, and (4) the vehicle length L is the time A from the time Tstart to the time Tmin with the minimum similarity and the time with the smallest similarity from Tmin. The vehicle length L = 4A / B is calculated using the value of the time B until the last Tend.

FIG. 9 is an explanatory diagram of the apparent passage time when the tail lamp is turned on. As shown in the figure, when there is a feature that the degree of similarity of the tail lamp part is the minimum, the relationship of (1) and (2) of Formula 3 is obtained between the vehicle speed V, the vehicle length L, and the vehicle height H. To be

[0048]

[Equation 3] L + H / tan (θ) = (A + B) * V1 (1) L + (H−h) / tan (θ) = A * V2 (2) where h is the height of the tail lamp section. is there.

The vehicle height H, the vehicle length L, and the tail lamp height h are unknown, and it is difficult to obtain them by image processing. Therefore, various combinations of vehicle height, vehicle length, and tail lamp height are prepared in advance. Table 1 shows an example of combination data and calculation results.

[0050]

[Table 1]

Substituting each combination of L, H, and h and the measured values of A and B into (1) and (2) of the equation 3, the speeds V1 and V
2 is obtained, and the value at which they are closest to each other is set as the vehicle speed V. Further, from the combination data corresponding to the vehicle speed V, the vehicle length L, the vehicle height H
To estimate. According to this, since it is possible to estimate a rough vehicle shape, it is possible to estimate, for example, the number of trucks or buses having a large amount of smoke emission and contribute to improvement of ventilation efficiency of ventilation control in the tunnel.

In the case of a vehicle in which the tail lamp is not turned on, the minimum similarity is the second threshold value (for example, 0.
2) Since it does not become less than the following, in this case, it will be calculated as a small vehicle, although there will be some error. Of course, it is possible to use the estimation method shown in FIG. 6 together.

Further, when the vehicle is photographed from the front,
Although the tail lamp has no characteristic, the similarity decreases, for example, in the windshield portion, and the similarity decreases when the headlight passes when the headlight is on. This can be dealt with by establishing a relational expression with the velocity similar to the equations 1 and 2 of the equation 3 from the positional relation of the characteristic points.

As described above, according to the modified embodiment, the vehicle speed, the vehicle length, etc. can be extremely easily obtained when the similarity pattern such as the tail lamp position of the passing vehicle appears.

[Third Embodiment] The template image registered in advance needs to be updated because it will be adversely affected if a building shadow or the like occurs on the monitored road surface. In the present embodiment, when the value of the similarity between the template and the input image continues to be high, attention is paid to the fact that there is no vehicle, and the template update process is performed simply and online.

FIG. 10 is a block diagram of a vehicle monitoring system according to the third embodiment of the present invention. Template image updating means 20 is added to the configuration of FIG. The template image updating means 20 has an update counter, and when the similarity higher than a threshold value (for example, 0.8) is continuously counted a predetermined number of times or more, the template image is cut out from the input image at that time, and the template image is extracted. The image memory 4 is updated. At this time, there is also a method of averaging by a linear combination operation instead of simply replacing.

FIG. 11 shows a flowchart of the vehicle detection processing according to this embodiment. In the processing step of FIG.
44 and S445 for updating the template are added. When the degree of similarity is lower than the threshold value and the start time start is set (S442), then,
The background update counter is set to 0 (S4421).

On the other hand, if the degree of similarity is higher than the threshold value and step S443 is processed, then the background update count is incremented (S444), and it is determined whether the background update count is a predetermined number or more (S445). If it is the predetermined number of times or more, the background update processing is executed (S4451), and the background update count is set to 0 (S4452). If it is less than the predetermined number of times, the process returns to S41.

By performing such a template image update process, when the brightness of the road surface changes due to the shadow of the building, the weather, or the passage of time, if the change is uniform within a small area of the vehicle detection window (almost If so, the similarity can be calculated accurately without being affected by the change in brightness. Therefore, there is no need to prepare and switch templates of various conditions in advance, which is convenient.

The image monitoring method and apparatus according to the present invention have been described above through a plurality of embodiments of the road monitoring apparatus.
However, the present invention can be applied not only to roads but also to various monitoring applications, and has the following applications, for example.

(1) An image near the entrance of an elevator is registered in a template, and when the degree of similarity is low, it is determined that a person is waiting.

(2) The image of the train platform is registered in the template to monitor the presence of people when the train approaches,
Roughly measure the number of people.

(3) The track under the platform is registered in the template image, and when the degree of similarity decreases, it is determined that a person is getting on the track.

(4) An image in which there is no vehicle at the railroad crossing is registered in the template, and the similarity with the input image is obtained to determine whether or not the vehicle is staying inside the railroad crossing when the railroad crossing is blocked.

[0065]

According to the present invention, the degree of similarity between the template image registered in the absence of an object and the input image of the same processing area that is sequentially captured is calculated, and the continuous decrease in the degree of similarity results in the monitoring area. Since the presence of the object is detected, the processing is simple and real-time intrusion detection is possible. Further, since it is not necessary to use a complicated algorithm or ultra-high-speed hardware, it is possible to provide an image monitoring device using hardware with low processing cost.

Further, according to the present invention, when the similarity continuously shows a high value while online, the input image is updated as the template image, so that it is possible to follow the change in the brightness of the monitoring environment and the detection accuracy is improved. There is an effect that can be improved. Further, it is possible to provide a highly reliable and easy-to-use image monitoring device.

According to the road monitoring method of the present invention to which the above image monitoring method is applied, the vehicle detection window of the TV camera is used as the processing area, and the passing vehicle is determined based on the change in the similarity between the template image in the absence of a vehicle and the online input image. Is detected, there is an effect that the processing is simpler and faster than the conventional method in which the background image of the entire monitoring area is tracked and the grayscale pattern matching is performed. Moreover, since the processing area is small, the sensitivity of vehicle detection is improved, and it becomes possible to detect vehicles without a lamp in the tunnel.

Further, according to the road monitoring method of the present invention, the template image can be easily updated while online, so that it is possible to avoid the adverse effects of the change in the brightness of the monitored road surface due to time, weather, shadows of buildings, etc. The detection accuracy is improved and it is easy to use.

Further, according to the road monitoring system of the present invention,
Since it is possible to easily estimate the vehicle speed and the vehicle length from the characteristics of the duration of change in the similarity or the state of change, it is possible to provide an inexpensive vehicle detection sensor suitable for monitoring abnormalities in traffic flow.

Further, according to the road monitoring system of the present invention,
Since remarkable similarity changes depending on vehicle lighting positions such as tail lamps can be used, it is possible to identify large vehicles with large smoke emissions from estimation of vehicle length, vehicle height, etc., and provide useful information to tunnel ventilation control etc. It will be possible.

Furthermore, the image monitoring method of the present invention can be applied to an elevator hall monitoring device, a train platform monitoring device, a railroad crossing monitoring device, and the like, and has an effect that these monitoring devices having a huge number of application points can be provided at low cost. .

[Brief description of drawings]

FIG. 1 is a configuration diagram of a vehicle monitoring system according to a first embodiment of the present invention.

FIG. 2 is an explanatory diagram showing how to import a template and an input image according to the first embodiment.

FIG. 3 is a transition diagram of input images in the order of import and similarity.

FIG. 4 is a graph showing a temporal change in similarity.

FIG. 5 is a flowchart of vehicle detection processing according to the first embodiment.

FIG. 6 is a processing block diagram of vehicle length / speed processing means according to a second embodiment of the present invention.

FIG. 7 is an explanatory diagram of a mechanism in which an apparent passage time becomes long.

FIG. 8 is a graph showing a temporal change in similarity between a small vehicle and a large vehicle when a tail lamp is turned on.

FIG. 9 is an explanatory diagram of an apparent passage time when the tail lamp is turned on.

FIG. 10 is a configuration diagram of a vehicle monitoring system according to a third embodiment of the present invention.

FIG. 11 is a flowchart of vehicle detection processing according to a third embodiment.

[Explanation of symbols]

1 ... Television camera, 2 ... A / D converter, 3 ... Image memory, 4 ... Template image memory, 5 ... Template setting means, 6 ... Timer, 7 ... Similarity processing means, 8 ... Similarity storage section, 9 ... Similarity analysis processing unit, 20 ... Template updating means.

Claims (16)

[Claims] 1. An image monitoring method for inputting an image of a surveillance area, and performing gray-scale matching with a normal state image to detect an abnormal state of the surveillance area, wherein a predetermined area is cut out from the normal state image in advance. Stored as a template image, the similarity between the input image corresponding to the predetermined area and the template image is calculated every time the monitoring area is repeatedly photographed, and the monitoring is performed when the decrease in the similarity continues for a predetermined number of times or more. An image monitoring method characterized by determining an area to be in an abnormal state. 2. An image monitoring method for inputting an image of a surveillance area, and performing gray-scale matching with an image in a normal state to detect an abnormal state in the surveillance area, wherein a predetermined area is cut out from the image in the normal state in advance. Stored as a template image, the similarity between the input image corresponding to the predetermined area and the template image is obtained every time the monitoring area is repeatedly photographed, and the similarity equal to or less than a predetermined threshold is the first predetermined number of times. The monitoring area is determined to be in an abnormal state when continuous for the above, and the template image is updated with the input image when the similarity exceeding the predetermined threshold continues for a second predetermined number of times or more. Image monitoring method. 3. The image monitoring method according to claim 1, wherein the predetermined area is a linear area or a linear rectangular area smaller than the monitoring area. 4. The image monitoring method according to claim 1, 2 or 3, wherein the similarity is calculated by a correlation process between the template image and the input image. 5. The image monitoring method according to claim 1, wherein the abnormal state is a state in which a foreign substance is present in or passed through the monitoring area. 6. A television camera for photographing the surveillance area,
In an image monitoring device including an image processing device that captures and processes a camera image, the image processing device may include a normal image in advance,
A template storage unit for registering an image of a predetermined area as a template image, a similarity calculation processing unit for obtaining a correlation between the template image and an input image corresponding to the predetermined area each time the monitoring area is photographed, and the similarity degree. An image monitoring apparatus, comprising: a similarity determination processing unit that determines an abnormal state of the monitoring area based on a changing state in which 7. The image processing apparatus according to claim 6, further comprising template updating means for updating the template image with an input image when the reduction in the similarity does not occur continuously. Image monitoring device. 8. A road monitoring device comprising a television camera for taking a bird's-eye view of a surveillance area on a road, and an image processing device for capturing and processing a camera image, wherein the image processing device preliminarily detects an image of a vehicle or the like that does not exist. A template storage unit for registering an image of a predetermined area as a template image; a similarity calculation processing unit for obtaining a correlation between the template image and an input image corresponding to the predetermined area each time the monitoring area is photographed; And a similarity determination processing unit that determines the presence of a vehicle when the degree of the vehicle is continuously lower than a threshold value. 9. The similarity determination processing unit according to claim 8, wherein the similarity determination processing unit determines that a vehicle is present when the reduced state of the similarity continues for a predetermined number of times or a predetermined time or more. Road monitoring device. 10. The decreasing section according to claim 8, wherein the similarity determination processing unit, when the state in which the similarity is equal to or less than the threshold value continues and then recovers beyond the threshold value, A road monitoring device characterized in that the number of vehicles passing through is measured as one vehicle. 11. The similarity determination processing unit according to claim 8, wherein the similarity determination processing unit calculates speeds for a plurality of pre-assumed vehicle lengths by inputting a time of a section in which the similarity continuously decreases. A road monitoring device characterized by selecting the closest value from a predetermined predetermined vehicle speed among a plurality of the obtained speeds to obtain the vehicle speed and the vehicle length of the vehicle. 12. The similarity determination processing unit according to claim 8, wherein the threshold is a first threshold, and the minimum value of the similarity is always less than that at a tail lamp position or the like. 2 threshold value is set in advance, and when there is a minimum value that is equal to or less than the second threshold value within a section continuously lower than the first threshold value, the A road monitoring device, characterized in that the vehicle speed and the vehicle length of the vehicle are obtained based on the time before and after the minimum value. 13. The similarity determination processing unit according to claim 8, wherein the threshold is a first threshold, and the minimum value of the similarity is less than or equal to a first threshold in the case of vehicle determination. And a threshold value of 2 is set, and when the minimum value of the similarity in the section continuously lower than the first threshold value is larger than the second threshold value, the vehicle is displayed in the monitoring area. A road monitoring device characterized by determining that there is an object other than. 14. An elevator waiting customer monitoring device comprising a television camera for photographing a surveillance area near an entrance of an elevator and an image processing device for capturing and processing an image of the camera, wherein the image processing device has no waiting passengers in advance. Of the images, a template storage unit for registering an image of a predetermined area as a template image, and each time the monitoring area is photographed,
A similarity calculation processing unit that obtains a correlation between the template image and the input image corresponding to the predetermined region, and a similarity determination process that determines the presence of a waiting customer when the state in which the similarity is less than or equal to a threshold continues. And an elevator waiting passenger monitoring device. 15. A train platform monitoring device comprising a television camera for photographing a surveillance area near a train platform or a railroad track, and an image processing device for capturing and processing a video image of the camera, wherein the image processing device is for passengers in advance. A template storage unit for registering an image of a predetermined area as a template image among the non-existing images, and a similarity calculation processing unit for obtaining a correlation between the template image and an input image corresponding to the predetermined area each time the monitoring area is photographed. And a similarity degree determination processing unit that determines the presence of a passenger or the like when the state in which the degree of similarity is equal to or less than a threshold value continues, and a train platform monitoring apparatus. 16. A railroad crossing monitoring device comprising a television camera for photographing a surveillance area near a railroad crossing and an image processing device for capturing and processing an image of the camera, wherein the image processing device does not have an object such as a vehicle or a person in advance. A template storage unit for registering an image in a predetermined area of an input image of the images as a template image, and a similarity calculation for obtaining a correlation between the template image and the input image corresponding to the predetermined area each time the monitoring area is photographed. A processing unit, a similarity determination processing unit that determines the presence of an object when the state in which the similarity is less than or equal to a threshold value continues,
A railroad crossing monitoring device comprising: