JP3360163B2 - Traffic flow monitoring device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a traffic flow monitoring device, and more particularly to a traffic flow monitoring device using image processing means.
The present invention relates to means for determining a vehicle speed and a vehicle type, that is, a type of a large vehicle / small vehicle.

[0002]

2. Description of the Related Art As a vehicle sensor, a so-called existence sensor represented by an ultrasonic sensor has become widespread. The presence sensor outputs a signal indicating the presence of the vehicle when the vehicle passes below the sensor, and is mainly used for counting the number of vehicles.

On the other hand, recently, a system has been developed in which a road image taken by a television camera is processed by image processing means to measure the number of vehicles. For example, in Japanese Patent Application Laid-Open No. Hei 7-244795, a periodic mark is installed on the road surface,
The density value at the mark position is binarized, and the presence or absence of a vehicle is determined based on whether or not it is the same as the mark.

[0004] Such an image type sensor has a merit that a single camera can measure a plurality of lanes, as compared with a conventional ultrasonic sensor, since it can process a monitored object in a two-dimensional manner. Therefore, image-type sensors are used to monitor vehicle speed, determine whether the vehicle is running at low speed due to traffic congestion, etc., and monitor for abnormalities, such as monitoring traffic flow in a tunnel. ing. For example, in Japanese Patent Application Laid-Open No. 5-298591, a vehicle is detected, and an image of the vehicle is tracked by shading pattern matching to measure a vehicle speed.

[0005]

However, Japanese Patent Application Laid-Open No. 7-24 / 1990
In No. 4795, it is necessary to set a mark on the road surface, and furthermore, since the binarization process detects whether the vehicle has blocked the mark, a method that can change the threshold value flexibly and quickly is not realized. As far as it is concerned, there are drawbacks such as difficulty in responding to changes in brightness.

[0006] Further, there is a great need for a vehicle detection sensor with a function capable of measuring not only the number of vehicles but also the vehicle speed and vehicle type, that is, the type of large / small vehicles. However, it has not been able to adequately meet this need.

On the other hand, in the light and shade pattern matching method disclosed in Japanese Patent Application Laid-Open No. 5-298591, the vehicle speed and the number of vehicles can be measured with high accuracy. However, in order to track a part of the vehicle, the vehicle width must be determined. Therefore, highly accurate judgment cannot be expected.

Further, in Japanese Patent Application Laid-Open No. Hei 6-309588, a grid-like pattern is provided on the road surface to determine the absolute difference between the background image and the new image of the vehicle. Therefore, it is necessary to always maintain the grid-like pattern. However, there has been a drawback that the detection sensitivity is greatly affected by a change in brightness due to a change in weather or the passage of time.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a traffic flow monitor provided with means for detecting a vehicle with high accuracy and measuring the number of vehicles passing therethrough, the vehicle speed and the type of vehicle without setting a special mark on a road surface. It is to provide a device.

[0010]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention provides a road surface pattern when a vehicle does not exist in a vehicle detection area defined in an input image from a television camera looking down at a road. Road surface template setting means for registering as a road surface template image, a similarity calculation processing unit for calculating a similarity between the road surface template image and an input image sequentially taken in, and a vehicle for analyzing the transition of the similarity and determining the passage of the vehicle A passage determination processing unit, a vehicle image registration processing unit that registers a vehicle end position when the vehicle passes through the vehicle detection area, and a vehicle tracking that sequentially updates the vehicle end position from an sequentially captured input image to track the vehicle A processing unit;
A speed calculation processing unit that calculates the vehicle speed from the tracked distance and the time therebetween, and a vehicle type determination processing unit that determines the vehicle type of the passing vehicle from the vehicle speed and the time required for the vehicle to pass through the vehicle detection area. We propose a traffic flow monitoring device consisting of

According to the present invention, in order to achieve the above object, a road surface pattern when a vehicle does not exist in a vehicle detection area defined in an input image from a television camera looking down at a road is defined as a road surface template image. Road surface template setting means to be registered, a similarity calculation processing unit for calculating the similarity between the road surface template image and the input image sequentially taken, and a vehicle passage determination processing unit for analyzing the transition of the similarity and determining the passage of the vehicle A vehicle image registration processing unit that determines a vehicle end position when the vehicle passes through the vehicle detection area and registers a vehicle end image near the traveling direction as a tracking vehicle template, and an input image and a vehicle template that are sequentially captured. A vehicle tracking processing unit that calculates a similarity by a similarity calculation processing unit, sequentially updates a position having a high similarity as a vehicle position, and tracks the vehicle; A speed calculation processing unit that calculates the vehicle speed from the traced distance and the time in between, a vehicle type determination processing unit that determines the vehicle type of the passing vehicle from the vehicle speed and the time required for the vehicle to pass through the vehicle detection area We propose a traffic flow monitoring device consisting of

The vehicle passage determination processing section determines that the state in which the degree of similarity is smaller than the predetermined value continues for a predetermined time T and the predetermined time T
If the minimum similarity is less than or equal to a predetermined value, it is determined that the vehicle is present, and if the similarity is higher than the predetermined value, it is determined that the vehicle has passed.

[0013] The similarity calculation processing section is also means for calculating the similarity between the background template image set in the road surface template setting means and the input image, and the vehicle image registration processing section determines whether the similarity is equal to or less than a predetermined value. Means for determining that a vehicle is present.

In any case, the vehicle type determination processing section calculates speeds from several assumed vehicle lengths L and times T passing through the vehicle detection area, respectively, and calculates the actual speeds and vehicle speeds obtained in the vehicle tracking process. This is a means for comparing a plurality of speeds calculated assuming the length L and determining the vehicle type based on the vehicle length L corresponding to the calculated speed with a small error.

The vehicle image registration processing section sets an existence determination processing area on another lane near the traveling direction of the vehicle detection area, stores a road surface pattern of the existence determination processing area in advance, and When the similarity with the corresponding input image is equal to or less than a predetermined value, the presence of the vehicle is determined, and the detection of overlap on another lane of the large vehicle using only the vehicle detection area is eliminated.

Further, the vehicle tracking processing section obtains an instantaneous speed of the vehicle, temporarily determines that the vehicle is stopped when the instantaneous speed is equal to or less than a predetermined value, and determines whether the vehicle template area used for tracking is a vehicle or a road surface. And means for determining a stopped vehicle only when determined to be a vehicle.

The road surface template setting means is a background image update processing means for updating the stored road surface template image using the input image when the similarity between the road surface template image and the input image is continuously higher than a predetermined value. including.

According to the present invention, a vehicle detection area is provided at a predetermined location of an image obtained by the image sensor, a road surface pattern of the detection area is registered in advance, and the registered road surface pattern and an image to be sequentially input are provided. The similarity at the same position is calculated, and if the similarity falls below a predetermined value, it is determined that an object is present in the detection area. If the similarity rises above the predetermined value, it is determined that the vehicle has passed. I do. Find the end coordinates of the vehicle in an area that is close in the direction of travel of the vehicle detection area when it is determined that the vehicle has passed, register the image pattern around this coordinate, and where the registered image pattern moves in the next input image The vehicle is tracked by searching whether the vehicle has been operated, and the vehicle speed is calculated. Further, in the similarity processing with the road surface pattern, the time during which the similarity is small, that is, the time required for the vehicle to pass through the detection area, is obtained, and the length of the vehicle is determined using this time and the vehicle speed obtained in the tracking processing. The vehicle type, that is, the type of the large vehicle / small vehicle is determined. In this method, (Length of a large car) / (Length of a small car) has a large ratio, so the headlamp interval, which does not significantly differ between the vehicle types, is detected to determine the vehicle type. In comparison, a small car / large car can be determined with high accuracy.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, referring to FIGS.
An embodiment of the traffic flow monitoring device according to the present invention will be described.

FIG. 1 is a block diagram showing a system configuration of an embodiment of a traffic flow monitoring device according to the present invention. The traffic flow monitoring device 1 includes an A / D converter 2 that converts an image of a television camera 16 into a digital signal, an image memory 3 that stores a digitized image signal, and a correlation or similarity between two images. A similarity calculation processing unit 4 to calculate, a similarity storage unit 5 to store the calculated similarity, and a vehicle passage determination processing unit 6 to analyze a change in the stored similarity and determine the passage of the vehicle.
And a transit time measuring unit 7 for obtaining a transit time of the vehicle, and a template setting unit 8 for setting a vehicle detection area at a predetermined position.
A road surface template image memory 9 for storing a road surface template; a vehicle image registration processing unit 10 for obtaining the end coordinates of the vehicle at a timing when the vehicle passage determination processing unit 6 detects the passage of the vehicle; a vehicle template image memory 11 A vehicle tracking processing unit 12 that calculates the similarity between the image of the vehicle stored in the vehicle template image memory 11 and the input image stored in the image memory 3 and tracks the vehicle; A vehicle speed calculation unit 13 for obtaining a speed, a vehicle type determination processing unit 14 for determining a vehicle type from the passing time obtained by the passing time measurement unit 7 and the vehicle speed obtained by the vehicle speed calculation unit 13, and a clock for these processing units. It comprises a timer 15 to be supplied and a CPU 17 for controlling all operations.

FIG. 2 is a flowchart showing the general procedure of vehicle detection, counting, vehicle speed measurement, and vehicle type determination in the traffic flow monitoring device of FIG. In the initial setting 100,
The CPU 17 sets various parameters, initializes various tables, the template setting unit 8 creates a background image, sets a vehicle detection area at a predetermined position of the video of the television camera 16, and detects a vehicle in the vehicle detection area. The road surface template image when it is not used is registered in the road surface template image memory 9 to prepare for the subsequent measurement processing. In an infinite loop 101, a background image updating process 1
Repeat steps up to 10.

In the image input 102, the image of the television camera 16 is captured, converted into a digital signal by the A / D converter 2, and the digitized input image is stored in the image memory 3. In the current time reading 103, the time at which the video of the television camera 16 is captured is recorded. In the road surface similarity determination process 104, the similarity calculation processing unit 4 calculates the similarity between the template image registered in the road surface template image memory 9 and the input image taken in the image memory 3. The similarity storage unit 5 stores the similarity calculated by the similarity calculation processing unit 4. The vehicle passage determination processing unit 6 analyzes the transition of the degree of similarity between the road surface template image and the input image to determine whether the vehicle has passed. The passing time measuring unit 7 obtains a vehicle passing time when the vehicle passing determination processing unit 6 determines that the vehicle has passed.

In the tracking template registration processing 105, when the vehicle passes through the vehicle detection area, the end coordinates of the vehicle are obtained, and an image around the coordinates is registered in the vehicle template image memory 11. In the tracking processing 106, the vehicle tracking processing unit 12 calculates the similarity between the vehicle template images registered in the vehicle template image memory 11 and the sequentially input images, and calculates the coordinates most similar to the template. Track the vehicle. In the tracking template update processing 107, when the coordinates having the maximum similarity are found, the vehicle template image already stored is updated with a rectangular image in the vicinity thereof.

In the speed calculating process 108, the vehicle speed calculating unit 13 calculates the vehicle speed based on the initial registration coordinates and time and the final tracking coordinates and time registered by the vehicle tracking processing unit 12. In the vehicle type determination process 109, the vehicle type determination unit 14 determines the vehicle type based on the vehicle speed obtained in the vehicle tracking process 106 and the passage time obtained by the passage time measurement unit 7.
In the background image updating process 110, when the state where the degree of similarity is higher than the predetermined value, that is, the state where there is no vehicle, continues, the stored road surface template image is updated with the new input image.

Next, each processing procedure of FIG. 2 will be described in more detail.

100: << Initial Setting >> FIG. 3 is a flowchart showing an example of the processing procedure of the initial setting 100. Initial setting 100, before the measurement process,
Parameter setting 111 and initialization of various tables 11
2, the creation 113 of the background image and the setting 114 of the road surface template.

FIG. 4 is a diagram showing an example of setting a processing area for obtaining parameters. On the input image 115,
For example, the range 116 of the tracking process, the vehicle detection windows 117 and 118 for calculating the road surface similarity, and the vehicle end coordinate detection window 11
9, 120 and a vehicle presence determination window 121 are set.

After setting the parameters, various tables are initialized.

FIG. 5 is a diagram showing an example of setting a storage area for a background image. The background image generation 113 is a process of storing in advance a background image in which no vehicle exists, that is, a road surface pattern. There are several methods for setting the storage area, as shown in FIG. In the example of FIG. 5A, only the vehicle detection windows 117 and 118 are stored. In FIG. 5B, the vehicle detection windows 117 and 118 and the vehicle presence determination window 121
Is stored. As shown in FIG. 5 (2), the presence of the vehicle is determined by the background subtraction processing, the road surface pattern of this portion is stored, and if the similarity with the input image is low, the vehicle is determined to be present. It is also possible to make a determination. Further, as shown in FIG. 5C, the entire tracking range 116 including the vehicle end coordinate detection windows 119 and 120 may be stored. In this case, tracking processing range 1
The background can be used for the determination of a stopped vehicle, which will be described in detail later in 16, and the like.

101: << Infinite Loop >> Unless there is an instruction to stop or stop the processing, the procedure from the image input 102 to the background image update processing 110 described below is repeated.

102: << Image Input >> The image of the television camera 16 is captured, converted into a digital signal by the A / D converter 2, and the digitized input image is stored in the image memory 3.

103: << Read current time >> The time at which the video of the television camera 16 was captured for speed determination processing and the like is recorded.

104: << Road Surface Similarity Judgment Processing >> The similarity calculation processing section 4 calculates the similarity between the template image registered in the road surface template image memory 9 and the input image taken in the image memory 3. The similarity storage unit 5 stores the similarity calculated by the similarity calculation processing unit 4. The vehicle passage determination processing unit 6 calculates the similarity between the road surface template image and the input image, and analyzes the result to determine whether the vehicle has passed. The passing time measuring unit 7 obtains a vehicle passing time when the vehicle passing determination processing unit 6 determines that the vehicle has passed.

FIG. 6 is a diagram showing an example of the setting of the vehicle detection window. The template setting unit 8 presets an area to be stored as a road surface template, that is, a rectangular vehicle detection window 23 at a location where the vehicle enters. An area specified by the vehicle detection window 23 is cut out from the road image 20 where the vehicle 200 is not present, and is stored in the road surface template image memory 9 as a road surface template image 21. The similarity calculation processing unit 4
The similarity between the template image 21 and the image 23 sequentially input from the same position is obtained, and the obtained similarity is stored in the similarity storage unit 5.

FIG. 7 is a diagram showing an example of the state of change of the similarity, that is, the correlation value, and FIG. 8 is a diagram showing the state of change of the similarity in a bar graph. As shown in the input image 24 in FIG. 7, when the vehicle 200 is not photographed, the registered template image and the input image f (t) are similar, and the similarity close to 1.0 is obtained. can get. Vehicle 200
Gradually enters the field of view of the television camera 16, the image in the vehicle detection window 23 changes and the input images 25, 26,
27, 28, and 29, the similarity decreases. When the similarity is smaller than a predetermined value (for example, 0.8) as shown in FIG. 8, it is determined that the vehicle 200 is present in the processing area (ON), and the time of this ON state, that is, When the time required for the vehicle 200 to pass is equal to or greater than a predetermined value, it is determined that the vehicle 200 is present. in this case,
As shown in FIG. 8, when the minimum similarity min value in a section smaller than the threshold value is larger than a predetermined value, it can be determined that noise is present.

FIG. 9 is a flowchart showing an example of the processing procedure of the road surface similarity determination processing according to the present invention. Processing 3
0, an image in which a vehicle does not exist is fetched in advance, and processing 31 is executed.
In step 32, a road surface template image of the processing region is stored in the background template, that is, the road surface template image memory 9. In a process 33, an ON flag indicating vehicle passing is initialized to 0, and in a process 34,
Enter an infinite loop.

At step 35, an image is input, and at step 36,
The time is read, and in step 37, the input image and
The similarity with the road surface template image registered in step (1) is stored in the similarity storage unit 5 by the similarity calculation processing unit 4.

In a process 38, it is determined whether or not the obtained similarity is smaller than a predetermined value. If the similarity is smaller, in a process 39, it is determined whether or not the on flag is 0. It is determined that at least a vehicle or something is on the road surface, and in a process 40, the on-flag is changed to 1 and a process 41 is executed.
Sets the start time of the transit time measurement. In process 38,
If the correlation value is high and the previous time is in the on state in process 42,
In process 43, the end time of the passage time measurement is set, and in process 4
In step 4, the passing time measuring unit 7 obtains the number of measurements up to that time, that is, the vehicle passing time. If it is smaller than the threshold value, it is determined that noise is present. Then, the processing exits the infinite loop 34.

The similarity calculation processing section 4 is means for calculating the similarity between the template image registered in the road surface template image memory 9 and the input image fetched into the image memory 3, and includes a technique such as a normalized correlation processing. Is used. In the normalized correlation processing, the input image f (x, y) is converted into a template t (p,
When the correlation is normalized by q), the similarity r (u, v) of the input image f (u, v) point is calculated by Expression 1. Where p,
q is the x, y size of the template image. In addition,
It is known that the normalized correlation processing can be performed at high speed by a dedicated LSI (“High-performance image processing LSI [ISP-X]).
Development and Application ”: See Video Information, 1996.2). In addition, Equation 1 is divided into Equations 2 to 6, and these correlations can be obtained at high speed by a dedicated LSI.

[0040]

(Equation 1)

[0041]

(Equation 2)

[0042]

(Equation 3)

[0043]

(Equation 4)

[0044]

(Equation 5)

[0045]

(Equation 6)

Here, the normalized correlation processing is described as an example of a method for obtaining the correlation, but any other method may be employed as long as it is a method for obtaining the similarity between the template image and the input image.

FIG. 10 is a diagram showing an example of a measure for preventing erroneous detection peculiar to a large vehicle. As shown in FIG.
If the heavy vehicle 92 runs in the front lane, a problem occurs. In the input images 85, 86, 87, the detection windows 90,
91 is set for each lane, and when a large vehicle 92 passes through it, the large vehicle 92 also covers the detection window 91 of the lane in the right lane. As shown in FIG. 10B, a vehicle end coordinate 94 is generated in the vehicle end detection window 93.

To avoid this problem, in this embodiment, as shown in FIG. 10C, a vehicle presence determination window 95 is further provided in the vehicle end detection window 93, and the vehicle presence determination window 95 If no vehicle exists, the end of the vehicle is not detected in the right lane lane. According to this measure, erroneous detection of the number of passing vehicles by the large vehicle 92 can be prevented. Also in the case of the presence check by the vehicle presence determination window 95,
The background image of the vehicle presence determination window 95 is stored in advance, and the similarity between the stored image and the input image is determined by the similarity calculation processing unit 4.
When the similarity is equal to or smaller than a predetermined value, it is determined that the vehicle exists.

FIG. 11 is a diagram showing an example of a processing procedure in which a vehicle presence confirmation process 139 as a measure for preventing erroneous detection unique to a large vehicle as shown in FIG. It is.

105: << Tracking Template Registration Processing >> FIG. 12 is a view for explaining the setting status of the processing area in the tracking template registration processing. The tracking template registration processing 105 is started when it is determined that the vehicle has passed in the road surface similarity calculation processing 104. As shown in FIG. 12, with respect to the input image 50, the vicinity of the traveling direction of the region of the vehicle detection window 51
A predetermined processing area 52 is set in (in this case, the upper part). The processing area 52 includes the speed of the passing vehicle and the TV camera 16.
Changes depending on the shooting direction. For the processing area 52, a process of searching for the end coordinates 53 of the vehicle is executed. Several methods are conceivable for this processing. Here, a description will be given of a difference processing with respect to a background image as an example.

FIG. 13 is a diagram showing an example of a situation of searching for the end coordinates of the vehicle. As shown in FIG. 13, the difference between the background and the input image is obtained. When the white vehicle or the tail lamp is lit, the difference from the background is large,
The result is as shown in FIG. With respect to this image, the cumulative distribution of the density in the horizontal direction is obtained, and a position ymax where the distribution becomes maximum is searched for. To obtain the horizontal position, FIG.
As shown in (2), the processing area 58 is limited to the vicinity of the obtained coordinate ymax, and as shown in FIG. 13 (3), the density accumulation distribution in the vertical direction is obtained. Regarding this distribution, as shown in FIG. 13 (4), positions xs and xe larger than a predetermined threshold value.
Is calculated, and the rectangular area centered on ymax is set to the vehicle end coordinate 5
Register as 9.

FIG. 14 is a diagram showing a procedure for registering vehicle end coordinates, and FIG. 15 is a diagram showing a procedure for registering an image of vehicle end coordinates as a template image.
FIG. 17 is a diagram illustrating an example of a data structure of a tracking management table used for vehicle tracking, and FIG. 17 is a diagram illustrating an example of the configuration of the entire tracking management table used for vehicle tracking.

If there is a vehicle, the end coordinates of the vehicle are detected at 143 as shown in FIG.
Then, it is determined whether or not the vehicle end coordinates exist. If yes, the center coordinates x and y are registered in process 145. After the registration, as shown in FIG. 15, it is determined in a process 146 whether or not the vehicle end coordinates have been detected. In a process 147, a rectangular template region is determined around the detected coordinates x and y, and its position is determined. Is registered as a template image. further,
The data is registered in the tracking management table of FIG. Here, as shown in FIG. 17, 24 tracking management tables are prepared, but the number of tracking management tables is not limited as long as the memory capacity is sufficient.

In the tracking management table of FIG. 16, a flag indicates whether the tracking management table is valid or invalid. When the flag is 1, a tracking process 106 described later is executed. The hold indicates whether or not the vehicle is being tracked, and the time of 1 indicates that the vehicle is being tracked. In the context of the initial registration coordinates, the xc coordinate is the x coordinate of the center of the template, and the yc coordinate is the y coordinate of the center of the template. The time is, for example, a time in milliseconds at the time of template registration. The passing lane number is stored as the detected vehicle detection window number. In the context of tracking data, the tracking x coordinate is the current vehicle end x coordinate, the tracking y coordinate is the current vehicle end y coordinate, and the time is the current time. The used template number is a template management number shown in FIG. In the detection area transit time, a vehicle transit time is stored for use in vehicle type determination.

106: << Tracking Process >> FIG. 18 is a flowchart showing an example of a tracking process procedure. As shown in FIG. 18, in the process 150, the processes 151 to 156 are looped by the number of the tracking management tables. In step 151, the flag in the tracking management table is checked, and if the flag is 1, the tracking process is performed. In the tracking process, a search area is set in a process 152, a maximum similarity coordinate is searched in a process 153, and a process 154 is performed.
Then, it is determined whether or not a vehicle is detected. If a vehicle is detected, the tracking management table is updated in step 155. If no vehicle is found in step 154, the hold flag is set to 0 in step 156. To end tracking.

FIG. 19 is a diagram showing a situation in which a vehicle is tracked while searching for a place most similar to the template between the registered image and the next input image. The image of the registered vehicle end coordinates is stored in the vehicle template image memory 11, and as shown in FIG. 19, a location which is most similar to the template between the registered image and the image input at the next time is determined. Follow the vehicle while searching. A search area 57 is set in the vicinity of the registered vehicle end position 53, and a point indicating the maximum similarity, that is, an x, y coordinate point, is searched within the range using the similarity calculation processing unit 4.

FIG. 20 is a flowchart showing an example of the tracking processing procedure. In process 60, the vehicle end coordinates are detected. In a process 61, the image of the rectangular region at the end of the vehicle is stored in the vehicle template image memory 11, a flag is set to 1 in a process 62, a current time is read in a process 63, and a process 6 is executed.
At 4, update to the stored current time.

In step 65, steps 66 to 73 are repeated until the flag becomes 0. In a process 66, an image is input, in a process 67, the time at that time is stored, and in a process 68,
A search area is set based on the registered coordinates. Process 69
Then, the similarity calculation processing section 4 obtains the similarity between the image stored in the vehicle template image memory 11 and the input image of the process 66, and searches for the coordinate point of the point of maximum similarity. In a process 70, it is determined whether or not the maximum similarity is higher than a predetermined value. If the maximum similarity is higher, the coordinates are stored.

If it is low, the flag is set to 0, and the tracking ends.

If the flag is set to 0, the tracking has been completed, and in a step 74, the vehicle speed is calculated based on the difference between the registration time and the end time and the distance traveled during that time.
In the present embodiment, the similarity calculation processing is used for the tracking processing. However, there is a method of executing the vehicle end detection processing of FIG. 19 every time and performing tracking while updating the coordinates.

107: << Tracking template update processing >> When the vehicle travels in the direction away from the television camera, the vehicle becomes gradually smaller as it goes deeper in the image. Update the template image while tracking the vehicle. In the present embodiment, when the maximum similarity coordinate is found, the rectangular image in the vicinity is replaced with the image of the area of the template management number already stored.

108: << Vehicle Speed Calculation Processing >> FIG. 21 is a flowchart showing an example of a processing procedure of the vehicle speed calculation processing. In step 161, the processes 162 to 166 are repeated by the number of tables. In step 162, the tracking management table of FIG. 16 is searched, and if there is a table whose flag is 1 and the hold flag is 0, the actual distance is calculated from the initial registration y coordinate and the end point y coordinate of the tracking management table, and the initial distance is calculated. The tracking time is calculated from the registered time and the current time, and the vehicle speed is calculated based on the actual distance / time. This vehicle speed corresponds to the average vehicle speed.

When the vehicle speed is obtained when the hold flag is 1, an instantaneous vehicle speed is obtained. In this case, it is necessary to additionally store the previous tracking coordinates and the previous time in the tracking management table. If this instantaneous vehicle speed is obtained every time, if the vehicle speed is equal to or less than a predetermined vehicle speed, it is determined that the vehicle is a low speed vehicle,
When the vehicle stops, it can be determined that the vehicle is stopped.

However, as for the stopped vehicle, if the vehicle end is not registered correctly, the vehicle template may be left behind without being able to follow the vehicle, and may be erroneously recognized as stopped, so that the vehicle has really stopped. It is determined whether or not. FIG. 22 is a flowchart illustrating an example of a processing procedure for determining whether the vehicle is stopped. As shown in FIG.
At 70, it is determined whether or not the vehicle has stopped due to lack of speed. If there is a stopped vehicle, it is checked at step 171 whether the difference between the background difference images of the tracking vehicle template area is large. If it is larger, it is determined in step 172 that there is a stopped vehicle. If it is smaller, tracking is stopped in step 173, and the flag and the hold flag are initialized to 0.

FIG. 23 is a flowchart showing an example of a processing procedure for determining whether a template has been left or a real stopped vehicle. As shown in FIG. 23, the difference between the background image 175 and the input image 176 obtained by the subtractor 179 is large because the left template 177 has a small difference from the background image in the template area. There are few parts. On the other hand, in the case of a real vehicle, when the difference 181 is obtained by the subtractor 180, a portion where the difference is large occurs. Based on this difference, it can be determined whether the template is abandoned or a real stopped vehicle.

109: << Vehicle Type Determination Process >> The vehicle type determination process determines the vehicle length, that is, the vehicle length, by using the vehicle passage time in the road surface similarity calculation process 104 and the vehicle speed obtained in the vehicle tracking process 106. This is a process for determining the type of a large vehicle / small vehicle.

FIG. 24 is a side view showing the principle of the vehicle type judgment processing 109. In this case, since the television camera 16 is installed facing downward, the time when the television camera 16 passes through the vehicle detection window 82 in the field of view of the television camera 16 is affected by the vehicle height. Detection window length L2
And the vehicle length L3 become the total distance L0 within the field of view of the television camera 16. This total distance L0 is obtained by Expression 7.

[0068]

(Equation 7)

FIG. 25 is a diagram showing an example of the vehicle transit time obtained based on the similarity value. Vehicle transit time TT
Is obtained as shown in FIG. 25, so that the apparent vehicle speed can be calculated by Expression 8 from the transit time TT and the total distance L0.

[0070]

(Equation 8)

Here, the length of the small car is set to 4.5 m,
The length of a large vehicle is 8 m. When each vehicle speed is calculated by Expression 8, and a vehicle length having a smaller error between the vehicle speed and the vehicle speed obtained in the tracking process is selected, a large vehicle / small vehicle can be determined with high accuracy. That is, in this method,
(Length of large car) / (Length of small car) = 8 / 4.5 ≒ 1.78
Therefore, a small car / large car can be determined with higher accuracy as compared with the conventional case in which the distance between the headlights, which is not so large, is detected and the vehicle type is determined.

110: << Background Image Updating Process >> FIG. 26 is a flowchart showing an example of the processing procedure of the background image updating process. If the road surface template image stored in advance is not used, the road surface similarity decreases even if the vehicle does not pass when the brightness changes due to a change in weather or the passage of time. In order to avoid this decrease in the road surface similarity, in the road surface similarity calculation processing 104, if the state in which the similarity is higher than a predetermined value, that is, the state where no car exists, continues, the stored road surface template image is newly input. Update with image. That is, in FIG. 26, it is determined in a process 157 whether or not anything is registered in the management table. If nothing is registered, it is determined in process 158 whether or not all the road surface similarities are high.
If all the road surface similarities are high, it is determined in process 159 whether or not there is a moving object in the entire area. If there is no moving object such as a vehicle, the input image is copied in step 160 as a background template, that is, a road surface template image. By updating the template image in this manner, it is possible to flexibly cope with a change in brightness due to a change in weather or the like.

Conventionally, it has been difficult to detect a vehicle whose tail lamp is not turned on in a tunnel. However, in the detection of a vehicle by calculating the similarity using the road surface template according to the present embodiment, it is satisfactory even for a vehicle whose lamp is not turned on. , The accuracy of the number measurement can be improved. In addition, the vehicle speed is obtained by the vehicle tracking process, and the vehicle type can be determined with high accuracy from the vehicle speed and the passage time of the detection area.

In the above embodiment, the present invention has been described with an example in which the vehicle moves away from the front of the screen to the back. However, when the vehicle approaches the front from the back of the screen, the processing area is moved forward in the traveling direction. It will be apparent that the present invention can be applied if the device is installed on the lower side (on the lower side of the screen).

The present invention is effective not only in the example of tunnel monitoring, but also in the measurement of the number of vehicles passing, the vehicle speed, and the vehicle length on an outdoor general road.

[0076]

According to the present invention, a vehicle detection area is provided at a predetermined position of an image obtained by an image sensor, a road surface pattern of this detection area is registered in advance, and the registered road surface pattern is sequentially input. The similarity at the same position as the image to be processed is calculated, and if the similarity falls below a predetermined value, it can be determined that some object exists in the detection area.

When the similarity exceeds a predetermined value,
Determines that the vehicle has passed, finds the end coordinates of the vehicle in an area that is close in the traveling direction of the vehicle detection area when it determines that the vehicle has passed, registers the image pattern around this coordinate, and registers the registered image pattern The vehicle can be tracked by searching where it has moved in the next input image, and the vehicle speed can be calculated.

Further, in the similarity processing with the road surface pattern,
Since the time during which the degree of similarity is small, that is, the time required for the vehicle to pass through the detection area, is obtained, the length of the vehicle is calculated using this time and the vehicle speed obtained by the tracking processing, and the type of the vehicle, that is, the type of the large vehicle / small vehicle Can be determined. in this way,
Since (Length of large car) / (Length of small car) has a large ratio, small car / large car can be determined with high accuracy.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a system configuration of an embodiment of a traffic flow monitoring device according to the present invention.

FIG. 2 is a flowchart showing a schematic procedure of vehicle detection, vehicle counting, vehicle speed measurement, and vehicle type determination in the traffic flow monitoring device of FIG.

FIG. 3 is a flowchart illustrating an example of a processing procedure of an initial setting process.

FIG. 4 is a diagram showing an example of setting of a processing area for obtaining parameters.

FIG. 5 is a diagram illustrating a setting example of a storage area of a background image.

FIG. 6 is a diagram illustrating an outline of an operation of a similarity calculation processing unit that calculates a similarity between a road surface template and a current input image.

FIG. 7 is a diagram illustrating an example of a state of change of a similarity, that is, a correlation value.

8 is a diagram showing a change state of the similarity in FIG. 7 in a bar graph.

FIG. 9 is a flowchart showing a processing procedure up to a vehicle passage determination processing.

FIG. 10 is a diagram showing an example of a measure for preventing erroneous detection peculiar to a large vehicle.

FIG. 11 shows a vehicle passage determination process of FIG. 9;
It is a figure which shows an example of the processing procedure which added the vehicle presence confirmation process etc. which are the measures which prevent the false detection peculiar to a large vehicle like 0.

FIG. 12 is a diagram illustrating a setting state of a processing area in the tracking template registration processing.

FIG. 13 is a diagram illustrating an example of a situation of searching for end coordinates of a vehicle.

FIG. 14 is a diagram showing an example of a procedure for registering vehicle end coordinates.

FIG. 15 is a diagram showing an example of a procedure for registering an image of vehicle end coordinates as a template image.

FIG. 16 is a diagram illustrating an example of a data structure of a tracking management table used for vehicle tracking.

FIG. 17 is a diagram showing an example of the configuration of the entire tracking management table used for vehicle tracking.

FIG. 18 is a flowchart illustrating an example of a tracking processing procedure.

FIG. 19 is a diagram illustrating a situation in which a vehicle is tracked while searching for a location that is most similar to a template between a registered image and a next input image.

FIG. 20 is a flowchart illustrating an example of a tracking processing procedure.

FIG. 21 is a flowchart illustrating an example of a processing procedure of a vehicle speed calculation process.

FIG. 22 is a flowchart illustrating an example of a processing procedure for determining whether the vehicle is stopped.

FIG. 23 is a flowchart illustrating an example of a processing procedure for determining whether the vehicle is a leaving template or a true stopped vehicle.

FIG. 24 is a side view showing the principle of the vehicle type determination process.

FIG. 25 is a diagram illustrating an example of a vehicle passage time obtained based on a similarity value.

FIG. 26 is a flowchart illustrating an example of a processing procedure of a background image update process.

[Description of Signs] 1 Traffic flow monitoring device 2 A / D converter 3 Image memory 4 Similarity calculation processing unit 5 Similarity storage unit 6 Vehicle passage determination processing unit 7 Passage time measurement processing unit 8 Template setting unit 9 Background template image Memory 10 Vehicle image registration processing unit 11 Vehicle template image memory 12 Vehicle tracking processing unit 13 Speed calculation processing unit 14 Vehicle type determination processing unit 15 Timer 16 TV camera 17 CPU 20 Input image 21 Road surface template 22 Input image 23 Vehicle detection window 24 ２９29 Input image 50 Input image 51 Vehicle detection window 52 Processing area 53 End coordinate 54 to 56 Input image 57 Search area 58 Processing area 59 Vehicle end coordinate 80, 81 Vehicle 82 Vehicle detection window 85 to 89 Input image 90, 91 Detection window 92 Large vehicle 93 Vehicle tail detection window 94 Vehicle tail seat 95 Vehicle presence determination window 115 Input image 116 Range of tracking processing 117, 118 Vehicle detection window 119, 120 Vehicle end detection window 121 Vehicle presence determination window 175 Background image 176 Input image 177 Left behind vehicle template 178 Vehicle template being tracked 179,180 Subtractor 181 Difference 182 Difference 200 Vehicle

──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor ▲ Taka ▼ Kazuya Hashimoto 7-1-1, Omikacho, Hitachi City, Ibaraki Prefecture Within Hitachi Research Laboratory, Hitachi, Ltd. (72) Inventor Katsunori Kobayashi Yukicho, Hitachi City, Ibaraki Prefecture 3-2-2, Hitachi Nuclear Engineering Co., Ltd. (72) Inventor Yoshiyuki Sato 5-2-1, Omika-cho, Hitachi City, Ibaraki Prefecture Hitachi, Ltd. Omika Plant (56) References JP-A-5-298591 (JP, A) JP-A-9-97335 (JP, A) JP-A-2-166598 (JP, A) JP-A 8-69596 (JP, A) JP-A-3-22099 (JP, A) Kaihei 9-282455 (JP, A) (58) Field surveyed (Int. Cl. ⁷ , DB name) G08G 1/00 G08G 1/04

Claims (8)

(57) [Claims] 1. A road surface template setting means for registering a road surface pattern when a vehicle does not exist in a vehicle detection area defined in an input image from a television camera overlooking a road as a road surface template image, the road surface template image A similarity calculation processing unit that calculates a similarity with the input image sequentially captured, a vehicle passage determination processing unit that analyzes a transition of the similarity and determines whether a vehicle passes, and a vehicle that passes through the vehicle detection area. A vehicle image registration processing unit for registering the vehicle end position when the vehicle has been moved; a vehicle tracking processing unit for sequentially updating the vehicle end position from the sequentially captured input image to track the vehicle; and A speed calculation processing unit for calculating a vehicle speed; and determining a type of the passing vehicle based on the vehicle speed and a time required for the vehicle to pass through the vehicle detection area. A traffic flow monitoring device comprising a vehicle type determination processing unit to determine the traffic flow. 2. A road surface template setting means for registering a road surface pattern when a vehicle does not exist in a vehicle detection area defined in an input image from a television camera overlooking a road as a road surface template image; A similarity calculation processing unit that calculates a similarity with an input image that is sequentially captured, a vehicle passage determination processing unit that analyzes a transition of the similarity and determines whether a vehicle passes, and a vehicle that passes through the vehicle detection area. A vehicle image registration processing unit that obtains a vehicle end position and registers a vehicle end image in the vicinity of the traveling direction as a vehicle template for tracking, and a similarity calculation processing unit that calculates the similarity between the sequentially captured input image and the vehicle template A vehicle tracking processing unit that sequentially updates the position having a high degree of similarity and calculates the vehicle position as a vehicle position and tracks the vehicle; A traffic calculation processing unit that calculates a vehicle speed from a distance between the vehicle and a vehicle type determination processing unit that determines a vehicle type of a passing vehicle based on the vehicle speed and a time required for the vehicle to pass through the vehicle detection area. Monitoring device. 3. The traffic flow monitoring device according to claim 1, wherein the vehicle passage determination processing unit determines that the state in which the similarity is smaller than a predetermined value continues for a predetermined time T, and the minimum value in the predetermined time T is A traffic flow monitoring device characterized by means for judging that a vehicle is present when the similarity is equal to or less than a predetermined value, and judging that the vehicle has passed when the similarity becomes higher than the predetermined value. 4. The traffic flow monitoring device according to claim 2, wherein the similarity calculation processing unit is also a unit that obtains a similarity between the background template image set in the road surface template setting unit and the input image. The traffic flow monitoring device, wherein the vehicle image registration processing unit is means for determining that a vehicle is present when the similarity is equal to or less than a predetermined value. 5. The traffic flow monitoring device according to claim 1, wherein the vehicle type determination processing unit determines a speed based on several assumed vehicle lengths L and a time T passing through the vehicle detection area. Is calculated respectively,
Means for comparing the actual speed obtained in the vehicle tracking process with a plurality of speeds calculated assuming the vehicle length L, and determining a vehicle type based on the vehicle length L corresponding to the calculated speed with a small error. A traffic flow monitoring device characterized by the following. 6. The traffic flow monitoring device according to claim 1, wherein the vehicle image registration processing unit sets the presence determination processing area on another lane near the traveling direction of the vehicle detection area. Setting, the road surface pattern of the presence determination processing area is stored in advance, and when the similarity between the road surface pattern and the corresponding input image is equal to or less than a predetermined value, it is determined that the vehicle is present. A traffic flow monitoring device, which is means for eliminating the detection of duplication on another lane. 7. The traffic flow monitoring device according to claim 1, wherein the vehicle tracking processing unit determines an instantaneous speed of the vehicle, and determines that the vehicle stops when the instantaneous speed is equal to or less than a predetermined value. A traffic flow monitoring device including means for temporarily determining whether a vehicle template area used for tracking is a vehicle or a road surface, and determining a stopped vehicle only when the vehicle is determined to be a vehicle. 8. The traffic flow monitoring device according to claim 1, wherein the road surface template setting means is configured to continuously perform a state in which the similarity between the road surface template image and the input image is higher than a predetermined value. A traffic flow monitoring device comprising a background image update processing means for updating a road surface template image stored in a vehicle using an input image.