JP2009146337A - Vehicle presence determination device, vehicle presence determination method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To attain accurate vehicle detection by relatively simple arithmetic processing in an image type vehicle detector. <P>SOLUTION: In an image type vehicle sensor 1, a control device 20 performs (1) detection of stopping vehicle by calculating a matching rate R by pattern matching of a photographic image taken by a camera 10 with a predetermined template image and comparing the matching rate with a predetermined threshold, and (2) detection of moving vehicle using an inter-frame difference method, and determines whether a vehicle is present or not based on both the detection results. Based on a moving speed V of a moving vehicle, an area where it is estimated that the other vehicle is absent in the front of the moving vehicle (vehicle-free estimated area) is determined, and the threshold of the matching rate R in the vehicle-free estimated area is changed to a value according to the matching rate R in this area which is calculated when a stopping vehicle is detected. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a vehicle presence / absence determination device that determines the presence / absence of a vehicle in a sensing area in a captured image.

In the road traffic field, an image type vehicle detector using a camera image is used as a device for measuring traffic flow on a road. In this image-type vehicle detector, for example, by performing a predetermined image analysis process on a photographed image by a camera installed above the road, for example, the number of vehicles that pass while the green light is lit or the vehicle that stops while the red signal is lit The number is detected (see, for example, Patent Document 1).
Japanese Patent No. 3215596

  As an image analysis method of a captured image in an image type vehicle sensor, a method of detecting the presence or absence of a vehicle in a captured image by performing pattern matching processing between the captured image and a predetermined template image is known. . At this time, a photographed image (background image) in a state where no vehicle is present is used as the template image. However, the luminance value of each pixel of the captured image changes depending on the weather such as sunny or rainy and the time zone such as day or night. In addition, the luminance value of each pixel of the captured image also changes due to changes in the shape and size of shadows such as buildings and street trees, and lighting of street lamps. For this reason, if the template image is fixed, the detection of the vehicle in the captured image is often mistaken, which causes a decrease in detection accuracy. Therefore, there is a method of changing the template image according to the environment. However, if the update timing is not properly determined, the detection accuracy may be lowered. The present invention has been made in view of the above circumstances, and an image-type vehicle detector has an object of realizing highly accurate vehicle detection by relatively simple arithmetic processing.

The first invention for solving the above-described problems is
Setting means (for example, the processing unit 100) for setting an image portion to be detected in a captured image captured by a camera viewed from above (for example, the camera 10 in FIGS. 1 and 2) as a sensing area;
A dividing unit (for example, the processing unit 100) for dividing the set sensing region into a plurality of partial regions;
A moving vehicle determination means (for example, a moving vehicle detection unit 110) that compares and analyzes time-series changes of the captured image to determine whether or not there is a moving vehicle in the sensing area;
A matching processing unit (for example, a stopped vehicle detection unit 130 in FIG. 2) that compares and analyzes the captured image and a predetermined template image by a predetermined pattern matching process to calculate a matching rate for each of the plurality of partial regions;
Stopped vehicle determination means for determining whether there is a stopped vehicle in the sensing region by comparing the calculated matching rate for each of the plurality of partial regions with a given threshold value of the partial region. For example, the stopped vehicle detection unit 130 in FIG.
Based on a determination result by the moving vehicle determination unit, an estimation unit (for example, a matching threshold value changing unit 140 in FIG. 2) that estimates a partial region where no vehicle exists among the plurality of partial regions;
Based on the matching rate of the matching processing means for the estimated partial area, threshold update means for updating the threshold of the partial area (for example, the matching threshold changing unit 140 in FIG. 2);
Vehicle presence / absence determination means (for example, the processing unit 100 in FIG. 2) for determining presence / absence of a vehicle in the sensing area based on the determination result of the moving vehicle determination means and the determination result of the stopped vehicle determination means;
Is a vehicle presence / absence determination device (for example, the control device 20 of FIGS. 1 and 2).

In addition, the seventh invention,
A vehicle presence / absence determination method for determining presence / absence of a vehicle in a sensing area in the captured image based on a captured image of a camera viewed from above,
A dividing step of dividing the sensing area into a plurality of partial areas (for example, step A in FIG. 15).
1) and
A moving vehicle determination step (for example, step A7 in FIG. 15) for determining whether or not there is a moving vehicle in the sensing area by comparing and analyzing time-series changes of the captured image;
A matching processing step (for example, step C1 in FIG. 17) for comparing and analyzing the photographed image and a predetermined template image by a predetermined pattern matching processing to calculate a matching rate for each of the plurality of partial regions;
A stopped vehicle determination step for determining whether or not there is a stopped vehicle in the sensing area by comparing the calculated matching rate for each of the plurality of partial areas with a given threshold value of the partial area ( For example, step C3) in FIG.
An estimation step (e.g., step D5 in FIG. 18) for estimating a partial region in which the vehicle does not exist among the plurality of partial regions based on the determination result in the moving vehicle determination step;
A threshold update step (for example, step D7 in FIG. 18) for updating the threshold value of the partial region based on the matching rate of the matching process for the estimated partial region;
A vehicle presence / absence determination step (for example, steps A11 to A27 in FIG. 15) for determining the presence / absence of a vehicle in the sensing area based on the determination result of the moving vehicle determination step and the determination result of the stop vehicle determination step;
Is a vehicle presence / absence determination method.

Further, the eighth invention is
A computer (for example, the control device 20 of FIGS. 1 and 2)
Setting means for setting, as a sensing area, an image portion to be sensed in a photographed image photographed by a camera viewed from above;
A dividing means for dividing the set sensing area into a plurality of partial areas;
Moving vehicle determination means for comparing and analyzing time-series changes of the captured image to determine whether or not there is a moving vehicle in the sensing area;
A matching processing means for comparing and analyzing the captured image and a predetermined template image by a predetermined pattern matching process to calculate a matching rate for each of the plurality of partial regions;
Stopping vehicle determination means for determining whether or not there is a stopped vehicle in the sensing area by comparing the calculated matching rate for each of the plurality of partial areas with a given threshold value of the partial area;
Based on the determination result by the moving vehicle determination means, an estimation means for estimating a partial area where no vehicle is present among the plurality of partial areas;
Threshold updating means for updating the threshold of the partial area based on the matching rate of the matching processing means for the estimated partial area;
Vehicle presence / absence determination means for determining the presence / absence of a vehicle in the sensing area based on the determination result of the moving vehicle determination means and the determination result of the stop vehicle determination means;
For example, the vehicle sensing program 210 in FIG.

  According to any of the first, seventh, and eighth inventions, it is determined whether or not there is a moving vehicle in the sensing area by comparing and analyzing the time-series changes of the images taken by the camera. In addition, the captured image and the predetermined template image are compared and analyzed by a predetermined pattern matching process to determine whether or not there is a stopped vehicle in the sensing area, and based on the determination result of the moving vehicle and the stopped vehicle The presence / absence of a vehicle in the sensing area is determined. The determination of the presence or absence of a stopped vehicle is performed by calculating a matching rate for each of a plurality of partial areas into which the sensing area is divided, and comparing the calculated matching rate with a given threshold value for the partial area. Further, a partial region where no vehicle exists is estimated based on the determination result of whether or not there is a moving vehicle in the sensing region, and the threshold value of the partial region is determined based on the matching rate for the estimated partial region. Is updated. There is a space (so-called inter-vehicle space) where no other vehicle exists in front / rear of the moving vehicle. Therefore, a more accurate vehicle (details) can be obtained by determining a partial region in which the vehicle is assumed to be absent based on the detection result of the moving vehicle and changing the threshold of the matching rate set in the region. The determination of the presence or absence of a stopped vehicle) is realized. Further, since the threshold value is changed for each partial region, it is possible to determine whether or not there is a stopped vehicle with relatively high accuracy even when a fixed template image is used constantly.

A second invention is a vehicle presence / absence determination device according to the first invention,
The vehicle presence / absence determining device that performs the pattern matching process using the image in which the pixel processing blocks having different luminance values are arranged in a grid as the predetermined template image.

  According to the second aspect of the invention, the pattern matching process is performed by regularly using an image in which pixel blocks having different luminance values are arranged in a grid pattern as a template image. Here, the pixel block may be one pixel or may be composed of a plurality of pixels. Further, the lattice shape may be only a vertical lattice, only a horizontal lattice, or a vertical and horizontal lattice. Since the template image is an image in which pixel blocks having different luminance values are arranged in a grid pattern, a constant stop vehicle detection rate can be obtained in the pattern matching process regardless of the weather or day / night time zone. Can be fixed.

A third invention is a vehicle presence / absence determination device according to the first or second invention,
The matching processing unit selects an image in which the luminance difference between the maximum luminance value and the minimum luminance value in the image is 1/10 or less of the luminance difference between the maximum luminance value and the minimum luminance value that can be taken as the luminance value. It is a vehicle presence / absence determination device that performs the pattern matching process by regularly using it as an image.

  According to the third invention, in the pattern matching process, the luminance difference between the maximum luminance value and the minimum luminance value in the image is 1/10 of the luminance difference between the maximum luminance value and the minimum luminance value that can be taken as the luminance value. The following image is regularly used as a template image. Since the brightness difference between the maximum brightness value and the minimum brightness value in the template image is relatively small, the template image can be said to have almost no edge component. For this reason, it becomes possible to further improve the accuracy of the presence / absence determination of the stopped vehicle.

A fourth invention is a vehicle presence / absence determination device of the first to third inventions,
The moving vehicle determining means determines the position and moving speed of the moving vehicle in the captured image;
The estimation means estimates a partial area in which the vehicle does not exist among the plurality of partial areas, using the position and moving speed of the moving vehicle determined by the moving vehicle determination means.
It is a vehicle presence / absence determination device.

  According to the fourth aspect, the position and moving speed of the moving vehicle in the photographed image are determined, and the vehicle in the plurality of partial areas does not exist using the determined position and moving speed of the moving vehicle. A partial region is estimated.

A fifth invention is a vehicle presence / absence determination device according to any one of the first to fourth inventions,
The vehicle presence / absence determining means 1) determines that a vehicle is present in the sensing area when the determination result of the moving vehicle determining means is that there is a moving vehicle, and 2) the determination result of the moving vehicle determining means is a moving vehicle. It is determined that there is no vehicle in the sensing area from when the vehicle changes from being present to no moving vehicle until the stopped vehicle determining means determines that there is no stopped vehicle. 3) In cases other than 1) and 2) Is a vehicle presence / absence determination device that determines that there is no vehicle in the sensing area.

  According to the fifth aspect of the invention, 1) when there is a moving vehicle, it is determined that there is a vehicle in the sensing area, and 2) when there is a change from the presence of the moving vehicle to no moving vehicle, it is determined that there is no stopped vehicle. Until this time, it is determined that there is a vehicle in the sensing area, and in cases other than 1) and 2), it is determined that there is no vehicle in the sensing area.

A sixth invention is a vehicle existence determination device according to any one of the first to fifth inventions,
The camera is provided at a position overlooking the roadway or lane to be sensed from above,
The dividing means divides the sensing area into a plurality of partial areas along a vehicle traffic direction.
It is a vehicle presence / absence determination device.

  According to the sixth aspect of the invention, the camera is provided at a position overlooking the roadway or lane to be sensed from above, and the sensing area is divided into a plurality of partial areas along the vehicle traffic direction.

  According to the present invention, a time series change of a captured image by a camera is compared and analyzed to determine whether or not there is a moving vehicle in the sensing area, and the captured image and a predetermined template image are determined in advance. In the pattern matching process, it is determined whether or not there is a stopped vehicle in the sensing area, and the presence or absence of the vehicle in the sensing area is determined based on the determination result of the moving vehicle and the stopped vehicle. . The determination of the presence or absence of a stopped vehicle is performed by calculating a matching rate for each of a plurality of partial areas into which the sensing area is divided, and comparing the calculated matching rate with a given threshold value for the partial area. Further, a partial region where no vehicle exists is estimated based on the determination result of whether or not there is a moving vehicle in the sensing region, and the threshold value of the partial region is determined based on the matching rate for the estimated partial region. Is updated. There is a space (so-called inter-vehicle space) where no other vehicle exists in front / rear of the moving vehicle. Therefore, a more accurate vehicle (details) can be obtained by determining a partial region in which the vehicle is assumed to be absent based on the detection result of the moving vehicle and changing the threshold of the matching rate set in the region. The determination of the presence or absence of a stopped vehicle) is realized.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. In the following, the case where the present invention is applied to a right-turn applied image-type vehicle detector that senses a right-turn lane of a road will be described. However, applicable embodiments of the present invention are limited to this. is not.

[Constitution]
FIG. 1 is a layout diagram of the image type vehicle detector 1 at an intersection. According to the figure, this image-type vehicle detector 1 is a right-turn applied image-type vehicle detector that senses the presence or absence of a vehicle in a right turn lane at an intersection, and includes a camera 10 and a control device 20. .

  The camera 10 is provided on an upper part of a pole installed near a road so as to take a picture of a right turn lane as a sensing target from above. Images taken by the camera 10 are output to the control device 20 as needed. The control device 20 is installed below the pole and detects the presence or absence of a vehicle on the right turn lane based on the image taken by the camera 10. The detection result by the control device 20 is sent to a signal control device that controls signal display at the intersection, a centralized management center connected via a predetermined communication line, and the like.

  FIG. 2 is a block diagram showing an internal configuration of the control device 20. As shown in the figure, the control device 20 includes a processing unit 100, a storage unit 200, and a communication unit 300.

  The processing unit 100 is realized by, for example, a CPU and performs overall control of the control device 20. In addition, the processing unit 100 includes a moving vehicle detection unit 110, a stopped vehicle detection unit 130, a vehicle tracking unit 120, and a matching threshold change unit 140, and is based on a captured image input from the camera 10 or the like. A vehicle sensing process for determining the presence or absence of a vehicle is performed.

  Specifically, in the vehicle sensing process, first, a sensing area as a vehicle detection target is set in the captured image, and further, the sensing area is divided into a plurality of areas to set divided areas. FIG. 3A shows an example of an image captured by the camera 10. As shown in FIG. 5A, the right turn lane is set as the sensing area AR1 in the captured image. The sensing area AR1 is defined as, for example, an image portion corresponding to a section from the stop line of the right turn lane to the upstream 30m of traffic. This detection area AR1 is basically set according to the operation of the administrator while visually recognizing the image taken by the camera 10 when the image type vehicle sensor 1 is installed, but is automatically set by image recognition processing. Anyway. Further, as shown in FIG. 2B, the sensing area AR1 is divided into a plurality of areas (divided areas) AR2 in a direction substantially perpendicular to the traveling direction of the vehicle in the lane. This division is automatically performed by the processing unit 100, but may be performed according to an operation instruction from the administrator.

  Then, the processing unit 100 detects the presence / absence (presence / absence) of the vehicle in the sensing area AR1. The detection result may be a moving vehicle / stopped vehicle. Specifically, the moving vehicle and the stopped vehicle in the sensing area AR1 are detected in parallel by different detection methods, and the presence / absence of the vehicle in the sensing area is determined based on both detection results. That is, while the moving vehicle is detected, it is determined that the sensing result is “present”. Further, even after the moving vehicle is desired to be detected, the sensing result is determined to be “present” while the stopped vehicle is detected. When both the moving vehicle and the stopped vehicle are no longer detected, it is determined that the sensing result is “none”.

  The image captured by the camera 10 is a moving image, and is captured by the control device 20 as a still image at a predetermined time Δt. The captured image, which is a captured still image, is accumulated and stored as the captured image data group 231 in association with the captured time t.

The moving vehicle detection unit 110 detects the moving vehicle by the “interframe difference method” based on the captured image input from the camera 10. FIG. 4 is a diagram for explaining the principle of detection of a moving object by the interframe difference method. In the interframe difference method, a moving object in an image is detected by calculating a difference between a plurality of temporally continuous images. That is, when a moving body at a certain time t ₀ is detected, a difference image 1 between the image (t) at this time t ₀ and the image (t−1) at the immediately preceding time t ₋₁ is generated. Further, a difference image 2 between the image (t) at time t and the image (t + 1) at time t _{+ 1} immediately after that is generated. Then, to generate a common image of the two difference images 1 and 2 generated (image consisting of common pixels), from the common image, it is possible to detect the moving object at time t _0.

However, since an image at a future time cannot be obtained at the current time, the moving vehicle detection unit 110 detects a moving vehicle based on an image shifted by one frame time. That is, the difference image 1 between the captured image (t) at the current time t ₀ and the captured image (t−1) at the past time t ₋₁ for a predetermined time Δt corresponding to one frame time, and the captured image (t -1) and a difference image 2 between the photographed image (t-2) at the past time t- ₂ for a predetermined time Δt from the time t- ₁ . Next, a common image of the difference images 1 and 2 is calculated. And the moving vehicle in the time t- ₁ is detected from the image part applicable to sensing area AR1 in the calculated common image.

Moreover, the moving vehicle in the said additional area in the time t- ₁ is detected from the image part in the common image corresponding to an additional area. FIG. 5 is a diagram illustrating the additional area AR3. As shown in the figure, the additional area AR3 is a part of the right turn lane in the photographed image, and is an area defined adjacent to the sensing area AR1 in a direction (backward) opposite to the traveling direction of the vehicle. is there. That is, the additional area AR3 can be said to be an area where a vehicle that is about to enter the sensing area AR1 can exist.

The vehicle tracking unit 120 tracks the moving vehicle detected by the moving vehicle detection unit 110. That is, based on the detection result of the moving vehicle in the sensing area AR1 or the additional area AR3 by the moving vehicle detection unit 110, the moving speed and the moving direction are calculated from the change in the position of the moving vehicle at time t, and the next time t Track moving vehicles by predicting position at ₊₁ . Here, the tracking of the moving vehicle is performed only for the moving vehicle existing in the sensing area AR1 or the additional area AR3, and when the vehicle moves out of the sensing area AR1, the tracking is ended.

  Data on the moving vehicle tracked by the vehicle tracking unit 120 is stored in the moving vehicle data 232. FIG. 6 shows an example of the data configuration of moving vehicle data. According to the figure, the moving vehicle data 232 stores the current position 232b, the moving speed 232c, and the moving direction 232d in association with each other for each tracked moving vehicle 232a. The position 232b is expressed by the position of the representative point of the corresponding moving vehicle, for example. The moving direction 232d is expressed by a vector indicating the moving direction in the captured image.

  The stopped vehicle detection unit 130 detects the stopped vehicle by “pattern matching processing” based on the captured image input from the camera 10. Specifically, pattern matching is performed between the captured image and a predetermined template image, and the presence or absence of a stopped vehicle is determined based on the calculated matching rate.

  FIG. 7 shows an example of a template image. The template image is an image of the same size as the captured image, and is a simulated road surface image assuming a road surface when there is no vehicle. In addition, this template image is configured by arranging two types of pixels having different luminance values in a grid pattern. The luminance values of the two types of pixels to be arranged are such that the difference is 1/10 or less of the difference between the maximum value (maximum luminance value) and the minimum value (minimum luminance value) of the luminance values that can be taken in the image. It is stipulated in. That is, in the template image, the difference in luminance value between adjacent pixels is small, and the edge component is extremely small. In the drawing, the maximum luminance value is “255”, the minimum luminance value is “0”, and the template image is composed of pixels with luminance values “10” and “0”. Therefore, the matching rate between the captured image and the template image is high when no vehicle is present in the captured image, and conversely, is low when the vehicle is present.

  The template image may be an image having almost no edge component. For example, the template image may be an image in which the luminance of adjacent pixels changes so as to draw a gradation. Moreover, as shown to Fig.8 (a), the image by which the pixel block comprised by several pixels with equal luminance value was arrange | positioned at the grid | lattice form may be sufficient, as shown to the same figure (b), It may be an image in which pixels having the same luminance value are arranged side by side in the column / row direction. Data about the template image is stored in the template image data 221.

  FIG. 9 is a diagram illustrating an outline of stopped vehicle detection by pattern matching. As shown in the figure, for each divided area AR2 of the photographed image, pattern matching with the corresponding image portion of the template image is performed to calculate the matching rate R. Then, the matching rate R calculated for each of the divided areas AR2 is compared with a predetermined threshold value, and if the number of areas where the matching rate exceeds the threshold value R is equal to or greater than the predetermined number N, the stop vehicle is determined to be “present”, otherwise Then, it is determined that there are no stopped vehicles.

式（１）において、「ｆ［ｉ，ｊ］」は撮影画像の各画素の輝度値、「ｓ［ｉ，ｊ］」はテンプレート画像の各画素の輝度値、「Ｍ，Ｎ」は撮影画像及びテンプレート画像の画像サイズ（列／行方向の画素の数）である。また、「ｆ_ａ」は撮影画像の各画素の輝度値の平均値、「Ｓ_ａ」はテンプレート画像の各画素の輝度値の平均値である。このマッチング率Ｒの算出式（１）は、マッチング率算出式データ２２３として記憶されている。 The matching rate R is given by equation (1).

In Expression (1), “f [i, j]” is the luminance value of each pixel of the captured image, “s [i, j]” is the luminance value of each pixel of the template image, and “M, N” is the captured image. And the image size (number of pixels in the column / row direction) of the template image. “F _a ” is the average value of the luminance values of the pixels of the captured image, and “S _a ” is the average value of the luminance values of the pixels of the template image. The calculation formula (1) for the matching rate R is stored as matching rate calculation formula data 223.

  Here, the calculated matching rate R of each divided region is stored in the matching result data 233. FIG. 10 shows an example of the data configuration of the matching result data 233. According to the figure, the matching result data 233 stores the calculated matching rate 233b in association with each divided area 233a constituting the sensing area.

  Further, the matching rate threshold value is stored in the matching threshold value data 222. FIG. 11 shows an example of the data configuration of the matching threshold value data 222. According to the figure, the matching threshold value data 222 stores a matching rate threshold value 222b in association with each of the divided areas 222a constituting the sensing area.

  The detection result of the moving vehicle by the moving vehicle detection unit 110 and the detection result of the stopped vehicle by the stop vehicle detection unit 130 are stored in the vehicle detection result data 234. FIG. 12 shows an example of the data configuration of the vehicle detection result data 234. According to the figure, the vehicle detection result data 234 stores a detection result 234b in association with each type of vehicle 234a to be detected. The vehicle type 234a is a moving vehicle and a stopped vehicle. The detection result 234b stores a value indicating whether or not the corresponding vehicle is detected as existing in the sensing area AR1.

  The matching threshold value changing unit 140 changes the “threshold value” of the matching rate used in the stopped vehicle detection unit 130. Specifically, based on the moving speed of the moving vehicle detected by the moving vehicle detection unit 110, a divided area AR2 where it is estimated that no other vehicle exists is determined, and this divided area AR2 (the estimated area without vehicle) ) Is changed.

  There is a region (so-called inter-vehicle space) where no other vehicle exists in front / back of the vehicle traveling on the road. This inter-vehicle distance is longer as the traveling speed is faster.

In traffic engineering, the distance L at which a traveling vehicle can be safely stopped is given by the following equation (2).
L = τ · V + V ² / 2d (2)
In equation (2), “V” is the vehicle running speed when the signal display is switched, “d” is the average deceleration from when the brake starts to stop until it stops, and “τ” is the driver reaction time (the brake is effective). Time to start). FIG. 13 shows a graph of the formula (2). In the figure, the horizontal axis is the distance L and the vertical axis is the traveling speed V. However, the deceleration d is 0.7 [s] and the reaction time τ is 3.0 [m / s ² ]. The calculation formula (2) of the distance L is stored as the vehicle-less estimated distance calculation formula data 224.

  In the present embodiment, the distance L is regarded as the distance ahead of the moving vehicle (the vehicle-less estimated region) where it is estimated that no other vehicle exists, and the vehicle-free estimated region is determined. FIG. 14 is a diagram for explaining determination of the no-vehicles estimation region. As shown in the figure, for the moving vehicle C in the sensing area AR1 or the additional area AR3, the estimated distance L without vehicle is calculated according to the equation (2) based on the moving speed V. Subsequently, the divided area AR2 that is entirely included in the range from the position of the moving vehicle C to the preceding vehicle-less estimated distance L along the traveling direction is determined as the vehicle-free estimated area. Here, the moving vehicles in the sensing area AR1 or the additional area AR3 are tracked by the moving vehicle tracking unit 120, and the moving speed V and the current position are calculated. In addition, when a plurality of moving vehicles C are being tracked, an estimated no-vehicle area is determined for each. Then, for each of the estimated no-vehicle areas, the threshold of the matching rate set in the area is set to a predetermined value α (for example, 0.05 (5 %)) Is added to the added value.

  The storage unit 200 stores a system program for the processing unit 100 to control the control device 20 in an integrated manner, a program and data for realizing vehicle detection control of the present embodiment, and the operation of the storage unit 200. It is used as an area and temporarily stores calculation results, flag data, and the like executed by the processing unit 100 according to various programs. The storage unit 200 is realized by, for example, an IC memory, a hard disk, a ROM, a RAM, or the like. In the present embodiment, the storage unit 200 stores a vehicle sensing program 210 as a program, and as data, template image data 221, matching threshold data 222, matching rate calculation formula data 223, and vehicle-less estimated distance calculation formula. Data 224, captured image data group 231, moving vehicle data 232, matching result data 233, and vehicle detection result data 234 are stored.

[Process flow]
FIG. 15 is a flowchart for explaining the flow of the vehicle sensing process. This process is realized by the processing unit 100 executing the vehicle detection program 210. According to the figure, the processing unit 100 first sets a sensing area AR1 and an additional area AR3 in a captured image, and further divides the sensing area AR1 into a plurality of areas to set divided areas (step A1). ). In addition, the moving vehicle detection unit 110 starts the moving vehicle detection process (step A3), the stopped vehicle detection unit 130 starts the stopped vehicle detection process (step A5), and the vehicle tracking unit 120 starts the vehicle tracking process ( In step A7), the matching threshold value changing unit 140 starts the matching threshold value changing process (step A9).

  Subsequently, the processing unit 100 determines whether or not a moving vehicle is detected by the moving vehicle detection unit 110. If the moving vehicle is detected (step A11: YES), the vehicle detection result is “present” (step A13). ).

  On the other hand, if a moving vehicle is not detected (step A11: NO), it is determined whether or not the detection result of the moving vehicle has changed from “present” to “absent”. If the detection result of the moving vehicle changes from “present” to “absent” (step A15: YES), the flag F is set to “1” (step A17). This flag F is a flag indicating that the determination result of the stopped vehicle is the vehicle detection result, and is set to “1” at the timing when the detection result of the moving vehicle changes from “present” to “none”. When the detection result of the stopped vehicle becomes “none”, it is set (cleared) to “0”.

  Subsequently, it is determined whether or not a stopped vehicle is detected by the stopped vehicle detection unit 130. If it is detected (step A21: YES), the vehicle detection result is set to “present” (step A23). On the other hand, if a stopped vehicle is not detected (step A21: NO), the flag F is set to “0” (step A25), and the vehicle detection result is set to “none” (step A27).

  If the detection result of the moving vehicle remains “None” (step A15: NO), the flag F is determined. If the flag F is “1” (step A19: YES), then the stopped vehicle detection unit It is determined by 130 whether or not a stopped vehicle is detected. If a stopped vehicle is detected (step A21: YES), the vehicle detection result is “present” (step A23). If a stopped vehicle is not detected (step A21: NO), the flag F is set to “0”. "(Step A25) and the vehicle detection result is" none "(step A27). If the flag F is “0” (step A19: NO), the vehicle detection result is “none” (step A27).

  Thereafter, the processing unit 100 determines whether or not to end the vehicle sensing process. If the processing does not end (step A29: NO), the processing unit 100 returns to step A11. If the processing ends (step A: YES), the processing unit 100 is executing. After each process is finished (step A31), the vehicle sensing process is finished.

FIG. 16 is a flowchart of the moving vehicle detection process. According to the figure, the moving vehicle detection unit 110 performs the process of the loop A every predetermined time Δt. That is, in the loop A, to calculate the captured image at the current time _{t 0} (t), a difference image 1 between the photographed image from the current time _{t 0} the time Δt only past time _{t -1 (t-1)} (Step B1). Further, a difference image 2 between the photographed image (t−1) and the photographed image (t−2) at the time t − ₂ past from the time t ₋₁ by the time Δt is calculated (step B3). And the common image of the difference images 1 and 2 is produced | generated, and the presence or absence of the moving vehicle in sensing area AR1 in a picked-up image is judged based on this common image (step B5). Loop A is performed in this way. Then, when an instruction to end the process is given, the moving vehicle detection process ends.

FIG. 17 is a flowchart of the stopped vehicle detection process. According to the figure, the stopped vehicle detection unit 130 performs the process of the loop B every predetermined time Δt. That is, in the loop B, for the captured image (t) at the current time t ₀ , pattern matching with the corresponding image portion of the template image is performed for each divided region constituting the sensing region to calculate the matching rate R ( Step C1) Next, for each divided region, the calculated matching rate R is compared with a predetermined threshold, and the presence or absence of a stopped vehicle in the sensing region AR1 in the captured image is determined according to the comparison result (Step C3). ). Loop B is performed in this way. Then, when an instruction to end the process is given, the stopped vehicle detection process ends.

  FIG. 18 is a flowchart of the matching threshold value changing process. According to the figure, the matching threshold value changing unit 140 performs the process of the loop C every predetermined time Δt. That is, in the loop C, it is determined whether there is a moving vehicle tracked by the vehicle tracking unit 120. If there is a moving vehicle being tracked (step D1: YES), a no-vehicle estimated distance L is calculated based on the moving speed V of the moving vehicle (step D3). Next, a divided region included in the range up to the calculated distance L is determined and set as a vehicle-free estimated region (step D5). Then, the threshold value is changed based on the matching rate R of each of the determined vehicle-less estimated regions (step D7). Loop C is performed in this way. Then, when an instruction to end the process is given, the matching threshold value changing process ends.

[Action / Effect]
As described above, according to the present embodiment, the control device 20 of the image type vehicle sensor 1 1) calculates the matching rate R by performing pattern matching between an image captured by the camera 10 and a predetermined template image. Detection of a stopped vehicle by comparison with a threshold value and 2) detection of a moving vehicle using an inter-frame difference method are performed, and the presence or absence of the vehicle is determined based on both detection results. Further, based on the moving speed V of the moving vehicle, a divided region (a vehicle-free estimated region) where it is estimated that there is no other vehicle ahead of the moving vehicle is determined, and a matching rate R of the vehicle-free estimated region is determined. Is changed to a value corresponding to the matching rate R of the area calculated when the stopped vehicle is detected. As a result, it is possible to realize the presence / absence of a highly accurate vehicle by a relatively simple calculation. In addition, since the threshold value of the matching rate is changed for each partial area AR2, the accuracy of determining whether or not there is a stopped vehicle is relatively good even if the template image is fixed. Furthermore, since a template image is an image in which pixels having a small luminance difference are arranged in a grid and has almost no edge component, the system for detecting a stopped vehicle can be enhanced.

  In the above-described embodiment, the presence / absence of the vehicle in the right turn lane is determined, but it goes without saying that other lanes such as a straight lane and a right turn lane may be used as a lane between intersections.

The example of arrangement | positioning of the image type vehicle sensor in an intersection. The internal block diagram of a control apparatus. An example of a photographed image. Explanatory drawing of the difference method between frames. Explanatory drawing of an additional area | region. The data structural example of moving vehicle data. An example of a template image. Another example of a template image. Explanatory drawing of pattern matching. The data structural example of matching result data. The data structural example of matching threshold value data. The data structural example of vehicle detection result data. The graph which shows the relationship between the moving speed V and the estimated distance L between vehicles. Explanatory drawing of judgment of an estimation area without a vehicle. The flowchart of a vehicle detection process. The flowchart of a moving vehicle detection process. The flowchart of a stop vehicle detection process. The flowchart of a matching threshold value change process.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image type vehicle sensor 10 Camera 20 Control apparatus 100 Processing part 110 Moving vehicle detection part, 120 Vehicle tracking part 130 Stop vehicle detection part, 140 Matching threshold change part 200 Storage part 210 Vehicle detection program 221 Template image data, 222 Matching threshold Data 223 Matching rate calculation formula data 224 Vehicle-free estimated distance calculation formula data 231 Captured image data group 232 Moving vehicle data 233 Matching result data 234 Vehicle detection result data AR1 Sensing area, AR2 divided area, AR3 additional area

Claims (8)

A setting means for setting, as a sensing area, an image part to be sensed in a photographed image photographed by a camera viewed from above;
A dividing means for dividing the set sensing area into a plurality of partial areas;
Moving vehicle determination means for comparing and analyzing time-series changes of the captured image to determine whether or not there is a moving vehicle in the sensing area;
A matching processing means for comparing and analyzing the captured image and a predetermined template image by a predetermined pattern matching process to calculate a matching rate for each of the plurality of partial regions;
Stopping vehicle determination means for determining whether or not there is a stopped vehicle in the sensing area by comparing the calculated matching rate for each of the plurality of partial areas with a given threshold value of the partial area; ,
Based on a determination result by the moving vehicle determination unit, an estimation unit that estimates a partial region in which the vehicle does not exist among the plurality of partial regions;
Threshold updating means for updating the threshold of the partial area based on the matching rate of the matching processing means for the estimated partial area;
Vehicle presence / absence determination means for determining the presence / absence of a vehicle in the sensing area based on the determination result of the moving vehicle determination means and the determination result of the stopped vehicle determination means;
A vehicle presence / absence determination device.   The vehicle presence / absence determining apparatus according to claim 1, wherein the matching processing unit performs the pattern matching processing by regularly using an image in which pixel blocks having different luminance values are arranged in a grid as the predetermined template image.   The matching processing unit selects an image in which the luminance difference between the maximum luminance value and the minimum luminance value in the image is 1/10 or less of the luminance difference between the maximum luminance value and the minimum luminance value that can be taken as the luminance value. The vehicle presence / absence determining apparatus according to claim 1, wherein the pattern matching process is performed using the image as a regular image. The moving vehicle determining means determines the position and moving speed of the moving vehicle in the captured image;
The estimation means estimates a partial area in which the vehicle does not exist among the plurality of partial areas, using the position and moving speed of the moving vehicle determined by the moving vehicle determination means.
The vehicle presence determination apparatus as described in any one of Claims 1-3.   The vehicle presence / absence determining means 1) determines that a vehicle is present in the sensing area when the determination result of the moving vehicle determining means is that there is a moving vehicle, and 2) the determination result of the moving vehicle determining means is a moving vehicle. It is determined that there is no vehicle in the sensing area from when the vehicle changes from being present to no moving vehicle until the stopped vehicle determining means determines that there is no stopped vehicle. 3) In cases other than 1) and 2) The vehicle presence / absence determination apparatus according to claim 1, wherein it is determined that no vehicle exists in the sensing area. The camera is provided at a position overlooking the roadway or lane to be sensed from above,
The dividing means divides the sensing area into a plurality of partial areas along a vehicle traffic direction.
The vehicle presence determination apparatus as described in any one of Claims 1-5. A vehicle presence / absence determination method for determining presence / absence of a vehicle in a sensing area in the captured image based on a captured image of a camera viewed from above,
Dividing the sensing region into a plurality of partial regions;
A moving vehicle determination step of comparing and analyzing time-series changes of the captured image to determine whether or not there is a moving vehicle in the sensing area;
A matching process step of comparing and analyzing the captured image and a predetermined template image by a predetermined pattern matching process to calculate a matching rate for each of the plurality of partial regions;
A stopped vehicle determination step of determining whether there is a stopped vehicle in the sensing region by comparing the calculated matching rate for each of the plurality of partial regions with a given threshold value of the partial region; ,
Based on the determination result of the moving vehicle determination step, an estimation step of estimating a partial region where no vehicle exists among the plurality of partial regions;
A threshold update step of updating the threshold of the partial area based on the matching rate of the matching process for the estimated partial area;
A vehicle presence / absence determination step for determining the presence / absence of a vehicle in the sensing area based on the determination result of the moving vehicle determination step and the determination result of the stop vehicle determination step;
A vehicle presence / absence determination method including: Computer
Setting means for setting, as a sensing area, an image portion to be sensed in a photographed image photographed by a camera viewed from above;
A dividing means for dividing the set sensing area into a plurality of partial areas;
Moving vehicle determination means for comparing and analyzing time-series changes of the captured image to determine whether or not there is a moving vehicle in the sensing area;
A matching processing means for comparing and analyzing the captured image and a predetermined template image by a predetermined pattern matching process to calculate a matching rate for each of the plurality of partial regions;
Stopping vehicle determination means for determining whether or not there is a stopped vehicle in the sensing area by comparing the calculated matching rate for each of the plurality of partial areas with a given threshold value of the partial area;
Based on the determination result by the moving vehicle determination means, an estimation means for estimating a partial area where no vehicle is present among the plurality of partial areas;
Threshold updating means for updating the threshold of the partial area based on the matching rate of the matching processing means for the estimated partial area;
Vehicle presence / absence determination means for determining the presence / absence of a vehicle in the sensing area based on the determination result of the moving vehicle determination means and the determination result of the stop vehicle determination means;
Program to function as.

Images