JP2008052517A - Traffic measurement method and traffic measurement apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately count traffic even when vehicles travelling in parallel are imaged in an overlapped manner or when a vehicle has an illuminator similar to a headlight at night, in a traffic measurement apparatus for measuring traffic by analyzing an image obtained by imaging a vehicle travelling on a road by a camera. <P>SOLUTION: A processing part 12 detects a position (base point) of a specific part (a front end part of the vehicle in daytime, and a pair of headlights at night) of a vehicle in an image picked by a camera 11. A three-dimensional shape is calculated which is obtained by viewing from the camera 11 a prescribed standard vehicle disposed along the road so that the specific part is located at a position corresponding to the base point in a real world, and superimposed on the image. It is determined that a vehicle exists in the region surrounded by the three-dimensional shape on the image, and it is judged whether another vehicle exists outside the three-dimensional shape. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a traffic volume measuring method and a traffic volume measuring apparatus for analyzing traffic images obtained by photographing a vehicle traveling on a road with a camera.

  In the measuring device that measures the traffic volume by taking a picture of the road situation with a camera and analyzing the image, and judging the vehicle type and traveling speed of the passing vehicle, a support frame is bridged across the road If the camera is installed, the equipment becomes large. Therefore, it is often done to install the camera on a column that stands on the side of the road and photograph the vehicle from diagonally forward.

  However, if the vehicle is photographed obliquely in this way, as shown in FIG. 2, the vehicles A and B running side by side in the adjacent lane are overlapped and photographed, so there is a possibility that a plurality of parallel running vehicles are mistaken as one. There is. Therefore, for example, a method is used in which an image is extracted for each lane region, and a vehicle is detected from the image for each lane to separate the overlap of vehicles traveling in adjacent lanes and analyze the image.

  On the other hand, since the outline of the vehicle becomes unclear at night, image analysis is performed by extracting a high luminance part such as a headlight. Usually, since the headlight has a feature such as a pair on the left and right, using such a feature, a pair of headlights are extracted and distinguished from other lamps, and the extracted headlight pair is used as a reference. The traffic volume and the type of vehicle are determined (for example, see Patent Document 1).

JP 11-353580 A

  In the method of extracting and analyzing images for each lane region, it is possible to separate and recognize vehicles running in parallel in adjacent lanes. On the other hand, in large vehicles, the roof portion enters the adjacent lane. There is a risk of misidentifying two large vehicles as two. In addition, a vehicle that travels between lanes cannot be detected correctly. Furthermore, even when a camera is installed so that the vehicle is photographed from the front, the vehicles before and after traveling with a close distance between the vehicles may overlap and be photographed. Even if extracted, the front and rear vehicles cannot be separated.

  In addition, when detecting a vehicle at night based on a pair of headlights, the headlight pair and other light emitters are separated from each other by the characteristics of the headlights (pairing and the distance between the paired light emitters is about the vehicle width. In such a method, there is a limit to the accuracy of the determination, and there is a possibility that a light emitter other than the headlight or a reflection on the road surface is mistaken as a headlight pair. For example, many large trucks are often equipped with various lamps on the top and side, and if one of these light emitters is mistaken for a headlight pair, one truck is counted as multiple vehicles. As a result, there was a problem that the correct traffic volume could not be grasped.

  The present invention is intended to solve the above-mentioned problem, and is a case where a parallel running vehicle and a front and rear vehicle are photographed with overlapping, or a nighttime vehicle has a illuminant confusing with a headlight. Another object of the present invention is to provide a traffic volume measuring method and a traffic volume measuring apparatus capable of accurately measuring the traffic volume of a vehicle traveling on a road.

  The gist of the present invention for achieving the object lies in the inventions of the following items.

[1] In a traffic volume measuring method for measuring traffic volume by analyzing an image obtained by photographing a vehicle traveling on a road with a camera,
A detection step of detecting a position of the vehicle on an image obtained by photographing with the camera;
A superimposing step of obtaining a three-dimensional shape when a predetermined standard vehicle arranged on the road is viewed from the camera, and superimposing the three-dimensional shape on the image in accordance with the position of the detected vehicle;
And a determination step of determining that one vehicle is present in the range covered with the three-dimensional shape on the image.

  In the above invention, the position of the vehicle is detected in the image photographed by the camera, the three-dimensional shape when the predetermined standard vehicle passing through the road is viewed from the camera is obtained, and the three-dimensional shape is detected first. The image is aligned based on the position of the image and superimposed on the image. Then, it is determined that one vehicle exists in the range covered with the three-dimensional shape on the image.

  Thereby, even if vehicles running in parallel overlap each other, for example, the position of one vehicle in front is detected, and one vehicle in front on the image is superimposed on this position by superimposing the three-dimensional shape of the standard vehicle. The area occupied by the vehicle can be grasped, and two overlapping parallel vehicles can be recognized separately for each vehicle. Further, at night, for example, the position of the vehicle is detected based on the position where the headlight exists, and the three-dimensional shape of the standard vehicle viewed from the camera is superimposed on this position. As a result, even at night when the outline of the vehicle is unclear, the area occupied by one vehicle can be recognized on the image, and even if other high-intensity parts that are confusing are present in this area, they are integrated into 1 It can be determined as a single vehicle.

  The standard vehicle is a model of a standard three-dimensional shape of a vehicle traveling on a road. For example, a model obtained by modeling a vehicle in a rectangular parallelepiped is used. The three-dimensional shape of the standard vehicle may be calculated according to the position of the vehicle detected on the image, or a fixed three-dimensional shape that is calculated and stored in advance regardless of the detected position of the vehicle. Can be used. In addition, a plurality of representative vehicle positions on the image are set, and a three-dimensional shape corresponding to each position is calculated and stored in advance, and 3 corresponding to the detected vehicle position is selected from these. A dimensional shape may be selected and overlapped.

[2] In the superimposing step, calculating a three-dimensional shape when the standard vehicle arranged at a position in the real world corresponding to the position of the vehicle detected on the image is viewed from the camera. The traffic volume measuring method according to [1], which is characterized.

  In the above-described invention, even if the vehicle appears large in front of the image or the vehicle is relatively small in the back, the three-dimensional shape is calculated according to the position of the vehicle. The area occupied by the vehicle on the image can be specified with a small error.

[3] The traffic according to [1] or [2], wherein it is determined whether or not another vehicle exists in a region outside the range covered with the three-dimensional shape on the image. Quantity measurement method.

  In the above invention, since the range covered with the three-dimensional shape is the range where one vehicle exists, it is determined whether or not another vehicle exists outside the range. That is, the second detection of the parallel running vehicle is performed.

[4] As the standard vehicle, a plurality of types of standard vehicles having different vehicle types such as a large vehicle and a small vehicle are set in advance.
A vehicle type determination step for determining the vehicle type of the vehicle whose position is detected in the detection step;
The traffic volume measuring method according to any one of [1] to [3], wherein a three-dimensional shape of a standard vehicle corresponding to a vehicle type indicated by a determination result of the vehicle type determination step is superimposed on the image.

  In the said invention, a vehicle type is determined, the three-dimensional shape of the standard vehicle according to the vehicle type is calculated | required, and it superimposes on an image. The vehicle type can be determined by, for example, the vehicle width or the distance between the pair of headlights. As described above, the type of vehicle such as a large vehicle or a small vehicle is determined, and the three-dimensional shape of the standard vehicle corresponding to the vehicle type is superimposed on the image, so that the region where one vehicle exists can be specified with higher accuracy. Note that the vehicle type is not limited to a large vehicle or a small vehicle, and may be set as necessary.

[5] The traffic volume according to any one of [1] to [4], wherein in the detection step, the position of the vehicle is detected by detecting a base point where a specific location of the vehicle exists on the image. Measurement method.

  In the said invention, the position of a vehicle is detected on the basis of the specific location of a vehicle. The specific part may be a part that can identify the position of the vehicle on the image, such as the front end of the vehicle or a pair of headlights.

[6] The specific location is the front end of the vehicle,
The traffic measurement method according to [5], wherein, in the detecting step, a front end portion of a vehicle existing closest to a traveling direction of the vehicle in the image is detected as the base point.

  In the above invention, even if a plurality of vehicles are overlapped, the outline of the front end of the vehicle located on the most vehicle traveling direction side (lowermost in the image) in the image can be reliably recognized. Is preferred.

[7] The specific portion is a pair of headlights,
In the detection step, a pair of high-intensity portions having characteristics corresponding to a pair of headlights are detected as a pair of headlight candidates, and the vehicle traveling direction is the most in the image among the detected candidates. A traffic volume measuring method according to [5], wherein a pair of headlight candidates existing on the side is detected as the base point.

  In the above invention, in a state where the vehicle is photographed from the front, the headlight is always present in the traveling direction side of the vehicle in the image rather than the light on the roof or side of the large vehicle. It is possible to correctly recognize the position of the vehicle and apply a three-dimensional shape.

[8] A day / night discrimination step for discriminating between day and night,
If the discrimination result of the day / night discrimination step is noon,
The specific location is the front end of the vehicle, and in the detection step, the front end of the vehicle existing closest to the vehicle traveling direction in the image is detected as the base point,
When the discrimination result of the day / night discrimination step is night,
The specific portion is a pair of headlights, and in the detection step, a pair of high-intensity parts having characteristics corresponding to the pair of headlights is detected as a pair of headlight candidates, and the detected candidate The traffic volume measuring method according to [5], wherein a pair of headlight candidates that are closest to the vehicle traveling direction in the image is detected as the base point.

  In the above invention, day and night are discriminated, and since the outline of the vehicle can be recognized at daytime, the base point is detected with reference to the front end portion of the vehicle, and at night, the base point is detected with reference to a pair of headlights. The determination of day and night can be made based on, for example, the ambient illuminance at the time of shooting or whether the average brightness of the image is equal to or higher than a predetermined threshold value.

[9] In a traffic measurement device that measures traffic by analyzing an image obtained by photographing a vehicle traveling on a road with a camera,
A traffic volume measuring apparatus comprising: a processing unit that executes the traffic volume measuring method according to any one of [1] to [8] with respect to an image captured by the camera.

  According to the traffic measuring method and the traffic measuring device according to the present invention, the position of the vehicle is detected on the image obtained by photographing with the camera, and the predetermined standard vehicle arranged on the road is seen from the camera. The three-dimensional shape of the case is obtained, and this three-dimensional shape is superimposed on the image according to the position of the vehicle detected earlier, and it is determined that one vehicle exists in the range covered with the three-dimensional shape. Therefore, even when multiple vehicles are photographed in an overlapping manner or when a nighttime vehicle has a illuminant confusing with headlights, the area occupied by the vehicle on the image is accurately recognized one by one. Traffic volume can be measured correctly.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram showing a schematic configuration of a traffic volume measuring apparatus 10 according to the present invention. The traffic volume measuring device 10 is a device that measures traffic volume by analyzing an image obtained by shooting a road condition with the camera 11, and includes a processing unit 12 that processes an image shot with the camera 11. I have. The camera 11 is fixed to a pillar or the like standing on the side of the road, and is installed so as to capture an image illustrated in FIG.

  The processing unit 12 includes an image acquisition unit 13, an image memory 14, a preprocessing unit 15, a vehicle position / vehicle type determination unit 16, a coordinate conversion unit 18, a vehicle region separation / integration unit 19, and a vehicle tracking unit 21. The traffic volume measuring unit 22 is provided.

  The camera 11 captures an image of several tens of frames per second, and the image acquisition unit 13 captures the image data from the camera 11. The image memory 14 temporarily stores the image data captured from the camera 11 by the image acquisition unit 13. The preprocessing unit 15 extracts a moving object from the image stored in the image memory 14 as a vehicle area when it is determined that it is daytime and a function that determines day and night based on ambient brightness at the time of shooting. Further, when it is determined that it is night, the function of extracting a high luminance part having a certain brightness or more from the image stored in the image memory 14 is achieved. The ambient brightness may be detected by providing a separate sensor, or based on the average brightness of an image captured by the camera 11 or the brightness of a predetermined specific area (for example, a part of the roadside). You may judge. In addition, a method such as frame difference or background difference is used for extraction of the moving object.

  The vehicle position / vehicle type determination unit 16 detects the position of the vehicle by detecting the position where the specific part of the vehicle is present as a base point on the image after the vehicle region or the high luminance part is extracted by the preprocessing unit 15. A function and a function of determining a vehicle type of the vehicle from which the position (base point) is detected. Here, when it is determined by the preprocessing unit 15 to be daytime, the front end portion of the vehicle is used as a specific location, and when it is determined to be night, a pair of headlights is used as the specific location. The vehicle width and the vehicle length are used to determine the vehicle type.

  The coordinate conversion unit 18 is located at a specific location (front end of the vehicle or a pair of heads) at a position in the real world (actual world taken by the camera 11) corresponding to the base point detected by the vehicle position / vehicle type determination unit 16. A predetermined standard vehicle is arranged along the road so that the light is present, and the three-dimensional shape of the standard vehicle when viewed from the camera 11 is calculated.

  Several types of standard vehicles are set according to the size of the vehicle. Here, two types, a large vehicle and a small vehicle, are used. In addition, a standard vehicle in which a vehicle is modeled in a rectangular parallelepiped is used. For example, the size of the standard vehicle for large vehicles may be determined according to the maximum vehicle width defined by law. The coordinate conversion unit 18 calculates the three-dimensional shape of the standard vehicle of the vehicle type according to the vehicle type determination result in the vehicle position / vehicle type determination unit 16. For the calculation, posture information and angle of view information of the camera 11 set in advance are used.

  The vehicle area separation / integration unit 19 superimposes the three-dimensional shape calculated by the coordinate conversion unit 18 on the image used for detecting the base point, and determines that one vehicle exists in the range covered with the three-dimensional shape. To do. In addition, after the three-dimensional shape is superimposed on the image as described above, the vehicle position / vehicle type determination unit 16, the coordinate conversion unit 18, and the vehicle region separation / integration unit 19 exclude the range covered with the three-dimensional shape. It is further determined whether or not there is another vehicle for the remaining area (outer area).

  The vehicle tracking unit 21 performs a function of recognizing and tracking a moving vehicle, and the traffic volume measuring unit 22 passes through the road based on the determination result of the vehicle region separation and integration unit 19 and the tracking status of the vehicle tracking unit 21. The number of vehicles to be counted is counted and the traffic volume per unit time is measured.

  Next, a mutual conversion formula between coordinates (camera coordinates) on an image photographed by the camera 11 and coordinates in the real world (world coordinates) will be described.

FIG. 3 shows the relationship between camera coordinates and world coordinates. In this figure, the position, posture, and characteristics of the camera are defined as follows.
Lens focal length: λ
Camera head position: W ₀ = (X ₀ , Y ₀ , Z ₀ )
Vector from camera head to imaging plane: r = (r ₁ , r ₂ , r ₃ )
The rotation of the camera posture around the X, Y, and Z axes is α, β, and θ, respectively.
At this time, to which point in the camera coordinate system (x, y) the position (X, Y, Z) in the world coordinate system is projected can be obtained by the following equation (1).

  Conversely, the position of the point (x, y) in the camera coordinate system in the world coordinate system (X, Y, Z) is shown below by setting Z in the world coordinate system as a constant ( 2) It can obtain | require by Formula.

  FIG. 4 shows the flow of the entire process performed by the traffic measuring device 10. First, an image is captured by the camera 11 (step S101). Next, it is determined whether or not the ambient illuminance at the time of shooting is greater than or equal to a predetermined threshold value (step S102). When it is determined that it is daytime when the brightness is equal to or higher than the threshold (step S102; high), the feature amount of the entire vehicle can be taken (recognizing the contour of the vehicle). A vehicle separation process for separating overlapping vehicles such as running vehicles is performed (step S103). On the other hand, when it is determined that the brightness is not equal to or higher than the threshold and the night (step S102; low), a high luminance part such as a headlight or a lamp on the side of the vehicle body is extracted, and the extracted high luminance part is A light integration process for integrating the units is performed (step S104). Details of the vehicle separation process or the light integration process will be described later.

  After recognizing the area where each vehicle exists in the image by the vehicle separation process or the light integration process, a vehicle tracking process (step S105) and a traffic volume measurement process (step S106) are performed. By repeatedly performing such processing, the traffic volume of the vehicle passing through the road and the vehicle type of the passing vehicle are measured.

  FIG. 5 shows the flow of the vehicle separation process performed in S103 of FIG. 6 and 7 show images of the process of the vehicle separation process. In this example, it is assumed that an image 30 shown in FIG. 2 is taken by the camera 11, and a process of separating two vehicles A and B that are taken in an overlapping manner on this screen will be described. The preprocessing unit 15 extracts the moving object from the image 30 captured from the camera 11 (step S201 in FIG. 5). FIG. 6A shows an extracted image 31 obtained by extracting a moving body (vehicle region 32) from the image 30 of FIG.

  Next, as shown in FIG. 6 (b), the vehicle position / vehicle type determination unit 16 is arranged in the vertical direction (y direction) with respect to the lower part of the extracted vehicle region 32 (the lower part of the broken line L in the figure). By performing the projection process (step S202), the distribution 35 shown in FIG. 6C is acquired. The distribution 35 indicates the number of pixels (projection pixel number) of the vehicle region 32 existing in a portion below the broken line L at each coordinate position in the horizontal direction (x direction) of the image. In FIG. 6C, the operation of obtaining the added value of the number of projected pixels in the range W corresponding to the vehicle width (area of the hatched portion in the figure) is performed by moving the position of the vehicle width W in the horizontal direction (x direction). However, it calculates | requires each (step S203, S204), and calculates | requires the position of the range W when the addition value of the number of projection pixels becomes the maximum as a position where the front-end part of a vehicle exists (step S205). In the example shown in FIG. 6C, the position where the right end of the front end portion of the vehicle exists is detected as the base point K.

  Further, the vehicle type is determined from the size (vehicle width, vehicle length, etc.) of the vehicle area 32 in the vicinity of the detected vehicle location (step S206). This determination is made in several stages (here, two stages, a small car and a large car).

  Next, the coordinate conversion unit 18 arranges the standard vehicle corresponding to the vehicle type indicated by the determination result in the vehicle position / vehicle type determination unit 16 so that the specific part is located at the position in the real world corresponding to the base point, and the camera The three-dimensional shape (three-dimensional rectangular parallelepiped model) of the standard vehicle viewed from 11 is calculated (step S207). Here, vehicle width, vehicle length, and vehicle height data in the real world are determined and stored in advance for each vehicle type of the standard vehicle, and data of the standard vehicle corresponding to the vehicle type of the determination result is used for calculation.

  As shown in FIG. 7A, the vehicle region separation and integration unit 19 superimposes the three-dimensional shape 33 calculated by the coordinate conversion unit 18 on the image used for detecting the base point K, and uses the three-dimensional shape 33. It is determined that one vehicle exists in the covered range, and the outside of the range covered with the three-dimensional shape 33 is determined as a range in which another vehicle can exist (step S208). Then, it is determined whether or not an extraction portion that is considered to be a vehicle exists outside the range covered with the three-dimensional shape 33 (step S209). Here, as shown in FIG. 7B, the range covered with the three-dimensional shape 33 is deleted from the image, and the presence or absence of another vehicle is determined for the vehicle area 34 remaining after the deletion. For example, it is determined whether another vehicle exists in the vehicle area 34 based on the area and shape of the vehicle area 34 remaining after the deletion.

  If it is determined that there is another vehicle (step S209; Y), the process returns to step S202, the same processing is performed on this vehicle area 34 (S202 to S208), and if it is determined that there is no further vehicle. (Step S209; N), the vehicle separation process is terminated.

  By this process, a plurality of vehicles that are photographed on the screen can be separated and detected one by one, and the traffic volume measurement accuracy is improved.

  Next, the light integration process will be described.

  FIG. 8 shows the flow of the light integration process, and FIGS. 9 and 10 show an example of an image in the process. FIG. 9A shows an example of an image obtained by photographing the road situation with the camera 11 at night. In FIG. 9A, for convenience of explanation, the vehicle C and the background road are clearly shown. However, the ambient illuminance is actually low, and it is difficult to obtain the contour information of the vehicle C. Therefore, focusing on high-luminance parts such as headlights, the position of the vehicle is detected using a pair of headlights as specific locations of the vehicle.

  First, a high luminance part is extracted from the captured image (FIG. 8: step S301, FIG. 9B). In FIG. 9B, in addition to the headlight, lamps installed on the roof and side portions of the vehicle C are extracted as high-luminance portions.

  Next, a pair of high luminance portions having characteristics corresponding to a pair of headlights is searched from the extracted high luminance portions, and this is detected as a candidate for a pair of headlights (hereinafter referred to as a light pair). (Step S302). The light pair candidates are those in which the distance in the X direction and the distance in the Y direction in the world coordinates between the lights are equal to or less than a predetermined threshold value. As a result, a pair of high-intensity parts that are equal to or less than the predetermined light interval and are arranged almost horizontally are detected as light pair candidates. As for the X direction, the minimum value and the maximum value of the distance in the X direction in the world coordinates may be determined, and those that fall between the minimum value and the maximum value may be extracted as light pair candidates. In this way, it is possible to eliminate a pair of high luminance portions that are too close as a headlight.

  FIG. 10A shows a portion detected as a light pair candidate surrounded by a broken line. In this example, three light pair candidates P1, P2, and P3 are found. Among these light pair candidates P1 to P3, the light pair P1 having the lowest camera coordinate (the lowest vehicle traveling direction side in the image) is selected as a base point for determining the vehicle position.

  The vehicle type is determined from the distance in the X direction of the light pair P1 selected as the base point, that is, the vehicle width (step S303). Then, the three-dimensional shape of the standard vehicle corresponding to the vehicle type is calculated (step S304). In other words, at the position in the real world corresponding to the base point detected earlier, the standard vehicle of the vehicle type indicated by the previous determination result is arranged along the road so that the specific location (a pair of headlights in this example) exists. Then, a three-dimensional shape that can be seen when the standard vehicle is viewed from the camera 11 is calculated.

  As shown in FIG. 10B, the vehicle region separation and integration unit 19 superimposes the calculated three-dimensional shape 41 on the image used for detecting the base point P <b> 1, and covers the range covered by the three-dimensional shape 41. It is determined that one vehicle exists. That is, it is determined that the light pair candidates P2, P3 and other high-intensity parts captured inside the three-dimensional shape 41 are all included in the same vehicle as the light pair P1. As a result, the light pair candidates P2 and P3 are excluded from the base point candidates for detecting other vehicles. In this example, a large number of high-intensity parts existing on the roof, sides, etc. of the large vehicle C are all contained within the three-dimensional shape (three-dimensional rectangular parallelepiped model) 41 of the large vehicle. Are integrated and counted as one vehicle.

  Thereafter, it is determined whether or not there are other light pair candidates outside the range covered by the three-dimensional shape 41 (step S306). When there are other light pair candidates (step S306; Y), Returning to step S304, detection of the next vehicle position and fitting of a three-dimensional shape are performed on the other light pair candidates (steps S304 and S305). If no other write pair exists (step S306; N), the process is terminated.

  By performing such a light integration process, it is possible to prevent erroneous measurement of a single vehicle as a plurality of vehicles during nighttime measurement.

  The embodiment of the present invention has been described with reference to the drawings. However, the specific configuration is not limited to that shown in the embodiment, and there are changes and additions within the scope of the present invention. Are also included in the present invention.

  For example, the standard vehicle may be provided with a vehicle type other than a large vehicle or a small vehicle, or a long trailer may be set. Further, although a rectangular parallelepiped model is used as the standard vehicle, the shape of the standard vehicle is not limited to this, and may be set as appropriate according to the shape of the target vehicle.

  The detection of the position of the vehicle is not limited to detection based on a specific location such as the front end of the vehicle or a headlight pair. In the embodiment, the three-dimensional shape of the standard vehicle is calculated according to the position of the base point (vehicle position) detected from the image, but a fixed three-dimensional shape is used regardless of the detected base point position. It doesn't matter. For example, when the vehicle is photographed at a substantially constant position in the image, the three-dimensional shape of the standard vehicle corresponding to the position is calculated and stored in advance, and the three-dimensional shape is superimposed on the image. You may do it. Also, a plurality of representative base point positions are set, and a three-dimensional shape corresponding to each base point is calculated and stored in advance, and from these, the position of the base point actually detected on the image is determined. You may comprise so that a three-dimensional shape may be selected and used. In this way, the processing load is reduced as compared with the case where the three-dimensional shape is calculated each time.

  Further, in the embodiment, the case where the camera 11 is installed so as to photograph the vehicle from diagonally forward is illustrated, but the installation state of the camera 11 is not limited thereto, and the vehicle is photographed from the front or passed by Even when the camera 11 is installed so as to take a picture from behind, the present invention works effectively. For example, even when the camera 11 is installed so as to photograph the vehicle from the front, the vehicles before and after traveling in the same lane with a close distance may be photographed. At this time, in the present invention, the range in which one vehicle (for example, one vehicle in front) exists is specified by fitting the three-dimensional shape, so that the front and rear vehicles can be separated and recognized.

It is a block diagram which shows schematic structure of the traffic measuring device concerning embodiment of this invention. It is explanatory drawing which shows an example of the image image | photographed with the camera of the traffic volume measuring apparatus concerning embodiment of this invention. It is explanatory drawing which shows the relationship between a camera coordinate and a world coordinate. It is a flowchart which shows the flow of the whole process which the traffic measuring device concerning embodiment of this invention performs. It is a flowchart which shows the flow of a vehicle separation process. It is explanatory drawing which illustrated the image (vehicle area extraction, base point detection) of the process of a vehicle separation process. It is explanatory drawing which illustrated the image (superposition of a three-dimensional shape, the outer area | region of a three-dimensional shape) of the process of a vehicle separation process. It is a flowchart which shows the flow of a light integration process. It is explanatory drawing which illustrated the image (a picked-up image, a high-intensity part extraction image) of the process of a light integration process. It is explanatory drawing which illustrated the image (Light pair candidate detection, three-dimensional shape superimposition) of the process of a light integration process.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Traffic measuring device 11 ... Camera 12 ... Processing part 13 ... Image acquisition part 14 ... Image memory 15 ... Pre-processing part 16 ... Vehicle position and vehicle type determination part 18 ... Coordinate conversion part 19 ... Vehicle area | region separation integration part 21 ... Vehicle Tracking unit 22 ... Traffic volume measuring unit 32 ... Vehicle region 33, 41 ... Calculated three-dimensional shape 34 ... Vehicle region remaining after deletion A, B, C ... Vehicles P1, P2, P3 ... Light pair candidates K ... Base point

Claims (9)

In the traffic measurement method that measures the traffic volume by analyzing the image obtained by photographing the vehicle traveling on the road with the camera,
A detection step of detecting a position of the vehicle on an image obtained by photographing with the camera;
A superimposing step of obtaining a three-dimensional shape when a predetermined standard vehicle arranged on the road is viewed from the camera, and superimposing the three-dimensional shape on the image in accordance with the position of the detected vehicle;
And a determination step of determining that one vehicle is present in the range covered with the three-dimensional shape on the image. The superimposing step calculates a three-dimensional shape when the standard vehicle arranged at a position in the real world corresponding to the position of the vehicle detected on the image is viewed from the camera. The traffic volume measuring method according to claim 1. The traffic volume measuring method according to claim 1 or 2, wherein it is determined whether or not another vehicle exists in a region outside the range covered with the three-dimensional shape on the image. As the standard vehicle, a plurality of types of standard vehicles such as a large vehicle and a small vehicle are set, and
A vehicle type determination step for determining the vehicle type of the vehicle whose position is detected in the detection step;
The traffic volume measuring method according to any one of claims 1 to 3, wherein a three-dimensional shape of a standard vehicle corresponding to a vehicle type indicated by a determination result of the vehicle type determination step is superimposed on the image. 5. The traffic volume measuring method according to claim 1, wherein in the detecting step, the position of the vehicle is detected by detecting a base point where a specific location of the vehicle exists on the image. The specific location is the front end of the vehicle,
6. The traffic volume measuring method according to claim 5, wherein, in the detecting step, a front end portion of a vehicle existing closest to a traveling direction of the vehicle in the image is detected as the base point. The specific portion is a pair of headlights,
In the detection step, a pair of high-intensity portions having characteristics corresponding to a pair of headlights are detected as a pair of headlight candidates, and the vehicle traveling direction is the most in the image among the detected candidates. The traffic volume measuring method according to claim 5, wherein a pair of headlight candidates existing on a side is detected as the base point. A day / night discrimination step for discriminating between day and night,
If the discrimination result of the day / night discrimination step is noon,
The specific location is the front end of the vehicle, and in the detection step, the front end of the vehicle existing closest to the vehicle traveling direction in the image is detected as the base point,
When the discrimination result of the day / night discrimination step is night,
The specific portion is a pair of headlights, and in the detection step, a pair of high-intensity parts having characteristics corresponding to the pair of headlights is detected as a pair of headlight candidates, and the detected candidate The traffic volume measuring method according to claim 5, wherein a candidate of a pair of headlights existing closest to the vehicle traveling direction in the image is detected as the base point. In a traffic measuring device that measures traffic by analyzing images obtained by shooting a vehicle traveling on a road with a camera,
A traffic measuring device, comprising: a processing unit that executes the traffic measuring method according to claim 1 for an image obtained by photographing with the camera.

Images