JP3605892B2 - Vehicle sensor and vehicle sensing method - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a vehicle sensor for recognizing a running vehicle by recognizing a vehicle image in an image captured by an imaging device such as a CCD camera, and a vehicle sensing method.
[0002]
[Prior art]
For the purpose of measuring a traffic flow, a vehicle detector that detects a traveling vehicle using image processing technology has been developed. 2. Description of the Related Art Conventionally, this type of apparatus uses a CCD camera to photograph a running vehicle. As shown in FIG. 5, an image indicating a feature point of the vehicle (the front end or the rear end of the vehicle) and its position are detected from the photographing screen obtained at a constant period. Next, the feature points of the vehicle are tracked over time on the screen obtained a predetermined number of times. The time when the vehicle first enters the measurement area of the CCD camera (time t in FIG. 5), the detected position of the characteristic point at that time, the time when the vehicle leaves the measurement area (time t + n in FIG. 5), and the characteristic point at that time It is recognized from the detection position that the vehicle has passed the measurement area. By repeating this process, the vehicle sensor counts the number of vehicles passing through the measurement area and calculates parameters for measuring the speed.
[0003]
[Problems to be solved by the invention]
As described above, in the related art, in order to confirm the passage of the vehicle, it is necessary that the characteristic points of the vehicle entering the measurement area exist in the photographed screen taken a predetermined number of times thereafter, and the vehicle can be tracked. Condition. However, if the succeeding vehicle of a specific vehicle is a large vehicle, the characteristic points of the specific vehicle are hidden by the characteristic points of the large vehicle on the shooting screen. As a result, if a predetermined number of photographing screens cannot be obtained when the feature points are hidden, the vehicle cannot track the feature points and cannot recognize the passage of the vehicle. Therefore, the number of vehicles cannot be counted. When a predetermined number of photographing screens are obtained, the vehicle sensor may incorrectly determine that the vehicle has exited even though the vehicle is present in the measurement area. When this situation occurs in the speed measurement, the value of the measured speed of the vehicle becomes higher than the actual speed value. This causes a measurement error.
[0004]
In addition, the image of one vehicle acquired from the photographing screen is divided, and when the divided image includes the image feature of the front end portion or the rear end portion indicating the feature of the vehicle, two vehicles are included. There has been proposed a vehicle sensor that determines that the vehicle is close to each other and obtains two vehicle images by performing image separation (OMRON TECHNICS Vol No. 2 (No. 102)). It is effective when the speed is low, but the image separation processing is performed on each vehicle image, so the same processing as detecting a vehicle several times as much as the actual vehicle is performed, and finally the speed etc. It is suitable for sensing high-speed traveling vehicles because it takes time to obtain, and the divided images differ depending on the large and small vehicle types, so if the vehicle type is incorrectly determined, the sensing accuracy of the vehicle will be reduced. I can't say it yet It should be good was left.
[0005]
In view of the above, an object of the present invention is to provide a vehicle sensor and a vehicle sensing method for improving the vehicle sensing performance when a vehicle is adjacent to the vehicle.
[0006]
[Means for Solving the Problems]
The invention according to claim 1 is a vehicle sensor that captures a vehicle that has entered a measurement area and measures the behavior of the vehicle from the image position of the vehicle image in the captured screen. A storage unit for storing a position as history data for vehicle tracking, and detecting an abnormal situation in which two vehicles running in the measurement area are close to each other and are captured as one vehicle image in the capture screen Abnormality detecting means, and when the abnormal situation is detected, the vehicle position of one of the two vehicles at the present time is determined from the history data of the storage means before the current time when the abnormal situation is detected. Calculating processing means for correcting one of the two vehicles as the corrected vehicle position and the other as the vehicle position obtained from the one vehicle image. before the history data Processing means creates a, wherein one vehicle position is preceding vehicle the vehicle position in the two vehicles, the other vehicle position is in the follower vehicle of the vehicle position in the two vehicles The arithmetic processing means acquires the inter-vehicle distance between the two vehicles from the history data and the vehicle position of the following vehicle, and acquires the vehicle position of the preceding vehicle from the vehicle position of the following vehicle and the acquired inter-vehicle distance. characterized in that it.
[0010]
According to a second aspect of the present invention, there is provided a vehicle detector for photographing a vehicle entering a measurement area and measuring a behavior of the vehicle from an image position of a vehicle image in the photographing screen. In the vehicle sensing method when captured as one vehicle image, the vehicle detector has a storage unit, a detection unit, and an arithmetic processing unit, and a position of the same vehicle obtained in time series from a plurality of photographing screens. Is stored in the storage means in the vehicle sensor as history data for vehicle tracking, and two vehicles traveling in the measurement area approach each other and are captured as one vehicle image in the capturing screen. The abnormal situation is detected by the detecting means , and when the abnormal situation is detected by the detecting means , the one vehicle image is regarded as an image of a succeeding vehicle of the two vehicles and the calculation is performed. The processing means Get the vehicle position of the current connection car, the inter-vehicle distance of the two vehicles acquired by the processing means from the history data, the preceding vehicle from the following vehicle of the current vehicle position and the acquired vehicle distance The present vehicle position is obtained by the arithmetic processing means .
[0011]
In the present invention, attention is paid to the fact that the distance between adjacent vehicles is substantially constant for a short time, and an abnormal situation in which two vehicles are captured as one vehicle image is detected. The two vehicle images are regarded as one of the two vehicle images. Further, the other vehicle image is calculated from the preceding inter-vehicle distance and the current vehicle position.
[0012]
According to the present invention, further , since a small car is often hidden by a subsequent large car just before exiting the measurement area, one vehicle image is determined to be a following car, and the position of the preceding car that has been hidden is determined. It is acquired using history data, for example, the distance between vehicles.
[0014]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0015]
FIG. 1 shows the arrangement of the vehicle sensors. The vehicle detector includes a CCD camera 10 and a vehicle detector main body 20. The CCD camera 10 is installed at a high place at the end of a road, and photographs a vehicle traveling in a plurality of lanes. There are two types of directions of the CCD camera: a direction for photographing the vehicle from the front and a direction for photographing the vehicle from the rear. The vehicle detector main body 20 performs image processing on an image photographed by the CCD camera 10 as described later, so that an adjacent vehicle also detects the vehicle.
[0016]
FIG. 2 shows a system configuration of the vehicle detector. 2, the vehicle sensor main body 20 includes a data input / output unit 11 and an image processing unit 21. An image for one screen transmitted from the CCD camera 10 is converted from analog to digital by the A / D converter 12 and stored in the image memory 26. The preprocessor 28 dedicated to image processing mainly performs pixel processing, and performs the same image processing as before, such as background subtraction for extracting the entire object and spatial difference for obtaining the contour of the object, and pre-processes the captured image. To process. Various data used in the image processing are temporarily stored in the local memory 27.
[0017]
The vehicle characteristic data obtained as a result of the image processing is stored in the dual port memory 25 via the image bus 29. The main processor 22 reads out the vehicle characteristic data from the dual port memory 25, determines the type of the vehicle by a method such as pattern matching, detects the current traveling position of the vehicle, and detects the speed of the vehicle and the speed of the passing vehicle. Measure the number of The processing executed in the preprocessor 28 and the main processor 22 will be described later with reference to FIG. The local memory 23 temporarily stores various data used in the processing of the main processor 22. The data on the speed and the number of passing vehicles obtained as a result of the measurement are transmitted to the outside via the CPU bus 24 and the image input / output port (I / O) 13 together with other data related to the passing vehicle, for example, the vehicle type.
[0018]
FIG. 3 shows a measurement process executed in such a configuration. With reference to FIG. 3, the system operation of the vehicle detector and the vehicle detecting method according to the present invention will be described. The road is one lane for ease of explanation. Prior to the measurement, it is assumed that a measurement area (measurement sample point) used for tracking the vehicle is set on the local memory 23 (step S10 in FIG. 3).
[0019]
When the vehicle enters the measurement area, the vehicle is captured by the CCD camera 10. The preprocessor 28 reads an image for one screen taken by the CCD camera 10 from the image memory 26 and executes preprocessing for vehicle detection. More specifically, an image of the approaching vehicle is obtained by subtracting the background screen of the measurement area prepared in advance from the input screen from the CCD camera 10. In this image, a spatial difference process for detecting an edge portion (contour line) where the luminance value changes rapidly is executed. The edge-detected image is binarized and passed to the main processor 22 (steps S20 to S40 in FIG. 3).
[0020]
The main processor 22 detects the presence of the vehicle and the type of the vehicle by performing pattern matching of the outline image of the characteristic portion of the vehicle obtained by the binarization processing with the reference outline image of the known vehicle type. The position of the vehicle at the time of shooting is also calculated from the position of the image determined to be the vehicle. These acquired data are stored in the local memory 23 together with the time obtained from the clock function of the main processor 22 (step S50 in FIG. 3). When the presence of the vehicle is confirmed (Yes in step S60 in FIG. 3), the main processor 22 executes the tracking process of the vehicle (step S70 in FIG. 3).
[0021]
In this tracking process, the position data and speed data of the same vehicle in the measurement area are stored every time a shooting screen is input, and two vehicles approach abnormally, and the same as one large vehicle on the screen This is a process of predicting an abnormal situation that results in a strange image using the stored history data. The details of this process are shown in FIG. For the vehicle that has entered the measurement area for the first time, the history data relating to the position and speed is not stored in the local memory 23. Thereby, the main processor 22 determines that the vehicle has entered the measurement area for the first time (affirmative determination in step S200 in FIG. 4). The main processor 22 secures an area for storing the tracking history data of the vehicle in the local memory 23 for initial processing, and stores, as an initial value of the history data, data indicating the currently obtained vehicle position and vehicle type. (Step S210 in FIG. 4). In addition, with reference to the speed data of the history data of the preceding vehicle obtained so far, it is calculated to what position (referred to as a predicted position) the entering vehicle will have moved at the next image input time. This calculation can be performed by a simple formula (current vehicle position + moving distance (speed × photographing interval time)), so that detailed description will not be required. The area on the screen corresponding to the area of a certain area centered on the obtained predicted position becomes the next vehicle search area, in other words, the area on the screen for tracking the same vehicle (step S220 in FIG. 4).
[0022]
The vehicle that has entered the measurement area is captured by the CCD camera 10 a plurality of times while traveling in the measurement area. The second captured screen is preprocessed by the preprocessor 28 as described above, and delivered to the main processor 22 in the form of a binary image. The main processor 22 detects a vehicle position, a speed, and the like in the set search area on the input screen (steps S20 to S50 in FIG. 3). The processing procedure proceeds from step S60 to step S70, and in the tracking processing of FIG. 4, the processing of identifying the vehicle that has been captured for the first time last time is performed. More specifically, a vehicle type determination process for confirming that the vehicle type detected this time matches the vehicle type detected last time is executed. Further, the change in the speed of the vehicle before the present time is calculated, and it is confirmed that the speed of the current vehicle has not changed more rapidly than the previous time. As an example, it is confirmed by comparison processing that the speed difference between the previous time and the current time is smaller than a threshold value (steps S200 → S230, S240 in FIG. 4). When the vehicle types do not match or when the vehicle shows a rapid change in speed, it is determined that the two vehicles are close to each other and the vehicle is photographed as one vehicle in the photographing screen, and the abnormality is dealt with after step S80 in FIG. Move on to processing.
[0023]
On the other hand, if a normal determination is obtained in the above determination process, the detected position and speed ((current vehicle position−previous vehicle position) ÷ obtained by calculating the photographing interval time) are added to the local memory 23 as history data. I do. Further, the next search area is set (step S250 → S260 in FIG. 4). Therefore, while the proximity state of the two vehicles does not occur and the one vehicle is normally confirmed, the position and the speed detected by the loop processing of steps S20 to S80 in the processing procedure of FIG. Is stored in the local memory 23. When the vehicle exits the measurement area, the exit is detected on the condition that the vehicle image has disappeared from the shooting screen and that the previous vehicle position is located in the exit area. This terminates the tracking of the vehicle, and executes conventional processing such as calculation of traffic flow and detection of parking / abnormal traffic (steps S130 → S140 in FIG. 3).
[0024]
When an abnormality is detected in the tracking processing of FIG. 4, the entire history data in the local memory 23 is searched to determine whether or not a following vehicle exists within a certain distance in the previous shooting screen. As an example, the distance between a plurality of vehicle positions stored in the local memory 23 and the vehicle currently being tracked is calculated, and it is determined whether the distance is smaller than a certain distance. When it is determined that there is no following vehicle, the process proceeds to the tracking end determination process thereafter (step S90 in FIG. 4).
[0025]
On the other hand, when the determination that there is a following vehicle is obtained, it is determined that two vehicles have been photographed as one vehicle, and the position and speed of the vehicle that has been determined to be abnormal are corrected by the following processing. In this embodiment, the point that the vehicle traveling in the same direction cannot overtake the preceding vehicle unless the lane is changed, and that two vehicles are close to each other, and the distance between the vehicles is almost the same for a very short time of 1 second or less. Paying attention to the fixed point, when a situation occurs in which the preceding vehicle cannot be tracked, the vehicle that cannot be tracked at the current time is determined from the vehicle position of the following vehicle in the vicinity and the inter-vehicle distance determined from the two vehicle positions acquired last time. Calculate the position of (preceding vehicle). The speed is calculated by using the correction value obtained by this calculation (current corrected vehicle position−previous detected vehicle position) ÷ photographing interval time. (Step S100 → S110 in FIG. 4). In this way, even when two vehicles approach each other and are captured as one vehicle, sensing of the two vehicles can be performed without performing complicated image separation processing as in the related art.
[0026]
The following example can be implemented in addition to this embodiment.
[0027]
1) In this embodiment, the vehicle speed is calculated from the vehicle position obtained from the current screen, the vehicle position obtained from the previous screen, and the shooting interval time. Speed may be calculated.
[0028]
2) The image processing for one lane has been described in the present embodiment. However, when a plurality of lanes are included in the measurement area, the same processing as described above is performed for each lane.
[0029]
3) In this embodiment, the area for searching for a tracked vehicle when tracking the vehicle is set only to the road portion in the input screen. As a result, the entire screen is not a search target, so that high-speed processing can be realized. When it is desired to improve the tracking accuracy, the size of the search area may be variably set in association with the vehicle speed and a change (acceleration) in the vehicle speed. For example, when the vehicle speed is high, the search area is set because the speed change is large, and when the vehicle speed is increasing, the search area is set large.
[0030]
4) The vehicle position at which two vehicles are photographed as one image differs depending on the installation position of the vehicle sensor. Further, depending on the direction of the CCD camera, a case where the preceding vehicle is hidden by the following vehicle and a case where the following vehicle is hidden by the preceding vehicle may occur. When the causal relationship between the vehicle hiding position and the hiding vehicle is clear, the main processor 22 determines the type of the vehicle (preceding vehicle / subsequent vehicle) to be corrected from the positional relationship when the vehicle is one vehicle. Is also good.
[0031]
5) A method other than the method described in this embodiment can be used to detect an abnormal situation in which two vehicles are photographed as one vehicle. For example, when a plurality of vehicle features are detected in the vehicle image on the input screen, a method of determining that an abnormal situation has occurred can be used.
[0032]
【The invention's effect】
As described above, in the present invention, an abnormal situation in which two vehicles are captured as one vehicle image is detected, and this one vehicle image is used as one vehicle image out of two vehicles. Is considered. Further, since the other vehicle image is calculated from the vehicle interval at the previous time and the vehicle position at the current time, the positions of the two vehicles can be sensed by a simpler process than before. Further, the processing time is also reduced.
[0033]
Furthermore, in the present invention, a small vehicle is often hidden by a subsequent large vehicle just before exiting the measurement area. Therefore, one vehicle image is determined to be a following vehicle, and the position of the preceding vehicle that has been hidden is determined from history data. The information is obtained by using the obtained information, for example, the inter-vehicle distance. Thus, the positions of two vehicles can be accurately grasped without separating one vehicle image.
[Brief description of the drawings]
FIG. 1 is a perspective view showing a system installation state according to an embodiment of the present invention.
FIG. 2 is a block diagram illustrating a system configuration according to an embodiment of the present invention.
FIG. 3 is a flowchart showing a processing procedure according to the embodiment of the present invention.
FIG. 4 is a flowchart illustrating details of a vehicle tracking process in step S70 of FIG. 3;
FIG. 5 is an explanatory diagram for explaining a conventional vehicle sensing method.
[Explanation of symbols]
10 CCD camera 11 Terminal communication unit 12 A / D converter 13 I / O
Reference Signs List 20 vehicle sensor body 21 image processing unit 22 main processor 23 local memory 24 CPU bus 25 dual port memory 26 image memory 27 local memory 28 preprocessor 29 image bus

Claims (2)

In a vehicle sensor that captures a vehicle that has entered the measurement area and measures the behavior of the vehicle from the image position of the vehicle image in the captured screen,
Storage means for storing the position of the same vehicle obtained in time series as history data for vehicle tracking,
Abnormality detection means for detecting an abnormal situation in which two vehicles traveling in the measurement area are close to each other and are captured as one vehicle image in the imaging screen;
When the abnormal situation is detected, arithmetic processing means for correcting one vehicle position of the two vehicles at the present time from history data of the storage means before the present time when the abnormal situation is detected, and Calculating the current history data by considering one vehicle position of the two vehicles as the corrected vehicle position and the other vehicle position as a vehicle position obtained from the one vehicle image. Created by the processing means , the one vehicle position is a vehicle position of a preceding vehicle in the two vehicles, the other vehicle position is a vehicle position of a following vehicle in the two vehicles, The arithmetic processing means acquires the inter-vehicle distance between the two vehicles from the history data and the vehicle position of the following vehicle, and acquires the vehicle position of the preceding vehicle from the vehicle position of the following vehicle and the acquired inter-vehicle distance. Vehicle feeling characterized by that Vessel. In a vehicle sensor that captures a vehicle that has entered the measurement area and measures the behavior of the vehicle from the image position of the vehicle image in the capture screen, two vehicles are captured as one vehicle image in the capture screen. In the vehicle detection method when
The vehicle detector has a storage unit, a detection unit, and an arithmetic processing unit,
The position of the same vehicle obtained in time series from a plurality of shooting screens is stored in the storage unit in the vehicle sensor as history data for vehicle tracking,
An abnormal situation in which two vehicles traveling in the measurement area approach each other and are captured as one vehicle image in the capturing screen is detected by the detection unit ,
If the detecting means detects the abnormal situation, the one vehicle image is regarded as an image of a succeeding vehicle of the two vehicles, and the current vehicle position of the following vehicle is determined by the arithmetic processing means. And get
The inter-vehicle distance between the two vehicles is obtained from the history data by the arithmetic processing means ,
A vehicle sensing method for a vehicle sensor, wherein the arithmetic processing means acquires the current vehicle position of the preceding vehicle from the current vehicle position of the following vehicle and the acquired inter-vehicle distance.

Images