JPH1040490A - Road monitoring device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a road monitoring device with which a vehicle accident on a road can be exactly detected. SOLUTION: The road monitoring device RM is provided with an image pickup means 11 installed on a road 10 for picking up images catching road conditions along time sequence. Based on the time sequential images, a vehicle action detecting means 14 detects the action of vehicle. The detected vehicle action is compared with an accident pattern previously stored in a vehicle accident discriminating means 15. When the vehicle accident is coincident with the accident pattern, the vehicle accident is detected.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a road monitoring device for monitoring road conditions widely, and more particularly to a road monitoring device used in an automatic highway driving system.

[0002]

2. Description of the Related Art A road monitoring apparatus for detecting the presence or absence of a vehicle in a lane is disclosed in, for example, Japanese Patent Application Laid-Open No. 7-244796. In this road monitoring device, only the presence or absence of a vehicle on the road is detected.

[0003]

In road surveillance, if it is possible to grasp not only the presence or absence of a vehicle but also the presence or absence of a vehicle accident, the processing for responding to the vehicle accident is quickly performed and the traffic of the vehicle is spread in a short time. It is possible to do.

In particular, for example, in an automatic highway driving system, the speed and steering of a traveling vehicle are automatically operated. Therefore, it is desired to grasp the road condition ahead of the vehicle in advance and execute speed adjustment and steering control according to the road condition. It is important to accurately and quickly detect situations that hinder traffic flow such as vehicle accidents. In addition, it is necessary to prevent a vehicle following the accident vehicle from being secondarily involved in the accident.

[0005] The present invention has been made in view of the above circumstances, and has as its object to provide a road monitoring device capable of accurately detecting a vehicle accident on a road.

[0006]

According to a first aspect of the present invention, there is provided an image pickup means which is installed on a road and captures an image of a road in a time series. A road monitoring device is provided, comprising: vehicle behavior detecting means for detecting a vehicle behavior based on an image along the road; and vehicle accident determining means for detecting a vehicle accident based on the detected vehicle behavior. . According to such a configuration, a vehicle accident on the road can be quickly and accurately detected.

The vehicle accident determining means may detect a vehicle accident by comparing the detected vehicle behavior with a previously stored accident pattern. According to such detection, reliable accident detection can be realized with a simple configuration.

[0008] As the index of the vehicle behavior, a speed component such as the speed and acceleration of the vehicle may be used. Also, in a special situation such as an accident, it is considered that a remarkable change is observed in the lateral speed component that cannot be detected in normal traveling. Therefore, when a lateral speed component is used as an index of the vehicle behavior, more accurate accident detection can be performed.

In detecting the vehicle behavior, an optical flow processing circuit for calculating a relative velocity vector of a pixel in a moving image may be used. By using a so-called optical flow, the behavior of the vehicle can be easily calculated without specifying the vehicle itself. In detecting the vehicle behavior, a vehicle specifying circuit that specifies the position of the vehicle on the screen, an XY plane conversion circuit that projects the specified vehicle position on an XY plane along the road surface, and Alternatively, a vehicle trajectory drawing circuit that describes a trajectory of the vehicle from the projected position of the vehicle may be used. In the latter case, a vehicle accident may be detected by comparing the depicted traveling locus with an accident pattern stored in advance.

[0010]

Preferred embodiments of the present invention will be described below with reference to the accompanying drawings.

FIG. 1 shows the basic concept of a road monitoring device RM according to the present invention. The road monitoring device RM includes a camera device 11 that is installed on a road 10 and that captures an image of a road condition at an arbitrary time interval (for example, 30 μm).
The camera device 11 sequentially outputs images capturing road conditions in chronological order. The output image signal is, for example, MP
Data compression means 12 using an image compression technique such as EG
Then, it is sent to the control center 13.

The control center 13 processes the transmitted image signal using, for example, a computer system.
The computer system includes, for example, a vehicle behavior detecting unit 14 that detects a vehicle behavior based on a time-series image.
And a vehicle accident determining means 15 for detecting a vehicle accident based on the detected vehicle behavior. When a vehicle accident is detected, a response process such as sending a rescue vehicle can be performed in a short time, and the speed control and steering control of the automatic highway driving system can be accurately performed.

In response to a command from the vehicle accident determining means 15, the recording device 16 records an image signal for a predetermined time before and after the time when the accident was specified. Such records can be used to determine the cause of the accident. As the recording device 16, for example, a magnetic disk or an optical disk or any type of recording medium is used.

The vehicle accident judging means 15 refers to the accident pattern stored in the accident pattern storing means 17 when judging a vehicle accident. The accident pattern is created with various accident patterns in mind, as shown in FIG.
In the present embodiment, for example, a course-out accident (FIG. 2) and a rollover accident (FIG.
), Collision with the side wall and adjacent vehicle (Fig. 2
), A collision accident between the front and rear vehicles (FIG. 2) is specified.

FIG. 3 shows a computer system of the road monitoring device RM according to the first embodiment of the present invention. The vehicle behavior detecting means 14 includes an optical flow processing circuit 20 for calculating a relative velocity vector of a pixel in a moving image, that is, an optical flow. To calculate the optical flow,
Utilizing the assumption that the distribution of gray values of pixels is invariant during motion, the method of relating the optical flow velocity of pixels to the spatial / temporal gradient of the gray distribution of moving images (gradient method), and two consecutive images A method (matching method) of associating frames using a template of an appropriate size (matching method) may be adopted (for example, Norihito Majima et al., "A study on preceding vehicle region extraction by optical flow" IEICE Technical Report, IE94 -144, PRU9
4-144, pp. 113-120).

The optical flow area determining circuit 21 of the vehicle behavior detecting means 14 specifies an area in the screen where an optical flow is generated, and identifies the presence of a vehicle from the area M of the specified area. For example, when the calculated area M is smaller than the predetermined threshold area m, it is determined that a small animal or a paper bag crossing the road has generated an optical flow, and the accident judgment processing of the vehicle accident judgment means 15 is omitted.

The vehicle accident judging means 15 judges the presence or absence of a vehicle accident based on the detected vehicle behavior, that is, the optical flow. The behavior comparison circuit 22 compares the detected optical flow pattern with the accident optical flow pattern representing the vehicle behavior shown in FIG. If the detected optical flow pattern corresponds to one of the accident optical flow patterns, it is determined that "there is an accident".

The accident optical flow pattern storage means 23 stores, for example, an optical flow pattern shown in FIG. The optical flow pattern in FIG. 4 shows the vehicle behavior during normal running (FIG. 2). As long as this optical flow pattern continues, it can be determined that no accident has occurred. As shown in FIG. 4, according to the optical flow pattern in which an optical flow that opens right and left from the traveling direction is found, it is possible to specify a rollover accident caused by a sharp turn of the vehicle shown in FIG. 2. As shown in FIG. 4, according to the optical flow pattern in which an optical flow largely flowing to one of the left and right from the traveling direction is found, it is possible to specify a collision with the side wall or the adjacent vehicle (FIG. 2). In this case, immediately after that, if the flow of the optical flow is reversed as shown in FIG. 4, such a collision can be specified more reliably. This is because it is considered that the collision causes the accident vehicle to bounce in the opposite direction. Also, if the optical flow disappears immediately after the optical flow pattern of FIG. 4 is observed, as shown in FIG. 4, it can be determined that the vehicle has completely exited the course and that “there is an accident”. Further, if a sudden change in running speed occurs as shown in FIG. 4 immediately after the optical flow pattern of FIG. 4 is observed, it can be determined that a collision has been made from behind as shown in FIG. . This is because a collision from behind generates acceleration in the traveling direction.

Now, the road monitoring device R according to the first embodiment
Assume that M is monitoring a highway. As shown in FIG. 1, the vehicle VC enters the monitoring area of the camera device 11. An automatic driving vehicle (not shown) follows this vehicle VC. The camera device 11 repeats image capturing at predetermined time intervals. The captured image is compressed and sent to the control center 13.

In the control center 13, an optical flow processing circuit 20 detects an optical flow from two continuous images. Since the screen itself is fixed and the background is also fixed, the behavior of the vehicle is specified by the optical flow. The optical flow area determination circuit 21 distinguishes a small animal or a paper bag crossing the road from the traveling vehicle from the detected optical flow area M.

When the behavior of the vehicle is detected, the vehicle accident judging means 15 judges whether a vehicle accident has occurred. The detected optical flow pattern is compared with the accident optical flow pattern shown in FIG. 4, and when it is recognized that both patterns match, it is determined that “there is an accident”.

The information of "having an accident" is first transmitted to an autonomous vehicle following the accident vehicle. In this self-driving vehicle, the speed of the vehicle can be reduced, the steering can be turned to avoid the accident vehicle, or the self-driving can be canceled based on the "accident" information. Since "accident" can be instantaneously detected on the basis of at least a few continuous screens, it is possible to avoid as much as possible that the following vehicle is secondarily involved in the accident of the preceding vehicle. in this case,
Even if the accuracy of the “accidents” information is not high, it can be said that detecting the suspicious behavior of the preceding vehicle promptly has its own meaning.

At the same time, the vehicle accident determining means 15 sends a recording command to the recording device 16. The recording device 16 records an image signal for a certain period of time before and after the accident according to the recording command. In addition, such "accident" information may be used to call a rescue vehicle.

As described above, according to this road monitoring device, by adopting the optical flow, a vehicle accident can be detected promptly, and post-processing and accident prevention can be achieved instantaneously.

FIG. 5 shows a computer system of a road monitoring device RM according to a second embodiment of the present invention. The second embodiment is characterized in that a lateral speed (speed in the width direction of the road) and a lateral acceleration are used as the behavior of the vehicle. That is, as shown in FIG. 6, for example, in the case of FIGS. 2 to 2, it is considered that the front-rear direction speed suddenly decreases simultaneously with the occurrence of an accident, and the unprecedented lateral speed is instantaneously generated. is there. Therefore, the vehicle behavior detecting means 14
Is the travel position serving as a reference for the vehicle trajectory in each of the images 30 and 31, regardless of whether the image 30 captures one lane as shown in FIG. 7 or the image 31 captures two lanes as shown in FIG. 7. Thus, the lateral motion vector 33 from the target 32 is specified.

In this computer system, as shown in FIG. 5, the vehicle behavior detecting means 14 includes an optical flow processing circuit 34 similar to that described above. The motion vector detection circuit 35 calculates an average vector, that is, a motion vector of the vehicle, from the detected optical flows. Using this motion vector, the speed calculation circuit 36 calculates the lateral speed, and the acceleration calculation circuit 37 calculates the lateral acceleration. Here, the lateral speed and the lateral acceleration are both indicators of the vehicle behavior.

The behavior comparing circuit 38 of the vehicle accident judging means 15
Determines the presence or absence of an accident by comparing the calculated lateral speed and lateral acceleration with the threshold for accident determination stored in the threshold storage means 39. The behavior comparing circuit 38 may judge "there is an accident" when the lateral speed exceeds a predetermined threshold, and may judge "there is an accident" when the lateral speed exceeding the predetermined threshold continues for a predetermined time. , "Accident exists" by comparing acceleration with a predetermined threshold value in addition to these indices
May be determined. The threshold value is used, for example, to distinguish the lateral speed at the time of changing lanes from the lateral speed at the time of an accident. As is clear from FIG. 6, not only the lateral speed but also
If the speed is used in accordance with the front-rear direction speed, more reliable accident detection can be performed.

In the vehicle behavior detecting means 14 described above,
Instead of the optical flow processing circuit 34, a vehicle specifying circuit for recognizing the position of the vehicle on the screen using template matching or edge detection may be provided. The motion vector of the vehicle is detected from the vehicle position recognized on each screen. Further, instead of the motion vector detection circuit 35, X
A -Y plane conversion circuit may be provided. According to the XY plane conversion circuit, the motion vector on the screen can be converted into an XY plane vector along the road surface. Therefore, the x-direction component (forward-backward speed) and the y-direction component (lateral speed) of the motion vector can be easily obtained.

FIG. 8 shows a computer system of a road monitoring device RM according to a third embodiment of the present invention. The feature of the third embodiment is that a running locus is used as the behavior of the vehicle. Using the traveling trajectory of the vehicle, for example, a case where the vehicle skids in one lane as shown in FIG. 9 (corresponding to FIG. 2), or a case where the vehicle goes out of course instantaneously as shown in FIG. 2) It is possible to specify a case where the vehicle is split due to a collision as shown in FIG. Further, as shown in FIG. 9, when the vehicle collides with a side wall or an adjacent vehicle (corresponding to FIG. 2), the collision can be identified because the moving angle α of the vehicle changes suddenly. Further, as shown in FIG. 9, the collision probability is calculated based on the approach vector and the distance between the two vehicles, and when the collision probability is equal to or more than the threshold value, it is determined that “there is an accident”, or as shown in FIG. In addition, the collision probability is calculated based on the approach vector and the distance between the vehicle and the side wall, and when the collision probability is equal to or more than the threshold, it can be determined that “there is an accident”.

In this computer system, as shown in FIG. 8, the vehicle behavior detecting means 14 includes a vehicle specifying circuit 40 for recognizing the position of the vehicle on the screen using template matching and edge detection. The specified position of the vehicle is projected on the XY plane by the XY plane conversion circuit 41. The vehicle trajectory drawing circuit 42 draws a running trajectory of the vehicle based on the projected vehicle position.

The behavior comparing circuit 43 of the vehicle accident judging means 15
Determines the presence or absence of an accident by comparing the depicted traveling locus with the accident locus pattern stored in the accident locus pattern storage circuit 44.

In this case, as shown in FIG. 10, the monitoring area is covered by one or more screens. X-
The monitoring area may be formed by connecting the screens before performing the conversion on the Y plane, or the travel trajectory covering the entire monitoring area may be calculated after performing the conversion on the XY plane.

[0033]

As described above, according to the present invention, a road monitoring apparatus capable of accurately detecting a vehicle accident on a road is provided.

[Brief description of the drawings]

FIG. 1 is a diagram showing a basic concept of a road monitoring device according to the present invention.

FIG. 2 is a diagram showing an accident pattern on a screen.

FIG. 3 is a schematic block diagram of a computer system of the road monitoring device according to the first embodiment.

FIG. 4 is a diagram showing an optical flow pattern.

FIG. 5 is a schematic block diagram of a computer system of a road monitoring device according to a second embodiment.

FIG. 6 is a diagram showing behavior of an accident vehicle.

FIG. 7 is a diagram showing a lateral vector as an index of a vehicle behavior.

FIG. 8 is a schematic block diagram of a computer system of a road monitoring device according to a third embodiment.

FIG. 9 is a diagram showing behavior of an accident vehicle.

FIG. 10 is a diagram showing a monitoring area of each screen.

[Explanation of symbols]

10 road, 11 camera device as imaging means, 14
Vehicle behavior detecting means, 15 Vehicle accident determining means, 20
Optical flow processing circuit, 40 vehicle identification circuit, 4
1 XY plane conversion circuit, 42 vehicle trajectory description circuit.

Continuation of the front page (51) Int.Cl. ⁶ Identification number Reference number in the agency FI Technical display location H04N 7/18 G08B 13/196 // G08B 13/196 G06F 15/62 380

Claims (7)

[Claims] 1. An image pickup means installed on a road for capturing an image of a road condition along a time series, a vehicle behavior detection means for detecting a vehicle behavior based on the image along the time series, A road accident monitoring means for detecting a vehicle accident based on the detected vehicle behavior. 2. The road monitoring device according to claim 1, wherein the vehicle accident determining unit detects a vehicle accident by comparing the detected vehicle behavior with a previously stored accident pattern. Road monitoring device characterized by: 3. The road monitoring device according to claim 1, wherein a speed component of the vehicle is used as an index of the vehicle behavior. 4. The road monitoring device according to claim 3, wherein a lateral speed component is used as the speed component. 5. The road monitoring device according to claim 1, wherein the vehicle behavior detecting unit includes an optical flow processing circuit that calculates a relative speed vector of a pixel in a moving image. Road monitoring device. 6. The road monitoring device according to claim 1, wherein the vehicle behavior detecting means includes a vehicle specifying circuit for specifying a position of the vehicle on a screen, and an X-axis along a road surface along the road surface. A road monitoring device comprising: an XY plane conversion circuit for projecting on a Y plane; and a vehicle trajectory drawing circuit for drawing a trajectory of the vehicle from the projected position of the vehicle. 7. The road monitoring device according to claim 6, wherein the vehicle accident determination unit detects a vehicle accident by comparing the depicted traveling trajectory with a previously stored accident pattern. Road monitoring device characterized by: