KR20160088099A - Method for determinating collision avoiding path of vehicle - Google Patents

Abstract

Disclosed is a method for setting a collision avoidance path of a vehicle. The method for setting a collision avoidance path of a vehicle comprises the steps of: generating a displacement map from a stereo image; separating an obstacle region and a road region by using information acquirable from the displacement map; classifying the separated obstacle region into a fixed obstacle region including an immovable obstacle and a dynamic obstacle region including a movable obstacle; setting a global collision avoidance region in which the fixed obstacle region is removed from the road region; setting a local collision avoidance region in which the dynamic obstacle region is removed from the global collision avoidance region; and setting a final collision avoidance path in the set local collision avoidance region.

Description

METHOD FOR DETERMINING COLLISION AVOIDING PATH OF VEHICLE "

The present invention relates to a collision avoidance path setting method for a vehicle, and more particularly, to a collision avoidance path setting method for a vehicle that avoids a collision with an obstacle ahead of the vehicle.

Recently, a camera based collision avoidance system has been developed for avoiding collision with an obstacle such as a vehicle or a person located in front of a vehicle.

Conventional camera-based collision avoidance system calculates the risk of collision with forward obstacle through radar and monopod camera, compares the distance to the obstacle required to avoid collision and current distance, . ≪ / RTI >

Thus, since the conventional camera-based collision avoidance system warns the driver of the risk of collision, the steering control of the vehicle for avoiding the obstacle when the warning occurs is dependent on the driver's actions.

However, in an imminent situation such as a collision danger situation, since the driver's spatial perception force is lower than normal, it often leads to a serious accident because it can not find a route to avoid the obstacle quickly.

It is therefore an object of the present invention to provide a vehicle collision avoidance path setting method capable of automatically avoiding collision with an obstacle even in an imminent situation such as a collision danger situation.

According to an aspect of the present invention, there is provided a method of setting a collision avoidance path for a vehicle, the method comprising: generating a displacement map from a stereo image; Classifying the separated obstacle area into a fixed obstacle area including a motionless obstacle and a dynamic obstacle area including a motion obstacle, Setting a local collision avoiding area in which the dynamic obstacle area is removed in the global collision avoiding area, and setting a final collision avoiding path in the set local collision avoiding area.

According to the present invention, a displacement map is generated from left and right images input from a stereo camera, and fixed obstacles and dynamic obstacles are grasped from the generated displacement map to predict a collision avoidance path in which a vehicle can avoid a collision. .

FIG. 1 is a block diagram schematically showing an internal configuration of a collision avoidance system according to an embodiment of the present invention.
2 is a flowchart showing a method of setting a collision avoidance path for a vehicle according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, advantages and features of the present invention and methods of achieving them will be apparent from the following detailed description of embodiments thereof taken in conjunction with the accompanying drawings.

The present invention may, however, be embodied in many different forms and should not be construed as being limited to the exemplary embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, And advantages of the present invention are defined by the description of the claims.

It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In the present specification, the singular form includes plural forms unless otherwise specified in the specification. &Quot; comprises "and / or" comprising ", as used herein, unless the recited component, step, operation, and / Or added.

FIG. 1 is a block diagram schematically showing an internal configuration of a collision avoidance system according to an embodiment of the present invention.

Referring to FIG. 1, it is assumed that the collision avoidance system 300 according to an embodiment of the present invention is provided in a vehicle. Of course, the application field of the collision avoidance system 300 according to an embodiment of the present invention is not limited to a vehicle, and can be applied to various electronic apparatuses throughout the industry in which a collision avoidance design with an obstacle is required. For example, the collision avoidance system 300 according to an exemplary embodiment of the present invention has a moving means and can be applied to a robot system used for exploration, bomb removal, security, and the like.

A collision avoidance system (300) provided in a vehicle includes a collision avoidance path system (100) for automatically setting a vehicle avoidance path to avoid a collision with an obstacle, a vehicle braking and vehicle And an active braking / control system 200 that performs wheel control.

The collision avoidance path system 100 includes a stereo camera 110, a disparity map generator 120, a lane detector 130, an area separator 140, A collision avoidance avoidance path setting unit 150, a second avoidance collision avoidance path setting unit 160, a collision prediction and avoidance deciding unit 170, and an HMI (human machine interface) unit 180.

A stereoscopic camera 110 may be installed at a specific location on the vehicle capable of photographing the front of the vehicle. The stereo camera 110 includes a left camera for photographing the left front of the vehicle and a right camera for photographing the right front of the vehicle.

A stereoscopic camera 110 acquires a stereo image including a left image 11A photographing the left front side of the vehicle by the left camera and a right image 11B photographing the right front side of the vehicle by the right camera do.

The obtained stereo image is transmitted to the displacement map generator 120. At this time, any one of the left image 11A and the right image 11B included in the stereo image is also transmitted to the lane detecting unit 130. FIG.

The lane detecting unit 130 detects the lane information 13 from any one of the images transmitted from the stereo camera 110 according to a lane detection algorithm. The lane detection algorithm may be a Huff transform, a deformable template model, a training based algorithm, or a dynamic programming algorithm according to a method of detecting a lane. Detailed description of each algorithm is omitted.

The displacement map generation unit 120 generates a displacement map 12 including the distance information d from the stereo image transmitted from the stereo camera 110 to the obstacle from the image data.

The displacement map generation unit 120 calculates a disparity value between a minutiae point of the left image and a minutiae point of the right image based on the same pixel and generates a displacement map 12 from the calculated displacement value can do. The generated displacement map 12 is transmitted to the area extracting unit 140.

The region extracting unit 140 separates and extracts the obstacle region 14A and the road region 14B including the obstacle from the displacement map 12 received from the displacement map generating unit 120. [

The region extracting unit 140 extracts the position information (x, y) of the obstacle obtained from the displacement map 12, the distance information d from the vehicle to the obstacle, the height h information (or size information) And the relative speed of the obstacle to the obstacle region 14A. Here, the membership voting method can be used as a method of extracting the height (h) information (or size information) of the obstacle. This membership voting method consists of three steps. In the first step, the cost matrix is generated from the membership value based on the boundary detection result between the road surface and the obstacle described below. In the last step, dynamic programming is used to minimize the cost matrix while minimizing the spatially smooth obstacle Estimate height information.

The region extracting unit 140 classifies the obstacle region 14A into the fixed obstacle region 14A-1 and the dynamic obstacle region 14A-2 using the extracted information as described above. For example, since the fixed obstacle has no motion, the rate of change of the information in a successive frame has a linear rate of change property, whereas the rate of change of the information in a continuous frame has a nonlinear rate of change property . That is, the fixed obstacle region 14A-1 and the dynamic obstacle 14A-2 can be classified through the analysis of the change rate characteristic. The extracted fixed obstacle region 14A-1 is used as information for extracting the global collision avoiding space, and the extracted dynamic obstacle region 14A-2 is transmitted to the first collision avoidance space extracting unit 150 Is transmitted to the second collision avoiding space extracting unit 160 and used as information for extracting the local collision avoiding space.

When the obstacle area is extracted in the area extracting unit 140, a process of extracting the road area 14B is performed. In the road area 14B extraction method, a road surface estimation technique and a road and obstacle boundary detection technique can be used.

In the road surface estimation method, the road surface can be estimated by modeling the road surface with the piece-wise linear function in the YZ-plane domain to estimate the road surface topography in the forward direction. In this method, the points having 3D position information obtained from the displacement map are accumulated in the YZ-plane, and points having a high possibility of being present on the road surface are sampled. Then, the Random Sample Consensus (RANSAC) Thereby estimating various types of road surfaces.

In the boundary detection method of obstacles, the obstacle boundary is detected based on the characteristic that the distances measured on the boundary surface of the obstacle and the distance of the corresponding road surface must be the same in case of obstacles existing on the road surface. To this end, the distances of the obstacles are estimated by extracting representative values using the median operator in an area where obstacles are expected to exist on the road surface estimated by the road surface estimation technique. Then, the cost matrix is calculated based on the difference between the estimated obstacle distance and the corresponding road surface distance. Since the location with the smallest value for each column in the calculated cost matrix is the position of the obstacle boundary for each column, the cost matrix is minimized, and at the same time, a spatially smooth path is estimated through dynamic programming, Can be detected.

Thus, the road area road area 14B can be extracted from the estimated road surface and the obstacle boundary.

The extracted road region 14B is transmitted to the first collision avoiding space extracting unit 150 and used as information for extracting a global collision avoiding space.

The first collision avoidance path setting unit 150 sets the lane information 13 detected by the lane detecting unit 130 and the fixed obstacle area 14A-1 and the road area 14B extracted from the area extracting unit 140 And removes the fixed obstacle area 14A-1 from the road area 14B and establishes a global collision avoidance path in consideration of the lane information 13 within the removed area.

The second collision avoidance path setting unit 160 sets information such as position information, distance information, size information, and relative speed included in the dynamic obstacle region 14A-2 received from the region extracting unit 140 as a Kalman filter or the like So as to predict the motion of the dynamic obstacle region 14A-2.

The second collision avoidance path setting unit 160 reflects the motion prediction result of the dynamic obstacle region 14A-2 and outputs the local collision avoidance path 15 in the global collision avoidance path 15 received from the first collision avoidance path setting unit 150. [ The avoidance path 16 is set. That is, the second collision avoidance path setting unit 160 removes the motion prediction region of the dynamic obstacle region 14A-2 in the global collision avoidance path 15, and the motion prediction region of the dynamic obstacle region 14A- And sets the removed path to the local collision avoidance path 16. [

As such, the local collision avoidance path 16 can be set through the first collision avoidance path setting unit 150 and the second collision avoidance path setting unit 160, taking into consideration not only the fixed obstacle but also the dynamic obstacle.

The collision prediction and avoidance deciding unit 170 predicts collision based on the local collision avoidance path 16 and the position information, the distance information, and the relative speed information of the obstacle included in the obstacle region 14A extracted by the region extracting unit , And determines a collision avoidance when the collision risk situation is predicted, and transmits a collision avoidance command (17) for instructing collision avoidance to the HMI unit (180) and the active braking / controlling system (200), respectively.

 The HMI unit 180 generates visual or auditory information of a collision risk situation according to the collision avoidance command 17 and alerts the driver of the collision risk situation.

The active braking / control system 200 starts the operation in response to the collision avoidance command 17 from the collision prediction and avoidance decision unit 170. And performs the braking and wheel control of the vehicle so as to run on the local collision avoidance route 16 transmitted from the second collision avoidance route setting unit 160. [

2 is a flowchart showing a method of setting a collision avoidance path for a vehicle according to an embodiment of the present invention.

Referring to FIG. 2, a stereo image of the vehicle front is acquired from a stereo camera installed in the vehicle, and a disparity map is generated from the obtained stereo image (S210).

Then, the obstacle area and the road area are separated from the generated displacement map (S220). Here, the obstacle region can be separated and extracted through the feature point analysis, and the road region can be separately extracted according to the road surface estimation method and the obstacle detection method described above.

Then, the extracted and extracted obstacle region is classified into a fixed obstacle region and a dynamic obstacle region (S230). The classification method can classify the fixed obstacle area and the dynamic obstacle area by analyzing the change rate characteristics of the information such as the height information (size information), the position information, and the distance information of the obstacle that can be obtained from the displacement map. For example, since the fixed obstacle is motionless, the information will exhibit a linear rate of change property on successive frames. On the other hand, since the dynamic obstacle has motion, the information will exhibit a nonlinear rate of change characteristic on successive frames. Thus, it is possible to classify fixed obstacles and dynamic obstacles by analyzing the change rate characteristics of information that can be acquired on a displacement map in successive frames.

Next, a global collision avoidance area in which the classified fixed obstacle area is removed in the road area is set (S240).

Next, a local collision avoidance region in which the classified dynamic obstacle region is removed in the set global collision avoidance region is set (S250), and the set local collision avoidance region is set as a final collision avoidance path, The procedures of FIG. Here, if the dynamic obstacle region is not identified in the process of classifying the obstacle region into the fixed obstacle region and the dynamic obstacle region, that is, if the dynamic obstacle region is not identified, i.e., the height information (size information) If the nonlinear characteristic is not confirmed in the process of analyzing the rate of change of information, the global collision avoiding region is set as the final collision avoiding path.

The set final collision avoidance path is transmitted to the active braking / controlling system provided in the vehicle, as described above, to perform the vehicle braking and the wheel control so as to drive the vehicle in the active braking / And the like. At the same time, the HMI provides the driver with a collision risk situation in the form of visual or auditory information.

As described above, the displacement map is generated from the stereo image (or the left and right images) inputted from the stereo camera, and the fixed obstacle area and the dynamic obstacle area are grasped based on the information that can be acquired on the generated displacement map. Then, a global collision avoiding space is set in consideration of the identified fixed obstacle area, a local collision avoiding space is set in the global collision avoiding space set in consideration of the dynamic obstacle area, It is possible to predict the collision avoiding path, thereby enabling the safe driving of the vehicle.

The foregoing description is merely illustrative of the technical idea of the present invention, and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments of the present invention are not intended to limit the scope of the present invention but to limit the scope of the present invention. The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents thereof should be construed as being included in the scope of the present invention.

Claims (1)

A collision avoidance path setting method of a vehicle for avoiding a collision with an obstacle,
Generating a displacement map from the stereo image;
Separating an obstacle area and a road area using information obtainable from the displacement map;
Classifying the separated obstacle region into a fixed obstacle region including a motionless obstacle and a dynamic obstacle region including a motion obstacle;
Setting a global collision avoidance area in which the fixed obstacle area is removed from the road area;
Setting a local collision avoidance area in which the dynamic obstacle area is removed in the global collision avoidance area; And
Setting a collision avoiding path within the set local collision avoiding area
The collision avoidance path setting method comprising:

Images