CN-114994635-B - Intelligent driving drivable area detection method and device

CN114994635BCN 114994635 BCN114994635 BCN 114994635BCN-114994635-B

Abstract

The application provides an intelligent driving drivable region detection method and device, the method comprises the steps of obtaining point cloud data under the same time stamp of a plurality of laser radars, carrying out fusion processing on multi-frame point cloud data to obtain fusion point cloud sets, determining non-ground point cloud data from the fusion point cloud sets according to z-direction coordinate values of the fusion point cloud data contained in the fusion point cloud sets to obtain non-ground point cloud sets composed of the determined non-ground point cloud data, mapping the non-ground point cloud sets to a pre-established target polar coordinate system, determining nearest distances between obstacles at all angles around a vehicle and the vehicle according to the non-ground point cloud sets at the target polar coordinate system, and determining drivable regions of the vehicle according to the nearest distances between the obstacles at all angles around the vehicle and the vehicle. The application adopts a plurality of laser radars to collect point cloud data, the detection distance is greatly improved, the environmental adaptability is better, and the autonomous detection performance is high because the condition assistance of a high-precision map, a structured road and the like is not needed.

Inventors

HUANG WENJIN

Assignees

上海涵润汽车电子有限公司

Dates

Publication Date: 20260508
Application Date: 20220518

Claims (8)

1. An intelligent driving drivable area detection method is characterized by comprising the following steps: Acquiring point cloud data under the same time stamp of a plurality of laser radars, wherein the point cloud data are acquired by the laser radars installed on a vehicle, and the point cloud data are all point cloud data transformed into a vehicle coordinate system; carrying out fusion processing on the multi-frame point cloud data to obtain a fusion point cloud set; determining non-ground point cloud data from the fused point cloud set according to z-direction coordinate values of the fused point cloud data contained in the fused point cloud set to obtain a non-ground point cloud set composed of the determined non-ground point cloud data; Mapping the non-ground point cloud set to a pre-established target polar coordinate system to determine the nearest distances between the vehicle and the obstacles at all angles around the vehicle according to the non-ground point cloud set in the target polar coordinate system, wherein the target polar coordinate system is a polar coordinate system established by taking the origin of the vehicle coordinate system as a pole; determining a drivable region of the vehicle according to the nearest distances between the obstacles and the vehicle respectively under the full angle around the vehicle, comprising: mapping each non-ground point cloud data into a corresponding grid contained in a grid map according to an x-direction coordinate value and a y-direction coordinate value of each non-ground point cloud data in the non-ground point cloud set and a preset grid resolution, wherein the grid resolution is used for dividing the grid map into a plurality of grids; Determining non-ground point cloud data contained in the grid map as low-point data or high-point data according to a target fitting plane equation and a preset second distance threshold, wherein the target fitting plane equation is a fitting plane equation determined in the last iteration; Determining the mixed occupation probability corresponding to each grid according to the low-point data and the high-point data contained in each grid in the grid map, the low-point data contained in the adjacent grids, the preset grid occupation probability and the target fitting plane equation; Calculating the distance between each grid contained in the grid map and the center of the vehicle rear axle based on the coordinate transformation relation between the center of the vehicle rear axle and the origin of the grid map, and determining the initial occupation probability corresponding to each grid according to the calculated distance and the nearest distance between the obstacle and the vehicle under the corresponding angle of each grid, wherein the initial occupation probability corresponding to each grid is calculated according to the distance between each grid and the center of the vehicle rear axle and the nearest distance corresponding to each grid by adopting the following formula: in the formula, Refers to the initial occupancy probability corresponding to the grid, For each grid to vehicle rear axle center distance, The nearest distance corresponding to each grid; Determining the probability that each grid contained in the grid map contains an obstacle according to the mixed occupation probability and the initial occupation probability corresponding to each grid contained in the grid map; And determining the drivable area of the vehicle according to the probability that each grid in the grid map contains the obstacle.
2. The intelligent driving drivable area detection method as set forth in claim 1, wherein the fusing the multi-frame point cloud data to obtain a fused point cloud set comprises: Splicing the multi-frame point cloud data into one frame of point cloud data according to a preset laser radar sequence to obtain the fusion point cloud set; Or alternatively, the first and second heat exchangers may be, Calculating a horizontal angle and a vertical angle corresponding to the point cloud data aiming at each point cloud data contained in the multi-frame point cloud data, and filling the point cloud data into a corresponding point cloud grid according to the horizontal angle and the vertical angle corresponding to the point cloud data; And for each point cloud grid, if the point cloud grid contains a plurality of point cloud data, taking the gravity centers of the plurality of point cloud data as the fusion point cloud data under the point cloud grid to obtain a fusion point cloud set consisting of the fusion point cloud data under all the point cloud grids.
3. The intelligent driving drivable region detection method as set forth in claim 1, wherein the determining non-ground point cloud data from the fusion point cloud set according to z-coordinate values of the fusion point cloud data contained in the fusion point cloud set comprises: Removing fusion point cloud data in which abnormal z-direction coordinate values in the fusion point cloud set are located, and taking the fusion point cloud set after abnormal data are removed as a target fusion point cloud set; Determining ground point cloud data from the target fusion point cloud set according to z-direction coordinate values of fusion point cloud data contained in the target fusion point cloud set, so as to obtain a to-be-updated ground point cloud set composed of the determined ground point cloud data; determining a fitting plane equation according to the ground point cloud set to be updated; Calculating the distance between each fusion point cloud data in the target fusion point cloud set and the fitting plane equation, determining fusion point cloud data with the calculated distance smaller than a preset first distance threshold value as ground point cloud data, and taking the determined ground point cloud data as updated ground point cloud; And adding 1 to the updating iteration number, judging whether the updating iteration number added with 1 reaches the preset iteration total number, if not, taking the updated ground point cloud set as a ground point cloud set to be updated, and returning to execute the step of determining a fitting plane equation according to the ground point cloud set to be updated until the updating iteration number reaches the iteration total number, and taking other point cloud data except the updated ground point cloud set in the target fusion point cloud set as non-ground point cloud data, wherein the initial updating iteration number is 0.
4. The intelligent driving drivable region detection method as set forth in claim 3, wherein the determining the ground point cloud data from the target fusion point cloud set according to the z-coordinate values of the fusion point cloud data contained in the target fusion point cloud set comprises: selecting a preset number of fusion point cloud data from the target fusion point cloud set according to the z coordinate value of the fusion point cloud data contained in the target fusion point cloud set, and calculating the z average value of the preset number of fusion point cloud data; and aiming at each fusion point cloud data in the target fusion point cloud set, if the difference value between the z-direction coordinate value of the fusion point cloud data and the z-direction average value is smaller than a preset difference value threshold value, determining that the fusion point cloud data is ground point cloud data.
5. The intelligent driving drivable area detection method as set forth in claim 3, wherein determining a fitted plane equation from the to-be-updated ground point cloud comprises: Calculating an x-direction average value, a y-direction average value and a z-direction average value of the ground point cloud set to be updated; Determining a covariance matrix according to the x-direction average value, the y-direction average value and the z-direction average value of the ground point cloud set to be updated, and solving a plurality of eigenvalues and eigenvectors of the covariance matrix; Determining a minimum characteristic value from the plurality of characteristic values, and taking a characteristic vector corresponding to the minimum characteristic value as a normal vector of a fitting plane; and solving the fitting plane equation according to the to-be-updated ground point cloud set and the normal vector of the fitting plane.
6. The intelligent driving drivable region detection method as set forth in claim 3, wherein the target polar coordinate system comprises a plurality of angle grids, each angle grid comprising a plurality of radial grids; the mapping the non-ground point cloud set to a pre-established target polar coordinate system to determine the nearest distances between the obstacles and the vehicle in all angles around the vehicle according to the non-ground point cloud set in the target polar coordinate system, including: According to the x-direction coordinate value and the y-direction coordinate value of each non-ground point cloud datum in the non-ground point cloud set, solving a horizontal angle and a two-dimensional projection distance corresponding to each non-ground point cloud datum; mapping each non-ground point cloud data to the target polar coordinate system according to the horizontal angle and the two-dimensional projection distance corresponding to each non-ground point cloud data in the non-ground point cloud set to obtain a non-ground point cloud set in the target polar coordinate system; And for each angle grid included in the target polar coordinate system, determining a non-empty radial grid closest to a pole from the angle grids, calculating the distance average value of non-ground point cloud data included in the non-empty radial grid and the pole as the closest distance between the obstacle and the vehicle under the angle corresponding to the angle grid, and obtaining the closest distances between the obstacle and the vehicle under the all-angle surrounding of the vehicle.
7. The intelligent driving drivable area detecting method as set forth in claim 1, wherein the determining the hybrid occupancy probability corresponding to each grid based on the low-point data and the high-point data contained in each grid in the grid map, and the low-point data contained in the neighboring grids, the preset grid occupancy probability, and the target fitting plane equation comprises: Calculating the number of high-point data contained in each grid in the grid map, and determining the high-point occupation probability corresponding to each grid in the grid map according to the number and the preset grid occupation probability; Determining the low point occupation probability corresponding to each grid in the grid map according to the low point data contained in each grid in the grid map, the low point data contained in the adjacent grids and the target fitting plane equation; and determining the mixed occupation probability corresponding to each grid according to the high point occupation probability and the low point occupation probability corresponding to each grid in the grid map.
8. An intelligent driving drivable area detection apparatus, comprising: The system comprises a point cloud data acquisition module, a point cloud data acquisition module and a point cloud data processing module, wherein the point cloud data acquisition module is used for acquiring point cloud data under the same time stamp of a plurality of laser radars, the point cloud data are acquired by the laser radars arranged on a vehicle, and the point cloud data are all point cloud data transformed to a vehicle coordinate system; The point cloud data fusion module is used for carrying out fusion processing on the multi-frame point cloud data to obtain a fusion point cloud set; The non-ground point cloud set determining module is used for determining non-ground point cloud data from the fusion point cloud set according to the z-coordinate value of the fusion point cloud data contained in the fusion point cloud set so as to obtain a non-ground point cloud set composed of the determined non-ground point cloud data; The nearest distance determining module is used for mapping the non-ground point cloud set to a pre-established target polar coordinate system so as to determine the nearest distances between the obstacles at all angles around the vehicle and the vehicle respectively according to the non-ground point cloud set in the target polar coordinate system, wherein the target polar coordinate system is a polar coordinate system established by taking the origin of the vehicle coordinate system as a pole; the drivable area determining module is used for determining the drivable area of the vehicle according to the nearest distances between the obstacles and the vehicle under the full angle around the vehicle; The drivable region determining module is specifically configured to: mapping each non-ground point cloud data into a corresponding grid contained in a grid map according to an x-direction coordinate value and a y-direction coordinate value of each non-ground point cloud data in the non-ground point cloud set and a preset grid resolution, wherein the grid resolution is used for dividing the grid map into a plurality of grids; Determining non-ground point cloud data contained in the grid map as low-point data or high-point data according to a target fitting plane equation and a preset second distance threshold, wherein the target fitting plane equation is a fitting plane equation determined in the last iteration; Determining the mixed occupation probability corresponding to each grid according to the low-point data and the high-point data contained in each grid in the grid map, the low-point data contained in the adjacent grids, the preset grid occupation probability and the target fitting plane equation; Calculating the distance between each grid contained in the grid map and the center of the vehicle rear axle based on the coordinate transformation relation between the center of the vehicle rear axle and the origin of the grid map, and determining the initial occupation probability corresponding to each grid according to the calculated distance and the nearest distance between the obstacle at the corresponding angle of each grid and the vehicle, wherein the initial occupation probability corresponding to each grid is calculated according to the distance between each grid and the center of the vehicle rear axle and the nearest distance corresponding to each grid by adopting the following formula: in the formula, Refers to the initial occupancy probability corresponding to the grid, For each grid to vehicle rear axle center distance, The nearest distance corresponding to each grid; Determining the probability that each grid contained in the grid map contains an obstacle according to the mixed occupation probability and the initial occupation probability corresponding to each grid contained in the grid map; And determining the drivable area of the vehicle according to the probability that each grid in the grid map contains the obstacle.

Description

Intelligent driving drivable area detection method and device Technical Field The application relates to the technical field of intelligent driving, in particular to a method and a device for detecting an intelligent driving drivable area. Background Intelligent driving drivable region detection is a technique that provides a safety margin for intelligent driving, and detects a region in which a vehicle can run through the surrounding environment of the vehicle. The existing intelligent driving drivable area detection method mainly comprises two methods, wherein one method is to use a camera for target detection of a road type, identify a road to determine a drivable area, and the other method is to use a laser radar and a high-precision map to obtain boundary points to solve the drivable area. However, the method for detecting the road type target by aiming at the camera has the advantages of limited detection distance, high light requirement, generally being applicable to structural scenes and poor environmental adaptability, and the method for detecting the driving area of the laser radar and the high-precision map needs to be periodically maintained and updated. Disclosure of Invention In view of the above, the present application provides a method and apparatus for detecting a drivable area for intelligent driving, which are used for solving the problem of poor environmental adaptability in a method for detecting a road type object by using a camera, and the problem of needing to periodically maintain and update a high-precision map in a method for detecting a drivable area by using a laser radar and a high-precision map, and the technical scheme thereof is as follows: An intelligent driving drivable region detection method includes: Acquiring point cloud data under the same time stamp of a plurality of laser radars, wherein the point cloud data are acquired by the laser radars installed on a vehicle, and the point cloud data are all point cloud data converted into a vehicle coordinate system; carrying out fusion processing on multi-frame point cloud data to obtain a fusion point cloud set; determining non-ground point cloud data from the fused point cloud set according to the z-coordinate value of the fused point cloud data contained in the fused point cloud set to obtain a non-ground point cloud set consisting of the determined non-ground point cloud data; mapping the non-ground point cloud set to a pre-established target polar coordinate system, so as to determine the nearest distances between the vehicle and the obstacles at all angles around the vehicle according to the non-ground point cloud set in the target polar coordinate system, wherein the target polar coordinate system is established by taking the origin of the vehicle coordinate system as a pole; and determining the drivable area of the vehicle according to the nearest distances between the obstacles and the vehicle under the full angle around the vehicle. Optionally, the fusing processing is performed on the multi-frame point cloud data to obtain a fused point cloud set, including: splicing the multi-frame point cloud data into one frame of point cloud data according to a preset laser radar sequence to obtain a fusion point cloud set; Or alternatively, the first and second heat exchangers may be, Calculating a horizontal angle and a vertical angle corresponding to the point cloud data aiming at each point cloud data contained in the multi-frame point cloud data, and filling the point cloud data into a corresponding point cloud grid according to the horizontal angle and the vertical angle corresponding to the point cloud data; And for each point cloud grid, if the point cloud grid contains a plurality of point cloud data, taking the gravity centers of the plurality of point cloud data as the fusion point cloud data under the point cloud grid to obtain a fusion point cloud set consisting of the fusion point cloud data under all the point cloud grids. Optionally, determining non-ground point cloud data from the fused point cloud set according to z-coordinate values of the fused point cloud data included in the fused point cloud set includes: removing fusion point cloud data in which abnormal z-direction coordinate values in the fusion point cloud set are located, and taking the fusion point cloud set after abnormal data are removed as a target fusion point cloud set; determining ground point cloud data from the target fusion point cloud set according to the z-direction coordinate value of the fusion point cloud data contained in the target fusion point cloud set, so as to obtain a to-be-updated ground point cloud set composed of the determined ground point cloud data; determining a fitting plane equation according to the ground point cloud set to be updated; Calculating the distance between each fusion point cloud data in the target fusion point cloud set and the fitting plane equation, determining fusion point cloud data with the calculate