External parameter calibration method, device and related equipment of laser radar

Abstract

The present application discloses a method, device, and electronic device for calibrating the external parameters of a laser radar. The method includes: obtaining laser point cloud data through a vehicle's laser radar scan; obtaining the angle between the vehicle and a preset lane line on the lane, and a preset number of coordinate points located on the preset lane line in the laser point cloud data; obtaining a road area plane equation based on a road plane fitting point set; determining a transformation matrix based on the road area plane equation; and determining the external parameters of the laser radar based on the angle between the vehicle and the preset lane line, a preset number of coordinate points located on the preset lane line in the laser point cloud data, and the transformation matrix. The external parameters of the laser radar are calibrated using the angle between the vehicle's position where the laser radar is located and the lane line, as well as the coordinates on the lane line. This calibration method does not require calibration tools, reduces calibration hardware costs, and achieves rapid calibration of the laser radar.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and more specifically, to a method, device, electronic device, and storage medium for calibrating external parameters of a laser radar. Background Art

[0002] In recent years, the intelligentization of vehicles has become a trend, necessitating the emergence of various onboard sensors. LiDAR-based environmental perception technology has garnered widespread attention in the autonomous driving field. LiDAR calibration is fundamental to its perception performance, primarily transforming the LiDAR's coordinate system into the vehicle's coordinate system. For users, LiDAR calibration primarily involves extrinsic calibration, while intrinsic parameters are typically provided by the LiDAR manufacturer. Extrinsic calibration primarily includes coordinate system rotation parameters (yaw, pitch, and roll angles), as well as coordinate system translation parameters (X, Y, and Z axis translations).

[0003] In related technologies, the calibration of the external parameters of the lidar mainly relies on manually placed targets (such as cardboard boxes, cones, calibration rods, etc.), and the calibration efficiency is not high. Summary of the Invention

[0004] In view of the problems raised in the background technology, the present application proposes a method, device and electronic equipment for calibrating the external parameters of a laser radar.

[0005] In a first aspect, an embodiment of the present application provides a method for calibrating the external parameters of a laser radar, the method comprising: acquiring laser point cloud data through a laser radar scan of a vehicle; acquiring the angle between the vehicle and a preset side lane line of the lane where the vehicle is located, and a preset number of coordinate points in the laser point cloud data located on the preset side lane line, where the preset side is the left or right side; determining a transformation matrix of the road plane where the vehicle is located from the laser radar coordinate system to the vehicle body coordinate system based on the laser point cloud data; determining the external parameters of the laser radar according to the angle between the vehicle and the preset side lane line, the preset number of coordinate points, and the transformation matrix.

[0006] In some embodiments, determining the transformation matrix of the road plane of the vehicle from the laser radar coordinate system to the vehicle body coordinate system based on the laser point cloud data includes: determining a road plane fitting point set of the road on which the vehicle is located based on the laser point cloud data; fitting the lane in the vehicle body coordinate system based on the road plane fitting point set to obtain a road area plane equation in the laser radar coordinate system, wherein the road area plane equation is the plane corresponding to the road plane fitting point set in the vehicle body coordinate system; and determining the transformation matrix based on the road area plane equation.

[0007] In some embodiments, determining the road plane fitting point set of the road where the vehicle is located based on the laser point cloud data includes: projecting the laser point cloud data into a two-dimensional grid map to obtain a two-dimensional point cloud grid map; screening candidate grids from the two-dimensional point cloud grid map based on the local point cloud height difference between each grid in the two-dimensional point cloud grid map; segmenting and extracting the candidate grids to obtain an obstacle point cloud set and a non-obstacle point cloud set; searching and processing the obstacle point cloud set to obtain a road boundary point cloud set in the obstacle point cloud set; and obtaining the road area point cloud set based on the non-obstacle point cloud set and the road boundary point cloud set.

[0008] In some specific embodiments, the method of screening out candidate grids from the two-dimensional point cloud grid image based on the local point cloud height difference between each grid in the two-dimensional point cloud grid image includes: traversing the grids in the two-dimensional point cloud grid image, taking the currently traversed grid as the first central grid, and obtaining the maximum height point cloud value based on the coordinates of the point cloud data in the first central grid; taking the first central grid as the center, calculating the height difference between the maximum height point cloud value and the minimum height point cloud value corresponding to each grid within a preset size range, wherein the minimum height point cloud value is calculated based on the coordinates of the point cloud data in the grid; if the height difference is higher than the first preset height difference, adding 1 to the count value; within the preset size range, if the count value exceeds a pre-designed numerical threshold, marking the first central grid as a candidate grid.

[0009] In some specific embodiments, the segmentation and extraction of the candidate grid to obtain an obstacle point cloud set and a non-obstacle point cloud set includes: traversing the laser point cloud data, and if the currently traversed laser point cloud data is in the candidate grid, using the laser point cloud data as the center point cloud, and calculating the slope of the point cloud data within a first preset radius and the center point cloud; if the slope is greater than a preset slope threshold, the slope count is increased by 1; if the height difference between the point cloud heights of the laser point cloud data greater than the preset slope threshold and the center point cloud exceeds a second preset height difference, the height count is increased by 1; if the slope count corresponding to the laser point cloud data within the first preset radius of the center point cloud is not less than the slope count threshold and the height count is not less than the height count threshold, determining that the center point cloud is an obstacle point cloud.

[0010] Specifically, the segmentation and extraction of the candidate grid to obtain an obstacle point cloud set and a non-obstacle point cloud set also includes: taking the candidate grid where the central point cloud is located as a second central grid, and determining that the point cloud data in the candidate grid within a second preset radius whose point cloud height is higher than that of the central point cloud is an obstacle point cloud.

[0011] In some specific embodiments, searching and processing the obstacle point cloud set to obtain a road boundary point cloud set in the obstacle point cloud set includes: projecting the obstacle point cloud set into a grid map to obtain an obstacle grid; performing a row-by-row search for the obstacle point cloud in each obstacle grid, using the lateral center of the vehicle as the search center and a third preset radius as the search radius, and obtaining the road boundary point cloud set in the obstacle point cloud set from the search road boundary point clouds within a search range corresponding to the search radius.

[0012] In some embodiments, the conversion matrix includes: a roll angle coordinate transformation matrix corresponding to the roll angle of the laser radar, a pitch angle coordinate transformation matrix corresponding to the pitch angle, and a Z-axis translation parameter coordinate transformation matrix corresponding to the Z-axis translation parameter; determining the external parameters of the laser radar based on the angle between the vehicle and the preset side lane line, the preset number of coordinate points and the conversion matrix, includes: transforming a preset number of coordinate points on the preset side lane line according to the roll angle coordinate transformation matrix, the pitch angle coordinate transformation matrix and the Z-axis translation parameter coordinate transformation matrix to obtain transformed coordinate points; obtaining the straight line equation corresponding to the transformed coordinate point; obtaining the angle between the straight line equation and the X-axis according to the straight line equation; determining the yaw angle of the laser radar according to the angle between the vehicle and the preset side lane line and the angle between the straight line equation and the X-axis.

[0013] In a second aspect, an embodiment of the present application provides a laser radar external parameter calibration device, the device comprising: a laser point cloud data acquisition module, for acquiring laser point cloud data through a laser radar scan of a vehicle; a lane line data acquisition module, for acquiring the angle between the vehicle and a preset side lane line of the lane where the vehicle is located, and a preset number of coordinate points in the laser point cloud data located on the preset side lane line, where the preset side is the left or right side; a transformation matrix determination module, for determining the transformation matrix of the road plane where the vehicle is located from the laser radar coordinate system to the vehicle body coordinate system based on the laser point cloud data; an external parameter determination module, for determining the external parameters of the laser radar based on the angle between the vehicle and the preset side lane line, the preset number of coordinate points and the transformation matrix.

[0014] In a third aspect, an embodiment of the present application provides an electronic device comprising: one or more processors; a memory; and one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the external parameter calibration method of the laser radar provided in the first aspect above.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which program code is stored. The program code can be called by a processor to execute the external parameter calibration method of the laser radar provided in the first aspect above.

[0016] The solution provided in this application calibrates the external parameters of the laser radar through the angle between the current lane position of the vehicle where the laser radar is located and the lane line, as well as multiple coordinate points on the lane line. This calibration method does not require additional calibration tools, reducing the hardware cost during the calibration process. Moreover, it is only necessary to obtain the angle between the vehicle and the lane line and the coordinates on the lane line to achieve rapid calibration of the laser radar, shortening the calibration time. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 A flow chart of a method for calibrating external parameters of a laser radar provided in an embodiment of the present application is shown.

[0019] Figure 2 A detailed flow chart of step S130 of calibrating the external parameters of a laser radar according to an embodiment of the present application is shown.

[0020] Figure 3 A grid diagram schematically illustrates a method for calibrating external parameters of a laser radar provided in an embodiment of the present application.

[0021] Figure 4 A detailed flow chart of step S132 of calibrating the external parameters of the laser radar provided in one embodiment of the present application is shown.

[0022] Figure 5 A detailed flowchart of step S1322 of calibrating the external parameters of the laser radar provided in an embodiment of the present application is shown.

[0023] Figure 6 A detailed flowchart of step S1323 of calibrating the external parameters of the laser radar provided in one embodiment of the present application is shown.

[0024] Figure 7 A schematic diagram of an obstacle point cloud for calibrating external parameters of a laser radar provided in an embodiment of the present application is shown.

[0025] Figure 8A schematic diagram of a road boundary point cloud search using external parameter calibration of a laser radar provided in an embodiment of the present application is shown.

[0026] Figure 9 A road boundary area fitting diagram for calibrating external parameters of a laser radar provided in an embodiment of the present application is shown.

[0027] Figure 10 A detailed flowchart of step S160 of calibrating the external parameters of a laser radar provided in an embodiment of the present application is shown.

[0028] Figure 11 The figure shows a structural block diagram of the external parameter calibration device of the laser radar provided in an embodiment of the present application.

[0029] Figure 12 A structural block diagram of an electronic device provided in an embodiment of the present application for executing the external parameter calibration method of a laser radar according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0031] In related technologies, there are two commonly used methods for calibrating the external parameters of lidar: one is a joint calibration method based on lidar and camera; the other is a method of directly placing targets (such as cardboard boxes, cones, calibration rods, etc.) and inverting the calibration matrix through measurement.

[0032] Joint calibration of the lidar and camera: First, complete the calibration of the camera to obtain the internal and external parameters of the camera. Then, match the laser point cloud with the target pixel points in the camera image coordinate system to obtain the rotation and translation matrix of the lidar relative to the camera space, completing the lidar external parameter calibration.

[0033] Direct solution of LiDAR calibration parameters: Place multiple targets (such as cartons, cones, calibration poles, etc.) in an orderly manner on the site, measure and obtain the coordinates of the targets in the vehicle coordinate system, and simultaneously obtain the coordinates of each target in the LiDAR coordinate system. Associate the points in the two coordinate systems, inversely calculate the rotation and translation matrices, and complete the calibration of the LiDAR external parameters.

[0034] The inventors have found that commonly used calibration methods are highly dependent on manual labor and other calibration tools, and the calibration efficiency is low.

[0035] In response to the above-mentioned problems and the technical problems raised by the background technology, the embodiments of the present application propose a method, device and electronic device for calibrating the external parameters of the laser radar. The external parameters of the laser radar are calibrated by the angle between the current lane position of the vehicle where the laser radar is located and the lane line, as well as multiple coordinate points on the lane line. This calibration method does not require additional calibration tools, reducing the hardware cost during the calibration process. Moreover, it is only necessary to obtain the angle between the vehicle and the lane line and the coordinates on the lane line to achieve rapid calibration of the laser radar, shortening the calibration time.

[0036] See also Figure 1 , Figure 1 The figure shows a flow chart of the external parameter calibration method of the laser radar provided by an embodiment of the present application. In a specific embodiment, the external parameter calibration method of the laser radar in this embodiment is applied to a server, and the external parameter calibration method of the laser radar is applied to Figure 10 The external parameter calibration device 300 of the laser radar and the electronic device 100 equipped with the external parameter calibration device 300 of the laser radar are shown.

[0037] The following will target Figure 1 The process shown is described in detail. The external parameter calibration method of the laser radar may specifically include the following steps:

[0038] Step S110: Obtain laser point cloud data through laser radar scanning of the vehicle.

[0039] LiDAR is a radar system that uses laser beams to detect the position, speed and other characteristic quantities of a target.

[0040] One or more LiDAR sensors are installed on the vehicle, which collect laser point cloud data composed of multiple feature quantities within a preset range. The preset range can be centered around the LiDAR and based on the LiDAR's maximum scanning distance, or it can be set based on actual user needs and is not limited here.

[0041] Step S120: Obtain the angle between the vehicle and a preset side lane line of the lane where the vehicle is located, and a preset number of coordinate points located on the preset side lane line in the laser point cloud data, where the preset side is the left side or the right side.

[0042] The angle is calculated by measuring the distance between the front and rear wheels of the vehicle and the preset side lane line, denoted as FL and RL respectively. The angle between the vehicle and the preset side lane line is: Where D is the wheelbase of the vehicle and the width of the vehicle is recorded as W. The angle α between the vehicle and the preset side lane line is calculated according to formula (1).

[0043] In this embodiment of the present application, the vehicle's lane can be a multi-lane road, which is at least a two-lane road. Considering the ground distance that current LiDAR can collectively scan, the selected lane length can be 35 meters. The lane can be three lanes or fewer, but the lane markings must be clear and unworn to facilitate intensity differentiation of the laser point cloud data. The vehicle's stopping direction should be consistent with the overall lane marking direction, but does not need to be strictly parallel to the lane.

[0044] A radar coordinate system is established based on the current position of the vehicle, and the coordinate points on the preset side lane line are coordinate points obtained based on the radar coordinate system.

[0045] In some embodiments, the preset side lane marking is the lane marking on the left side of the vehicle. Using the laser point cloud data corresponding to the left lane marking, a preset number of coordinate points is obtained. The preset number can be 2, 4, or 5, and the number of coordinate points selected is not specifically limited herein. In the embodiment of the present application, the preset number is preferably 2.

[0046] Step S130: Determine a transformation matrix of the road plane where the vehicle is located from the laser radar coordinate system to the vehicle body coordinate system based on the laser point cloud data.

[0047] The laser point cloud data is fitted using a fitting algorithm to obtain the road plane of the vehicle in the LiDAR coordinate system. The road plane in the LiDAR coordinate system is converted to the road plane in the vehicle coordinate system using a preset conversion method. The conversion matrix is ​​obtained during the conversion process from the LiDAR coordinate system to the vehicle coordinate system.

[0048] See also Figure 2 Determining the transformation matrix of the road plane where the vehicle is located from the laser radar coordinate system to the vehicle body coordinate system based on the laser point cloud data specifically includes steps S132 to S136, which are detailed as follows:

[0049] Step S132: Determine a road plane fitting point set of the road where the vehicle is located based on the laser point cloud data.

[0050] The vehicle's LiDAR detects signals from the target and then processes the received signal reflected from the vehicle's lane with the transmitted signal to obtain laser point cloud data corresponding to the vehicle's lane. Feature extraction from the laser point cloud data is performed using a preset method to determine the road plane fitting point set corresponding to the vehicle's lane.

[0051] The vehicle's lane includes, but is not limited to, road areas, obstacle areas, and non-obstacle areas. The road area includes the road surface, the obstacle area includes trees and public facilities along the lane, and the non-obstacle area includes the road boundary. Based on the laser point cloud data corresponding to the vehicle's lane, the laser point cloud data received by the lidar can be divided into a road area point cloud set, an obstacle point cloud set, and a non-obstacle point cloud set.

[0052] In some embodiments, feature extraction can be performed on laser point cloud data using a raster image. A raster image is a method of converting image data into image data consisting of multiple rectangular grids. Figure 3 , Figure 3 The grid diagram in the embodiment of the present application is shown. The grid size can be 201×101, with the origin at the upper left corner (0,0). The pixel point (50,200) in the image is the origin of the LiDAR coordinates. The actual geometric distance represented by each pixel is 20 cm, so the vertical range is approximately 40 meters and the horizontal range is 10 meters on each side.

[0053] The laser point cloud data acquired by the lidar relies on the raster map for feature extraction to obtain a road plane fitting point set consisting of a non-obstacle point cloud set and a road area point cloud set.

[0054] See also Figure 4 Extracting the laser point cloud data based on the grid map to obtain a road plane fitting point set specifically includes steps S1321 to S1325, which are detailed as follows:

[0055] Step S1321: Project the laser point cloud data into a two-dimensional grid map to obtain a two-dimensional point cloud grid map.

[0056] Project the laser point cloud data onto a grid map to obtain a 2D point cloud grid map. The grid size in the 2D point cloud grid map can be a preset grid size, and the size of the 2D point cloud grid map can be enlarged or reduced according to user needs. By enlarging or reducing the size, 2D point cloud grid maps can be obtained according to different needs. The enlargement and reduction methods are not limited here.

[0057] The point clouds of each grid are sorted from small to large according to their height, and the number of point clouds, minimum height value, maximum height value and height difference of the grid are counted.

[0058] Step S1322: selecting candidate grids from the two-dimensional point cloud grid image according to the local point cloud height differences between the grids in the two-dimensional point cloud grid image.

[0059] The local point cloud height difference can be obtained by comparing the point cloud height corresponding to the central grid with the point cloud heights of other grids within a preset range. Alternatively, a grid in the two-dimensional point cloud grid image is first determined as a reference grid, and the point cloud height corresponding to the reference grid is obtained. The point cloud height difference of the remaining point cloud grids within the preset range is calculated, and the point cloud height difference is used as the local height difference. Alternatively, the point cloud height corresponding to each grid in the two-dimensional point cloud grid image is obtained, and each grid is used as a reference grid. The point cloud height difference between each grid within the preset range and the reference grid is obtained, and the point cloud height difference is used as the local height difference.

[0060] Candidate grids are filtered according to the local point cloud height difference. The filtering condition can be that the local point cloud height difference is greater than the local point cloud height difference, or the local point cloud height difference is not greater than the local point cloud height difference. The filtering condition can also be based on the specific filtering conditions set by the user to filter the candidate grids in the two-dimensional point cloud grid map. The filtering method and filtering conditions are not specifically limited here.

[0061] For details, please refer to Figure 5 , Figure 5 The specific steps of screening the candidate grids in the two-dimensional point cloud grid map in the embodiment of the present application are shown, including steps S13221 to S13224, which are described in detail as follows:

[0062] Step S13221: traverse the grids in the two-dimensional point cloud grid map, use the currently traversed grid as the first central grid, and obtain the maximum height point cloud value according to the coordinates of the point cloud data in the first central grid.

[0063] The way to traverse the grid in the two-dimensional point cloud grid map can be from left to right, from top to bottom, or from the center of the two-dimensional point cloud grid map as the starting point to the surrounding area. The way to traverse the two-dimensional point cloud grid map is not specifically limited here.

[0064] If there is at least one point cloud data point in the currently traversed grid, the height point cloud value is obtained based on the coordinates of the point cloud data in the radar coordinate system. If there is only one point cloud data point in the grid, the height point cloud value corresponding to the point cloud data point is used as the maximum height point cloud value. If there is more than one point cloud data point in the grid, the height point cloud value corresponding to each point cloud data point is calculated separately to obtain the maximum height point cloud value in the grid.

[0065] Step S13222: Taking the first central grid as the center, calculate the height difference between the maximum height point cloud value and the minimum height point cloud value corresponding to each grid within a preset size range, and the minimum height point cloud value is calculated based on the coordinates of the point cloud data in the grid.

[0066] With the first central grid as the center, a rectangular range of preset dimensions is used to obtain the corresponding minimum height point cloud value for all grids within the rectangular range, excluding the first central grid. The minimum height point cloud value is obtained in the same way as the maximum height point cloud value, so this is not repeated here. The side length of the rectangular range can be 10cm to 20cm, or it can be set by the user. The preset dimensions are not specifically limited here.

[0067] Step S13223: If the height difference is higher than the first preset height difference, add 1 to the count value.

[0068] Step S13224: within the range of the preset size, if the count value exceeds a pre-designed numerical threshold, the first central grid is marked as a candidate grid.

[0069] Within the preset size range, when the height difference between the minimum height point cloud value corresponding to each grid except the first central grid and the maximum height point cloud value corresponding to the first central grid is higher than the first preset height difference, the count value is increased by 1. When the count value exceeds the pre-designed numerical threshold, the first central grid is marked as a candidate grid. The first preset height difference is set by the user. In the present application, the first preset height difference can be set to 10cm to 20cm. The pre-designed numerical threshold is related to the preset size. The pre-designed numerical threshold is preferably 0.25 times the preset size. The pre-designed numerical threshold can also be specifically set by the user and is not limited here.

[0070] The road segmentation method based on local cumulative height difference greatly reduces the interference of noise by comparing the laser point cloud in the surrounding area grid multiple times, and has strong robustness.

[0071] In the specific implementation process, when traversing the two-dimensional point cloud grid map, for a certain grid, first obtain the maximum height point cloud value Z of the grid max [i, j] (i, j are the coordinates in the point cloud radar coordinate system), with the grid as the center, calculate Z in the m×n area max [i,j] corresponds to the minimum height point cloud value Z of each grid min The height difference between [i',j'], if Z max [i,j]-Z min [i',j']>Threshold1, then the count value N is counted once, if the cumulative count of N exceeds the threshold N min , then the grid is marked as a candidate segmentation grid. Wherein, Threshold1 is the first preset height difference.

[0072] Step S1323: Segment and extract the candidate grid to obtain an obstacle point cloud set and a non-obstacle point cloud set.

[0073] Obstacle point cloud sets include, but are not limited to, tree point cloud sets, public facility point cloud sets, and other non-road point cloud sets. Non-obstacle point cloud sets include road boundary point cloud sets.

[0074] The point cloud data in the candidate grid is segmented by segmentation to obtain obstacle point cloud sets and non-obstacle point cloud sets.

[0075] See also Figure 6 , Figure 6 The specific steps of segmenting and extracting the candidate grid to obtain the obstacle point cloud set and the non-obstacle point cloud set in the embodiment of the present application are shown. Step S1323 includes steps S13231 to S13234, which are described in detail as follows:

[0076] Step S13231: traverse the laser point cloud data. If the currently traversed laser point cloud data is in the candidate grid, use the laser point cloud data as the center point cloud and calculate the slope of the point cloud data within a first preset radius and the center point cloud.

[0077] Calculate the slope between the central point cloud and all other laser point cloud data within a first preset radius, centered on the central point cloud. The slope is determined by the height difference and distance between the central point cloud and the laser point cloud data within the first preset radius.

[0078] The calculation formula of slope K is: abs(ΔZ ij / D ij )(2)

[0079] where ΔZ ij D is the height difference between the center point cloud and a point cloud data within the first preset radius. ij The distance between the center point cloud and a point cloud data within the first preset radius.

[0080] The slope between the two point cloud data is calculated according to formula (2). The first preset radius can be set according to the specific needs of the user. In the present application, the first preset radius is preferably 5 cm.

[0081] Step S13232: If the slope is greater than the preset slope threshold, the slope count is increased by 1.

[0082] Step S13233: If the height difference between the laser point cloud data greater than the preset slope threshold and the center point cloud exceeds a second preset height difference, the height count is incremented by 1.

[0083] Within the preset range, the slope between the center point cloud and other point cloud data is calculated, and when the slope is greater than the preset slope threshold, the slope count is incremented.

[0084] The point cloud data with a slope greater than a preset threshold within a preset range is further calculated to determine whether the height difference between the laser point cloud data and the center point cloud is greater than a second preset height difference. If it is greater than the second preset height difference, the height count is incremented.

[0085] Step S13234: If the slope count corresponding to the laser point cloud data within the first preset radius range of the central point cloud is not less than the slope count threshold and the height count is not less than the height count threshold, the central point cloud is determined to be an obstacle point cloud.

[0086] Whether the central point cloud is an obstacle point cloud is comprehensively judged according to the slope count and height count corresponding to the central point cloud, thereby avoiding the randomness of determining the obstacle point cloud and improving the accuracy of determining the obstacle point cloud.

[0087] In the specific implementation process, a certain laser point cloud data in a certain grid is determined as the center point cloud from the candidate grids, and the slope K=abs(ΔZ ij ) / (D ij ), where ΔZ ij is the height difference between the two laser point cloud data, D ij is the distance between two laser point cloud data. If K exceeds the set slope threshold K min , that is, K>=K min , then the slope count K_Count is increased by 1, and the point cloud height that meets the slope threshold calculation is stored. If the difference between the stored point cloud heights exceeds the threshold Threshold2, the height difference count H_Count is increased by 1. If K_Count>=TK min &&H_Count>=TH min , then the point cloud is an obstacle point cloud.

[0088] In other embodiments, segmenting and extracting the candidate grid to obtain an obstacle point cloud set and a non-obstacle point cloud set also includes: taking the candidate grid where the central point cloud is located as a second central grid, and determining that the laser point cloud data in the candidate grid within a second preset radius whose point cloud height is higher than that of the central point cloud is an obstacle point cloud.

[0089] The grid containing the center point cloud is used as the second center grid. The point cloud height corresponding to the laser point cloud data in the grid within the second preset radius is calculated. If the point cloud height is higher than the point cloud height corresponding to the center point cloud, the laser point cloud data is determined to be an obstacle point cloud. The first preset radius can be the same as or different from the second preset radius, and is not specifically limited here. In the present application, the second preset radius is preferably 5 cm, the same as the first preset radius.

[0090] Taking the grid where the central point cloud is located as the center and the second preset radius as the extraction range, the point cloud data within the extraction range is calculated and the obstacle point cloud within the extraction range is determined. Taking the grid as the center can further expand the extraction range and improve the speed of obstacle point cloud extraction.

[0091] Step S1324: performing search processing on the obstacle point cloud set to obtain a road boundary point cloud set in the obstacle point cloud set.

[0092] Considering that some point clouds on the road boundary are far away from the road and belong to the plane area relative to other road boundary parts, the point cloud set corresponding to the road boundary cannot be completely divided into the obstacle point cloud set. Therefore, it is necessary to search and extract the obstacle point cloud from the laser point cloud data again according to the characteristics of the road boundary to obtain the road boundary point cloud set in the obstacle point cloud set. The search method can be a search method from far to near or a search method from near to far. The search method is not specifically limited here. In the embodiment of the present application, a search method from near to far is adopted, and the search method is preferably a middle-to-side search method.

[0093] See also Figure 7 , Figure 7 A schematic diagram of an obstacle point cloud set for extrinsic parameter calibration of a LiDAR provided in one embodiment of the present application is shown. 001 represents the non-road boundary point cloud set fitted from the obstacle point cloud set, and 002 represents the road boundary point cloud set fitted from the road boundary point cloud set. The LiDAR uses a light beam to acquire point cloud data corresponding to the vehicle's surrounding environment, processes this point cloud data to generate an obstacle point cloud set, and then performs search processing on the obstacle point cloud set, thereby dividing the obstacle point cloud set into a road boundary point cloud set and a non-road boundary point cloud set, ensuring the accuracy of the road fitting process in the subsequent process.

[0094] Specifically, the obstacle point cloud set is projected onto a grid map to obtain an obstacle grid; and a row-by-row search is performed for each obstacle point cloud in the obstacle grid, with the lateral center of the vehicle as the search center and a third preset radius as the search radius, and a road boundary point cloud set in the obstacle point cloud set is obtained from the search road boundary point cloud within a search range corresponding to the search radius.

[0095] See also Figure 8 , Figure 8The schematic diagram of the search for road boundary point cloud set based on the external parameter calibration of the laser radar provided in one embodiment of the present application is shown. The lateral center of the vehicle is used as the search center, and the road boundary point cloud existing in the obstacle grid is searched row by row with a third preset radius as the search radius, so as to obtain the road boundary point cloud set in the obstacle point cloud set. The obstacle point cloud set is projected onto the grid map to obtain the obstacle grid map, and the road boundary point cloud data is searched from the obstacle grid map by the Middle-to-Side search method, and the road boundary point cloud data is fitted using the least squares method to obtain the road boundary area. For the specific fitting image, please refer to Figure 9 , Figure 9 This figure shows a road boundary region fitting diagram for extrinsic parameter calibration of a LiDAR provided by an embodiment of the present application. This method searches and extracts road boundary point cloud data from an obstacle point cloud, avoiding the need to search for road boundary point cloud data from a two-dimensional point cloud grid. This further improves the efficiency of searching for road boundary point clouds and ensures the accuracy of the road plane fitting point cloud data.

[0096] Step S1325: Acquire the road area point cloud set based on the non-obstacle point cloud set and the road boundary point cloud set.

[0097] Obstacle point clouds and non-obstacle point clouds are obtained from the candidate grids, and then the road boundary point cloud is obtained from the obstacle point cloud. The non-obstacle point cloud and road boundary point cloud are combined to form a road plane fitting point cloud. The candidate grids are first divided into obstacle point clouds and non-obstacle point clouds, and then the road boundary point cloud is extracted from the obstacle point cloud. Finally, the non-obstacle point cloud and road boundary point cloud are combined to form a road plane fitting point cloud, ensuring the accuracy of the road plane fitting point cloud.

[0098] Step S134: fitting the road in the vehicle coordinate system based on the road plane fitting point set to obtain a road area plane equation, where the road area plane equation is a plane corresponding to the road plane fitting point set in the vehicle coordinate system.

[0099] The road fitting point cloud is fitted using a fitting algorithm to obtain the road region plane equation corresponding to the road fitting point cloud in the LiDAR coordinate system. The fitting algorithm can be a Random Sample Consensus (RANSAC) algorithm or other point cloud data fitting algorithm, which is not specifically limited here.

[0100] Step S136: Determine the transformation matrix based on the road area plane equation.

[0101] The road area plane equation in the LiDAR coordinate system is converted using a preset conversion method to obtain a transformation matrix, which transforms the road area plane equation obtained from the LiDAR coordinate system into the vehicle coordinate system. The conversion method can be around the x-axis, the y-axis, the z-axis, or first around the x-axis, then around the y-axis, and finally around the z-axis. There is no specific conversion method.

[0102] Specifically, according to RANSAC, the obtained road fitting point cloud set is fitted to obtain the plane equation of the fitting road area in the lidar coordinate system. The plane equation of the fitting road area is:

[0103] AX+BY+CZ+D=0 (3)

[0104] The current road area plane is horizontal, and the origin of the vehicle coordinate system is the ground contact point between the center of the vehicle's rear axle and the ground. Therefore, the equation of the road area in the vehicle coordinate system is Z=0. To make the two planes coincide, the road area plane equation AX+BY+CZ+D=0 needs to be rotated around the x-axis, rotated around the y-axis, and translated about the z-axis to obtain the road area plane equation. Rotating about the x-axis can obtain the roll angle (roll) of the laser radar, rotating about the y-axis can obtain the pitch angle (pitch) of the laser radar, and translating the plane equation about the z-axis can calibrate the Z-axis translation parameter of the laser radar. According to the road area plane equation, the above three parameters can be directly solved as follows:

[0105] The equation to solve roll is:

[0106] The equation for solving pitch is:

[0107] The equation for solving the Z-axis translation parameter is:

[0108] The conversion matrix is ​​obtained according to formulas (4), (5), and (6). Specifically, the conversion matrix includes: a roll angle coordinate transformation matrix corresponding to the roll angle of the laser radar, a pitch angle coordinate transformation matrix corresponding to the pitch angle, and a Z-axis translation parameter coordinate transformation matrix corresponding to the Z-axis translation parameter.

[0109] The coordinate transformation matrix of roll is:

[0110] The coordinate transformation matrix of pitch is:

[0111] The coordinate transformation matrix of Z is:

[0112] Step S140: Determine the external parameters of the laser radar based on the angle between the vehicle and the preset side lane line, the preset number of coordinate points and the transformation matrix.

[0113] A preset number of coordinate points on the preset side lane line in the laser point cloud data are input into the transformation matrix to obtain the coordinate points in the vehicle coordinate system. The coordinate points in the vehicle coordinate system are combined with the angle between the vehicle and the preset side lane line, the vehicle parameters and the road area plane model to calibrate the external parameters of the lidar.

[0114] See also Figure 10 Step S140 specifically includes steps S142 to S148, which are described in detail as follows:

[0115] Step S142: transforming a preset number of coordinate points on the preset side lane line according to the roll angle coordinate transformation matrix, the pitch angle coordinate transformation matrix, and the Z-axis translation parameter coordinate transformation matrix to obtain transformed coordinate points.

[0116] A preset number of coordinate points on the preset side lane line are transformed according to formulas (7), (8), and (9) to obtain the transformed coordinate points.

[0117] For example, the coordinate points A(x1, y1, z1) and B(x2, y2, z2) of the preset number of points located on the preset side lane line in the laser point cloud data are transformed into A'(x1', y1', z1') and B'(x2', y2', z2') after the above transformation, where:

[0118] A'(x1',y1',z1') is obtained by converting A(x1,y1,z1) according to formulas (7), (8), and (9). The specific conversion equation is:

[0119]

[0120] A(x1,y1,z1) is obtained from A'(x1',y1',z1') according to formula (10).

[0121] B'(x2',y2',z2') is obtained by converting B(x2,y2,z2) according to formulas (7), (8), and (9). The specific conversion equation is:

[0122]

[0123] B(x2,y2,z2) is obtained according to formula (11).

[0124] Step S144: Obtain the equation of the straight line corresponding to the transformed coordinate point.

[0125] Specifically, from formulas (10) and (11), we can see that A'(x1', y1', z1') and B'(x2', y2', z2') are in the XOY plane, and A'B' can be expressed as the following linear equation aX+bY+c=0.

[0126] Step S146: According to the straight line equation, obtain the angle between the straight line equation and the X-axis.

[0127] According to the straight line equation, the angle between the straight line equation and the X-axis is obtained. For example, the angle δ between the straight line and the X-axis is:

[0128]

[0129] δ is obtained by formula (12).

[0130] Step S148: Determine the yaw angle of the laser radar based on the angle between the vehicle and the preset side lane line and the angle between the straight line equation and the X-axis.

[0131] Assuming that the angle between the vehicle and the preset side lane line is α, the calculation equation of the yaw angle yaw is:

[0132] yaw=α-δ (13)

[0133] The yaw angle of the lidar is calibrated according to formula (13).

[0134] During yaw angle calibration, the roll, pitch, and Z-axis translation parameters of the calibrated LiDAR are used for calibration. In addition to roll, pitch, yaw, and Z-axis translation parameters, calibrating the LiDAR's external parameters also requires calibrating the X-axis and Y-axis translation parameters. These X-axis and Y-axis translation parameters can be obtained through actual measurement; the specific measurement method is not specified here.

[0135] In an embodiment of the present application, the external parameters of the laser radar are calibrated by the angle between the current lane position of the vehicle where the laser radar is located and the lane line, as well as multiple coordinate points on the lane line. This calibration method does not require additional calibration tools, reducing the hardware cost during the calibration process. Moreover, it is only necessary to obtain the angle between the vehicle and the lane line and the coordinates on the lane line to achieve rapid calibration of the laser radar, greatly shortening the calibration time.

[0136] See also Figure 11, which shows a structural block diagram of an external parameter calibration device 300 for a laser radar provided in an embodiment of the present application. The external parameter calibration device 300 for the laser radar is applied to the electronic device 100 and includes: a laser point cloud data acquisition module 310 for acquiring laser point cloud data through laser radar scanning of a vehicle; a lane line data acquisition module 320 for acquiring the angle between the vehicle and the preset side lane line of the lane where the vehicle is located and a preset number of coordinate points in the laser point cloud data located on the preset side lane line, where the preset side is the left or right side; a conversion matrix determination module 330 for determining the conversion matrix of the road plane where the vehicle is located from the laser radar coordinate system to the vehicle body coordinate system based on the laser point cloud data; and an external parameter determination module 340 for determining the external parameters of the laser radar based on the angle between the vehicle and the preset side lane line, the preset number of coordinate points, and the conversion matrix.

[0137] In some embodiments of the present application, the transformation matrix determination module 330 includes: a road plane fitting point set acquisition module, which is used to determine the road plane fitting point set of the road where the vehicle is located based on the laser point cloud data; a road area plane equation acquisition module, which is used to fit the lane in the vehicle coordinate system based on the road plane fitting point set, and obtain the road area plane equation in the lidar coordinate system, wherein the road area plane equation is the plane corresponding to the road plane fitting point set in the vehicle coordinate system; and a transformation matrix acquisition module, which is used to determine the transformation matrix based on the road area plane equation.

[0138] In some embodiments of the present application, the road plane fitting point set acquisition module 330 includes: a two-dimensional point cloud grid map acquisition module, which is used to project the laser point cloud data into a two-dimensional grid map to obtain a two-dimensional point cloud grid map; a screening module, which is used to screen out candidate grids from the two-dimensional point cloud grid map based on the local point cloud height difference between each grid in the two-dimensional point cloud grid map; a segmentation and extraction module, which is used to segment and extract the candidate grids to obtain an obstacle point cloud set and a non-obstacle point cloud set; a search and processing module, which is used to search and process the obstacle point cloud set to obtain a road boundary point cloud set in the obstacle point cloud set; and a fitting point set acquisition module, which is used to obtain the road plane fitting point set based on the non-obstacle point cloud set and the road boundary point cloud set.

[0139] In some embodiments of the present application, the screening module includes: a maximum height point cloud value acquisition module, which is used to traverse the grids in the two-dimensional point cloud grid map, take the currently traversed grid as the first central grid, and obtain the maximum height point cloud value based on the coordinates of the point cloud data in the first central grid; a height difference acquisition module, which is used to calculate the height difference between the maximum height point cloud value and the minimum height point cloud value corresponding to each grid within a preset size range with the first central grid as the center, and the minimum height point cloud value is calculated based on the coordinates of the point cloud data in the grid; a count value counting module, which is used to add 1 to the count value if the height difference is higher than the first preset height difference; and a candidate grid marking module, which is used to mark the first central grid as a candidate grid if the count value exceeds a pre-designed numerical threshold within a preset size range.

[0140] In some embodiments of the present application, the segmentation and extraction module includes: a slope acquisition module, which is used to traverse the laser point cloud data. If the currently traversed laser point cloud data is in the candidate grid, the laser point cloud data is used as the center point cloud to calculate the slope of the point cloud data within a first preset radius range and the center point cloud; a slope counting module, which is used to add 1 to the slope count if the slope is greater than a preset slope threshold; a height counting module, which is used to add 1 to the height count if the height difference between the point cloud height of the laser point cloud data greater than the preset slope threshold and the center point cloud exceeds a second preset height difference; an obstacle point cloud determination module, which is used to determine that the center point cloud is an obstacle point cloud if the slope count corresponding to the laser point cloud data within the first preset radius range of the center point cloud is not less than the slope count threshold and the height count is not less than the height count threshold.

[0141] In some embodiments of the present application, the obstacle point cloud determination module also includes: a module for determining an obstacle point cloud based on a second central grid, which is used to take the candidate grid where the central point cloud is located as the second central grid, and determine that the point cloud data in the candidate grid within the second preset radius whose point cloud height is higher than the central point cloud is an obstacle point cloud.

[0142] In some embodiments of the present application, the fitting point set acquisition module includes: an obstacle grid acquisition module, which is used to project the obstacle point cloud set into a grid map to obtain an obstacle grid; a search module, which performs a row-by-row search on the obstacle point cloud in each obstacle grid, uses the lateral center of the vehicle as the search center, uses a third preset radius as the search radius, and obtains a road boundary point cloud set in the obstacle point cloud set from the search road boundary point cloud in the search range corresponding to the search radius.

[0143] In some embodiments of the present application, the conversion matrix includes: a roll angle coordinate transformation matrix corresponding to the roll angle of the laser radar, a pitch angle coordinate transformation matrix corresponding to the pitch angle, and a Z-axis translation parameter coordinate transformation matrix corresponding to the Z-axis translation parameter; the external parameter determination module 360 ​​includes: a transformed coordinate point acquisition module, which is used to transform a preset number of coordinate points on the preset side lane line according to the roll angle coordinate transformation matrix, the pitch angle coordinate transformation matrix, and the Z-axis translation parameter coordinate transformation matrix to obtain the transformed coordinate points; a straight line equation acquisition module, which is used to obtain the straight line equation corresponding to the transformed coordinate point; a straight line equation and X-axis angle acquisition module, which is used to obtain the angle between the straight line equation and the X-axis according to the straight line equation; a yaw angle acquisition module, which is used to determine the yaw angle of the laser radar according to the angle between the vehicle and the preset side lane line and the angle between the straight line equation and the X-axis.

[0144] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0145] In several embodiments provided in this application, the coupling between modules may be electrical, mechanical or other forms of coupling.

[0146] In addition, the functional modules in the various embodiments of the present application may be integrated into a processing module, or each module may exist physically separately, or two or more modules may be integrated into a single module. The above-mentioned integrated modules may be implemented in the form of hardware or software functional modules.

[0147] Please refer to Figure 12 , which shows a structural block diagram of an electronic device provided in an embodiment of the present application. The electronic device 100 can be a switch, a computer, or a control unit with data transmission. The electronic device 100 in the present application may include one or more of the following components: a processor 110, a memory 120, and one or more application programs, wherein the one or more application programs can be stored in the memory 120 and configured to be executed by one or more processors 110, and the one or more programs are configured to execute the method described in the aforementioned method embodiment.

[0148] The processor 110 may include one or more processing cores. The processor 110 utilizes various interfaces and circuits to connect various components within the electronic device 100. It executes instructions, programs, code sets, or instruction sets stored in the memory 120, and accesses data stored in the memory 120 to perform various functions and process data within the electronic device 100. Optionally, the processor 110 may be implemented using at least one of the following hardware forms: a digital signal processing (DSP), a field-programmable gate array (FPGA), or a programmable logic array (PLA). The processor 110 may integrate one or a combination of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. The CPU primarily processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing display content; and the modem handles wireless communications. It is understood that the modem may not be integrated into the processor 110 and may be implemented separately via a communications chip.

[0149] The memory 120 may include a random access memory (RAM) or a read-only memory (ROM). The memory 120 may be used to store instructions, programs, codes, code sets, or instruction sets. The memory 120 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the following various method embodiments, etc. The data storage area may also store data created by the electronic device 100 during use (such as a phone book, audio and video data, chat history data), etc.

[0150] An embodiment of the present application further provides a computer-readable storage medium, in which program code is stored. The program code can be called by a processor to execute the method described in the above method embodiment.

[0151] The computer-readable storage medium may be an electronic memory such as a flash memory, an EEPROM (Electrically Erasable Programmable Read-Only Memory), an EPROM, a hard disk, or a ROM. Alternatively, the computer-readable storage medium includes a non-transitory computer-readable storage medium. The computer-readable storage medium has storage space for program codes for executing any of the method steps described above. These program codes can be read from or written to one or more computer program products. The program codes can be compressed, for example, in an appropriate form.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for calibrating external parameters of a laser radar, characterized in that: The method comprises: Obtain laser point cloud data through vehicle laser radar scanning; Obtaining an angle between the vehicle and a preset lane line on a side of the lane where the vehicle is located, and a preset number of coordinate points located on the preset lane line in the laser point cloud data, where the preset side is the left side or the right side; Projecting the laser point cloud data onto a two-dimensional grid map to obtain a two-dimensional point cloud grid map; Screening candidate grids from the two-dimensional point cloud grid image according to the local point cloud height difference between each grid in the two-dimensional point cloud grid image; Segmenting and extracting the candidate grid to obtain an obstacle point cloud set and a non-obstacle point cloud set; Projecting the obstacle point cloud set onto a grid map to obtain an obstacle grid; Performing a row-by-row search for each obstacle point cloud in the obstacle grid, using the lateral center of the vehicle as the search center and a third preset radius as the search radius, and obtaining a road boundary point cloud set in the obstacle point cloud set from search road boundary point clouds within a search range corresponding to the search radius; Acquire a road plane fitting point set according to the non-obstacle point cloud set and the road boundary point cloud set; Fitting the road plane of the lane to the laser radar coordinate system based on the road plane fitting point set to obtain a road area plane equation in the laser radar coordinate system; Based on the road area plane equation, determining a transformation matrix of the road plane where the vehicle is located from the laser radar coordinate system to the vehicle body coordinate system; The external parameters of the laser radar are determined based on the angle between the vehicle and the preset side lane line, the preset number of coordinate points and the transformation matrix.

2. The method according to claim 1, characterized in that The step of selecting candidate grids from the two-dimensional point cloud grid image according to the local point cloud height differences between grids in the two-dimensional point cloud grid image comprises: Traversing the grids in the two-dimensional point cloud grid image, taking the currently traversed grid as a first central grid, and obtaining a maximum height point cloud value according to the coordinates of the point cloud data in the first central grid; Taking the first central grid as the center, calculate the height difference between the maximum height point cloud value and the minimum height point cloud value corresponding to each grid within a preset size range, where the minimum height point cloud value is calculated based on the coordinates of the point cloud data in the grid; If the height difference is higher than the first preset height difference, the count value is increased by 1; Within the range of the preset size, if the count value exceeds a pre-designed numerical threshold, the first central grid is marked as a candidate grid.

3. The method according to claim 1, characterized in that The step of segmenting and extracting the candidate grid to obtain an obstacle point cloud set and a non-obstacle point cloud set includes: Traversing the laser point cloud data, if the currently traversed laser point cloud data is in the candidate grid, using the laser point cloud data as the center point cloud, and calculating the slope of the point cloud data within a first preset radius and the center point cloud; If the slope is greater than the preset slope threshold, the slope count is increased by 1; If the height difference between the laser point cloud data greater than the preset slope threshold and the center point cloud exceeds a second preset height difference, the height count is incremented by 1; If the slope count corresponding to the laser point cloud data within the first preset radius range of the central point cloud is not less than the slope count threshold and the height count is not less than the height count threshold, the central point cloud is determined to be an obstacle point cloud.

4. The method according to claim 3, characterized in that The step of segmenting and extracting the candidate grid to obtain an obstacle point cloud set and a non-obstacle point cloud set further includes: The candidate grid where the central point cloud is located is used as the second central grid, and point cloud data in the candidate grids within a second preset radius whose point cloud height is higher than that of the central point cloud is determined as an obstacle point cloud.

5. The method according to any one of claims 1 to 4, characterized in that: The conversion matrix includes: a roll angle coordinate transformation matrix corresponding to the roll angle of the laser radar, a pitch angle coordinate transformation matrix corresponding to the pitch angle, and a Z-axis translation parameter coordinate transformation matrix corresponding to the Z-axis translation parameter; the external parameters of the laser radar are determined based on the angle between the vehicle and the preset side lane line, the preset number of coordinate points, and the conversion matrix, including: transforming a preset number of coordinate points on the preset side lane line according to the roll angle coordinate transformation matrix, the pitch angle coordinate transformation matrix, and the Z-axis translation parameter coordinate transformation matrix to obtain transformed coordinate points; Obtaining the equation of the straight line corresponding to the transformed coordinate point; According to the straight line equation, obtaining the angle between the straight line equation and the X-axis; The yaw angle of the laser radar is determined based on the angle between the vehicle and the preset side lane line and the angle between the straight line equation and the X-axis.

6. An electronic device, characterized in that: include: one or more processors; Memory; One or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the method according to any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program code, which can be called by a processor to execute the method according to any one of claims 1 to 5.

Images