CN114966630B - Multi-laser radar calibration method and system based on target sphere - Google Patents

Abstract

The invention discloses a multi-laser radar calibration method based on a target sphere; the method comprises the following steps: determining the type, number and detection distance of the laser radar according to the common field of view; calculating the size of the optimal target sphere; placing a target sphere at a common field of view of the lidar to be calibrated; segmenting foreground point clouds from the laser radar point clouds, and clustering by using a region growing method; fitting the sphere shape, and calculating the sphere center position of the target sphere; accumulating the estimated spherical center coordinates; the obtained spherical center coordinates are clustered in Euclidean distance, and the average value is taken as the final spherical center estimated position; selecting more than three spherical center coordinates, and calculating the position and posture differences of different laser radar coordinate systems; all laser radar coordinate systems are integrated into the same coordinate system, so that calibration among multiple laser radars is realized; the method is suitable for the conditions of different baselines, and the calibration and fusion of the multi-laser radar are carried out according to the position of the target sphere.

Description

A multi-lidar calibration method and system based on target sphere

Technical Field

The present invention relates to the field of laser radar technology, and in particular to a multi-laser radar calibration method and system based on a target sphere.

Background Art

With the development of intelligent systems, perception systems consisting of multiple 3D LiDARs (detection and ranging) are being gradually deployed. This is because 3D LiDARs have the following advantages: Compared with 2D cameras, 3D LiDARs can provide accurate 3D information about the environment and enrich input information such as 3D position, direction, and speed; Compared with a single 3D LiDAR, multiple 3D LiDARs can increase the perception field of view, thereby avoiding blind spots.

Based on the distance between multiple LiDARs, these perception systems can be divided into two categories. Short baseline scenario: the distance between LiDARs is very short (tens of centimeters to several meters), such as multiple LiDARs mounted on unmanned vehicles, obstacle avoidance LiDARs on autonomous mobile robots, perception LiDARs on unmanned road sweepers, etc. Long baseline scenario: the distance between LiDARs is very long (tens of meters), such as multiple LiDARs on large bucket wheel reclaimers, and a full range perception system composed of multiple LiDARs at an intersection.

In order to fuse the information of multiple 3D LiDARs, accurate external calibration between LiDARs is a basic and indispensable technology; however, there is no unified and precise calibration method for various intelligent systems at present.

Summary of the invention

The purpose of the present invention is to overcome the above-mentioned defects in the prior art and provide a multi-lidar calibration method and system based on a target sphere; it can be applicable to situations with different baselines, automatically detect the target sphere, calibrate and fuse the multi-lidar according to the position of the target sphere, and realize the construction of a multi-lidar system.

To achieve the above object, the present invention provides a multi-lidar calibration method based on a target sphere, comprising the following steps:

Step S11: Determine the model, number and detection distance of the laser radar according to the required common field of view; calculate the optimal size of the target sphere;

Step S12: placing the target sphere in the common field of view of the laser radar to be calibrated;

Step S21: segmenting the foreground point cloud from the laser radar point cloud;

Step S22: clustering the segmented foreground point clouds using the region growing method;

Step S23: fitting the sphere shape to calculate the center position of the target sphere;

Step S24: accumulating the estimated sphere center coordinates; performing Euclidean distance clustering on the processed sphere center coordinates, and taking the average value as the final sphere center estimated position;

Step S31: Select three or more spherical center coordinates to calculate the position and posture differences of different laser radar coordinate systems;

Step S32: Align all laser radar coordinate systems to the same coordinate system to achieve calibration between multiple laser radars.

Preferably, in step S11, the target sphere is a uniform sphere that can reflect the detection light of the laser radar, and the ideal size of the target sphere is calculated by the detection method of the laser radar.

Preferably, in step S21, when extracting the center of the sphere, the distance discontinuity of the lidar point cloud is used to set a distance mutation threshold, and if the distance change of adjacent points exceeds the mutation threshold, the point is recorded; thereby, the rising edge and the falling edge of the point cloud are identified, and the foreground point cloud is segmented from the lidar point cloud.

Preferably, in step S23, for each cluster obtained, a two-step random sampling consistency fast fitting from coarse to fine is used to determine the candidate clusters that may be spherical, the size of the candidate cluster data and the radius; then, based on the known target sphere size, clusters that are too large or too small are screened out, and the cluster that is closest to the preset radius and has the largest cluster data volume is selected as the target cluster; then, by performing fixed point iteration on the target cluster, the center position of the target sphere is calculated.

Preferably, in step S24, before each target sphere or the same sphere moves to the next position, its stay time should allow each laser radar to be calibrated to capture more than 60 frames of point cloud data, and 60 frames of point cloud data are collected cumulatively. Steps S21 to S23 are repeated for each frame of point cloud data, and the processed coordinates of the sphere center are clustered by Euclidean distance; after filtering out abnormal processing results, the remaining processing results are averaged as the final estimated position of the sphere center.

Preferably, in step S31, step S12 is repeated to set a plurality of identical target spheres or move a target sphere to a plurality of different positions, the number of the different positions should be greater than 3, and at least 3 different positions are not collinear.

Preferably, each target sphere or a target sphere moves to multiple different positions, and the target sphere at each position is visible to all lidars to be calibrated, so that during detection, the lidar can extract the coordinates of the center of the target sphere in the coordinate system of each lidar to be calibrated.

Preferably, steps S21 to S24 are repeated for each position of the target sphere and each laser radar; after obtaining the coordinate differences of the same sphere center of more than three different laser radar coordinate systems, the nearest point iteration is used to calculate the position and posture differences of different laser radar coordinate systems.

Preferably, in step S32, by inversely transforming the position and posture differences of different laser radar coordinate systems, all laser radar coordinate systems are aligned to the same coordinate system, thereby achieving calibration between multiple laser radars.

The present invention also provides a system using the above-mentioned multi-laser radar calibration method based on a target sphere, comprising: a target sphere, a point cloud accumulation module, a sphere center extraction module, a position and posture estimation module and a calibration reprojection module;

Multiple identical target spheres are placed or a target sphere is moved to multiple different locations in the common field of view of multiple laser radars;

Point cloud accumulation module: used to accumulate multi-frame radar point clouds, so that when extracting the center of the sphere, the coordinates of the sphere position in the laser radar coordinate system to be calibrated are determined by the center position of the sphere calculated through multi-frame clustering and average calculation;

Sphere center extraction module: segment, cluster, fit, and iterate each lidar point cloud to extract the sphere center coordinates in the corresponding lidar coordinate system;

Position and attitude estimation module: It performs the closest point iteration on the coordinate difference of the same sphere center in different LiDAR coordinate systems, and calculates the position and attitude difference of different LiDAR coordinate systems;

Calibration and reprojection module: Compensate the position and posture of the lidar based on the calculation results so that all lidars are reprojected to the same coordinate system to complete the lidar calibration.

Compared with the prior art, the present invention has the following beneficial effects:

The method and system provided by the present invention adopt one or more target spheres as calibration objects. The laser radar performs accurate and automatic detection on the target sphere, extracts the coordinates of the center of the target sphere in the laser radar coordinate system, calculates the coordinate difference of the same center in different laser radar coordinate systems, and estimates the position and posture differences between different laser radars; then, calibration between multiple laser radars is achieved by compensating for the position and posture differences; the detected center position of the target sphere is very accurate, which greatly improves the calibration accuracy, and the target sphere is a uniform sphere that can reflect the detection light of the laser radar and is easy to detect, so the overall calculation amount is greatly reduced; it can also be applied to different baseline conditions, automatically detect the target sphere, calibrate and fuse multiple laser radars according to the position of the target sphere, and realize the construction of a multi-laser radar system.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG1 is a flowchart of a multi-laser radar calibration method based on a target sphere provided by the present invention;

FIG2 is a two-dimensional plane schematic diagram of the laser radar provided by the present invention scanning a target sphere and placing the target sphere at different positions.

DETAILED DESCRIPTION

The following will be combined with the drawings in this embodiment of the present invention to clearly and completely describe the technical solution in this embodiment of the present invention. Obviously, the described embodiment is one embodiment of the present invention, not all embodiments of the present invention. Based on this embodiment of the present invention, all other embodiments of the present invention obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

Embodiment 1

Please refer to FIG. 1 and FIG. 2 , a first embodiment of the present invention provides a multi-lidar calibration method based on a target sphere.

First, the method provided by the present invention is described as a whole: the method sets multiple identical target spheres in the common field of view of multiple laser radars or moves a target sphere to multiple different positions; and uses laser radar for automatic detection to extract the coordinates of the center of the target sphere in the laser radar coordinate system. Thus, the laser radar can estimate the position and posture differences between different laser radars through the coordinate differences of the same center in different laser radar coordinate systems; and then calibrate the multiple laser radars by compensating for the position and posture differences. Through the above-mentioned corresponding calibration steps, the position and posture differences (rotation matrices and translation vectors) between the laser radar frames can be obtained; by compensating for such position and posture differences, the calibration and information fusion of the three-dimensional laser radar can be achieved.

Furthermore, the method provided by the present invention can also be applied to different baseline conditions, automatically detect the target sphere, calibrate and fuse multiple laser radars according to the position of the target sphere, and realize the construction of a multi-laser radar system.

Next, a multi-lidar calibration method based on a target sphere is introduced in detail, please refer to Figure 1; it includes the following steps.

Step S1: Processing the target sphere; specifically, step S1 includes:

Step S11: Calculate the optimal target sphere size; select the radar signal, number and detection distance to be used according to the common field of view required by the task; the target sphere is a uniform sphere that can reflect the detection light of the laser radar, and the ideal size of the target sphere is calculated by the detection method of the laser radar.

The required optimal target sphere size is calculated through relevant data; the relevant calculation formula can be derived according to the diagram shown in Figure 2.

Step S12: Place the target sphere at a suitable position; place the target sphere in the common field of view of the laser radars to be calibrated; it should be noted that for two or more laser radar systems, calibration can be performed in pairs or multiple times at the same time; when performed in pairs, it is only necessary to place the target sphere in the common field of view of two laser radars; when performed multiple times at the same time, it is necessary to place the target sphere in the common field of view of all laser radars to be calibrated.

Step S2: Processing the laser radar point cloud; specifically, step S2 includes:

Step S21: Segment the foreground point cloud from the LiDAR point cloud; use the distance discontinuity of the LiDAR point cloud to set a distance mutation threshold, and record the point if the distance change of adjacent points exceeds the mutation threshold; thereby identify the rising edge and falling edge of the point cloud, and segment the foreground point cloud from the LiDAR point cloud.

Step S22: clustering of foreground point clouds; clustering the segmented foreground point clouds using a region growing method.

Step S23: sphere fitting and sphere center position estimation; for each cluster obtained, use a two-step random sampling consistency fast fitting from coarse to fine to determine the candidate clusters that may be spherical, the size of the candidate cluster data and the radius; then, according to the known target sphere size, filter out clusters that are too large or too small, and select the cluster that is closest to the preset radius and has the largest cluster data as the target cluster; then perform fixed point iteration on the target cluster to infer the sphere center position of the target sphere.

Step S24: Accumulate the estimated sphere center coordinates; before each target sphere or the same sphere moves to the next position, its stay time should allow each laser radar to be calibrated to capture more than 60 frames of point cloud data, and 60 frames of point cloud data are collected cumulatively. Steps S21 to S23 are repeated for each frame of point cloud data, and the processed sphere center coordinates are clustered by Euclidean distance; after filtering out abnormal processing results, the remaining processing results are averaged as the final estimated sphere center position.

Step S3: Processing the estimated sphere center; specifically, step S3 includes:

Step S31: Select three or more spherical center coordinates and calculate the position and posture differences; next, repeat step S12, set multiple identical target spheres or move a target sphere to multiple different positions, the number of different positions should be greater than 3, and at least 3 different positions are not collinear.

Furthermore, each target sphere or a target sphere moves to multiple different positions, and the target sphere at each position is visible to all laser radars to be calibrated, so that during detection, the laser radar can extract the coordinates of the center of the target sphere in the coordinate system of each laser radar to be calibrated.

Furthermore, steps S21 to S24 are repeated for each position of the target sphere and each laser radar; after obtaining the coordinate differences of more than three identical sphere centers in different laser radar coordinate systems, the nearest point iteration is used to calculate the position and posture differences in different laser radar coordinate systems.

Step S32: unify the coordinate systems of multiple laser radars and complete the calibration; by inversely transforming the position and posture differences of different laser radar coordinate systems, all laser radar coordinate systems are unified into the same coordinate system to achieve calibration between multiple laser radars.

This method provides an efficient and accurate calibration solution for two or more fixed laser radars; the placement tolerance of each radar is high, and providing a suitable target sphere can enable the laser radar to complete convenient and accurate calibration under the condition of a distance difference of 30 meters or more and a viewing angle difference of 180 degrees.

Embodiment 2

Please refer to Figures 1 and 2. Embodiment 2 of the present invention provides a system that adopts the multi-lidar calibration method based on the target sphere described in Embodiment 1.

Please refer to Figures 1 and 2, the multi-lidar calibration system based on the target sphere includes: a target sphere, a point cloud accumulation module, a sphere center extraction module, a position and posture estimation module and a calibration reprojection module;

Multiple identical target spheres are placed or a target sphere is moved to multiple different locations in the common field of view of multiple laser radars;

Point cloud accumulation module: used to accumulate multi-frame radar point clouds, so that when extracting the center of the sphere, the coordinates of the sphere position in the laser radar coordinate system to be calibrated are determined by the center position of the sphere calculated through multi-frame clustering and average calculation;

Sphere center extraction module: segment, cluster, fit, and iterate each lidar point cloud to extract the sphere center coordinates in the corresponding lidar coordinate system;

Position and attitude estimation module: It performs the closest point iteration on the coordinate difference of the same sphere center in different LiDAR coordinate systems, and calculates the position and attitude difference of different LiDAR coordinate systems;

Calibration and reprojection module: Compensate the position and posture of the lidar based on the calculation results so that all lidars are reprojected to the same coordinate system to complete the lidar calibration.

The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims (10)

1. A multi-laser radar calibration method based on a target sphere, characterized in that it comprises the following steps: Step S11: Determine the model, number and detection distance of the laser radar according to the required common field of view; calculate the optimal size of the target sphere; Step S12: placing the target sphere in the common field of view of the laser radar to be calibrated; Step S21: segmenting the foreground point cloud from the laser radar point cloud; Step S22: clustering the segmented foreground point clouds using the region growing method; Step S23: fitting the sphere shape to calculate the center position of the target sphere; Step S24: accumulating the estimated sphere center coordinates; performing Euclidean distance clustering on the processed sphere center coordinates, and taking the average value as the final sphere center estimated position; Step S31: Select three or more spherical center coordinates to calculate the position and posture differences of different laser radar coordinate systems; Step S32: Align all laser radar coordinate systems to the same coordinate system to achieve calibration between multiple laser radars. 2. According to the multi-lidar calibration method based on target sphere according to claim 1, it is characterized in that: in the step S11, the target sphere is a uniform sphere that can reflect the detection light of the laser radar, and the ideal size of the target sphere is calculated by the detection method of the laser radar. 3. According to the multi-lidar calibration method based on the target sphere as described in claim 1, it is characterized in that: in the step S21, when extracting the center of the sphere, the distance discontinuity of the laser radar point cloud is used to set a distance mutation threshold. If the distance change of adjacent points exceeds the mutation threshold, the point is recorded; thereby, the rising edge and the falling edge of the point cloud are identified, and the foreground point cloud is segmented from the laser radar point cloud. 4. A multi-lidar calibration method based on a target sphere according to claim 3, characterized in that: in the step S23, for each cluster obtained, a two-step random sampling consistency fast fitting from coarse to fine is used to determine the candidate cluster that may be a sphere, the size of the candidate cluster data and the radius; then, according to the known size of the target sphere, the clusters that are too large or too small are screened out, and the cluster that is closest to the preset radius and has the largest cluster data volume is selected as the target cluster; then, by performing fixed point iteration on the target cluster, the center position of the target sphere is calculated. 5. According to the multi-lidar calibration method based on the target sphere according to claim 4, it is characterized in that: in the step S24, before each target sphere or the same sphere moves to the next position, its stay time should allow each laser radar to be calibrated to capture more than 60 frames of point cloud data, and 60 frames of point cloud data are collected cumulatively, and steps S21 to S23 are repeated for each frame of point cloud data, and the processed center coordinates are clustered by Euclidean distance; after filtering out abnormal processing results, the remaining processing results are averaged as the final estimated position of the center of the ball. 6. According to a multi-lidar calibration method based on a target sphere according to claim 1, it is characterized in that: in the step S31, step S12 is repeated to set multiple identical target spheres or move a target sphere to multiple different positions, the number of the different positions should be greater than 3, and at least 3 different positions are not collinear. 7. A multi-lidar calibration method based on a target sphere according to claim 6, characterized in that: each target sphere or a target sphere moves to multiple different positions, and the target sphere at each position is visible to all laser radars to be calibrated, so that during detection, the laser radar can extract the coordinates of the center of the target sphere in the coordinate system of each laser radar to be calibrated. 8. A multi-lidar calibration method based on a target sphere according to claim 7, characterized in that: steps S21 to S24 are repeated for each position of the target sphere and each laser radar; after obtaining the coordinate differences of more than three identical sphere centers in different laser radar coordinate systems, the nearest point iteration is used to infer the position and posture differences of different laser radar coordinate systems. 9. According to the multi-laser radar calibration method based on the target sphere as described in claim 8, it is characterized in that: in the step S32, by inversely transforming the position and posture differences of different laser radar coordinate systems, all laser radar coordinate systems are aligned to the same coordinate system to achieve calibration between multiple laser radars. 10. A system using any one of the target sphere-based multi-lidar calibration methods of claims 1 to 9, characterized in that it comprises: a target sphere, a point cloud accumulation module, a sphere center extraction module, a position and attitude estimation module and a calibration reprojection module; Multiple identical target spheres are placed or a target sphere is moved to multiple different locations in the common field of view of multiple laser radars; Point cloud accumulation module: used to accumulate multi-frame radar point clouds, so that when extracting the center of the sphere, the coordinates of the sphere position in the laser radar coordinate system to be calibrated are determined by the center position of the sphere calculated through multi-frame clustering and average calculation; Sphere center extraction module: segment, cluster, fit, and iterate each lidar point cloud to extract the sphere center coordinates in the corresponding lidar coordinate system; Position and attitude estimation module: It performs the closest point iteration on the coordinate difference of the same sphere center in different LiDAR coordinate systems, and calculates the position and attitude difference of different LiDAR coordinate systems; Calibration and reprojection module: Compensate the position and posture of the lidar based on the calculation results so that all lidars are reprojected to the same coordinate system to complete the lidar calibration.