Laser radar calibration method and device and readable storage medium

Abstract

The invention provides a laser radar calibration method and device and a readable storage medium. The laser radar calibration method comprises the steps of obtaining source point cloud data under a laser radar coordinate system; obtaining signboard area information based on the source point cloud data, extracting ground point cloud in the calibration site, and obtaining height jump edge points in the calibration site; determining a laser point cloud key point set according to the height jump edge points; acquiring a plurality of pieces of real angular point coordinate information of the signboard under a carrier vehicle coordinate system; according to the laser point cloud key point set and the real angular point coordinate information, determining a redundant corresponding relation between the laser point cloud key points and the real angular points; converting laser point cloud coordinates in the laser point cloud key point set into a rotation matrix and a translation vector required by a carrier vehicle coordinate system according to the redundancy corresponding relation; and obtaining calibration parameters of the laser radar according to a preset distance threshold, the rotation matrix and the translation vector. And surface constraints are converted into point features, so that the algorithm complexity is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned driving, in particular to a laser radar calibration method, device and readable storage medium. BACKGROUND

[0002] In the related art, in the field of current laser radar external parameter calibration, the calibration algorithm relies on establishing accurate geometric constraints in three-dimensional space to solve, that is, by introducing a mixed geometric constraint of various types such as points, lines, and surfaces, and using the least squares method to solve the external parameter optimal solution that minimizes the overall error. However, due to the fusion of features of different dimensions such as points, lines, and surfaces, the algorithm needs to handle complex feature extraction and association problems, thereby increasing the complexity and difficulty of implementation of the algorithm. SUMMARY

[0003] The present application aims to solve the technical problem that the algorithm needs to handle complex feature extraction and association in the prior art or related art.

[0004] To this end, the first aspect of the present application provides a laser radar calibration method.

[0005] The second aspect of the present application provides a laser radar calibration device.

[0006] The third aspect of the present application provides a laser radar calibration device.

[0007] The fourth aspect of the present application provides a readable storage medium.

[0008] Therefore, according to the first aspect of the present application, a laser radar calibration method is provided, the laser radar is arranged on a vehicle, and the laser radar calibration method comprises the following steps: controlling the laser radar to scan a pre-constructed calibration site to obtain source point cloud data in a laser radar coordinate system, wherein the calibration site contains a plurality of signboards distributed in different spatial poses; based on the source point cloud data, obtaining signboard region information and extracting ground point cloud in the calibration site; according to the signboard region information and the ground point cloud, obtaining height jump edge points in the calibration site; determining a laser point cloud key point set according to the height jump edge points; obtaining a plurality of real corner point coordinate information of the signboard in a vehicle coordinate system; determining a redundant correspondence relationship between the laser point cloud key points and the real corner points according to the laser point cloud key point set and the plurality of real corner point coordinate information; converting laser point cloud coordinates in the laser point cloud key point set to a rotation matrix and a translation vector required by coordinate conversion of the laser radar to the vehicle coordinate system according to the redundant correspondence relationship; and obtaining calibration parameters of the laser radar according to a preset distance threshold, the rotation matrix, and the translation vector.

[0009] The application provides a calibration method of a laser radar. The laser radar is arranged on a vehicle. The calibration method of the laser radar comprises the following steps: controlling the laser radar to scan a pre-constructed calibration site, and acquiring source point cloud data in a laser radar coordinate system. The calibration site comprises a plurality of signboards distributed in different spatial poses. The calibration site is constructed. A plurality of signboards are pre-set in the calibration site. The plurality of signboards are installed at different positions in front of the laser radar, and the heights of the plurality of signboards are different. The laser radar can be calibrated by increasing the rotatable multi-dimensional signboards, and the calibration accuracy can be further improved.

[0010] Based on the source point cloud data, signboard region information is acquired, and ground point cloud in the calibration site is extracted, that is, point cloud clusters belonging to each signboard are separated out by a point cloud segmentation algorithm to form the signboard region information. Meanwhile, ground points in the point cloud are extracted by using a plane fitting or filtering algorithm. The point cloud segmentation algorithm can be a clustering algorithm or a region growing algorithm. By clearly defining the point cloud regions of the signboards and the ground, the edge points at the intersection of the two can be accurately positioned. The plane fitting algorithm can be RANSAC (random sample consensus).

[0011] According to the signboard region information and the ground point cloud, height jump edge points in the calibration site are acquired. The algorithm can analyze point cloud sequences line by line according to the line bundle scanning characteristics of the laser radar. When a line bundle scans from a signboard facade to the ground in front of the signboard, the Y coordinates (height values) of adjacent laser points will experience a sharp and discontinuous jump (from the signboard height y1 to the ground height y2). Detecting these jump points is to locate the lower edge of the signboard and the projection edge of the signboard on the ground. Further, the algorithm complexity is reduced by converting the abstract plane constraint of the intersection of the “signboard plane” and the “ground plane” into specific and detectable point features, i.e. jump edge points.

[0012] According to the height jump edge points, a laser point cloud key point set is determined. The initially acquired jump edge points may contain noise, such as hitting grass and uneven ground. In this step, a robust algorithm such as RANSAC is used to purify these points. RANSAC distinguishes correct edge points from incorrect noise by iteratively fitting an ideal geometric model, and then forms a reliable laser point cloud key point set. The interference points caused by environmental noise and measurement errors can be effectively removed, and the data quality used for calibration calculation is ensured.

[0013] Obtain multiple real corner point coordinate information of the signboard in the vehicle coordinate system, and accurately measure the three-dimensional coordinates of specific corner points on each signboard in the vehicle coordinate system through high-precision external measurement equipment (such as GPS-RTK and total station). These coordinates serve as the "ground truth" for the calibration process, providing an accurate reference system for the entire calibration process. According to the key point set of the laser point cloud and the multiple real corner point coordinate information, determine the redundant correspondence between the key points of the laser point cloud and the real corner points; correlate the large number of automatically extracted laser key points (n in number) with the small number of manually measured real corner points (m in number, such as 24), forming a correspondence relationship with n much greater than m, and constructing an over-constrained system. When solving the over-constrained system, the least squares method can average out the influence of random measurement noise, thereby obtaining the optimal solution in a statistical sense and significantly improving the calibration accuracy.

[0014] According to the redundant correspondence relationship, convert the laser point cloud coordinates in the laser point cloud key point set to the rotation matrix and translation vector required for the transformation to the vehicle coordinate system, and obtain the calibration parameters of the laser radar according to the preset distance threshold, rotation matrix and translation vector. Using the redundant correspondence relationship, a set of rotation matrix R and translation vector T is determined by the least squares method (such as SVD decomposition) to minimize the overall error between all laser point cloud key points and the associated real corner points after transformation by P_vehicle=R×P_lidar+T. This realizes the automatic calculation from data to external parameter, and obtains the accurate transformation parameters required for converting the laser radar point cloud to the vehicle coordinate system. Using the preliminary solution of R and T, all laser key points are converted, and the 3D residual error of each point pair is calculated. By setting a distance threshold, abnormal point pairs with excessive residual error are removed. Then, using the cleaner point set after removal, R and T are recalculated. Through post-screening and iterative calculation, the accuracy and reliability of the final calibration parameters are further improved.

[0015] In some technical solutions, optionally, according to the signboard region information and the ground point cloud, the height jump edge points in the calibration site are obtained, including: based on the laser radar beam structure information, the height jump edge points are obtained from the adjacent place of the signboard region and the ground point cloud by analyzing the height jump of adjacent points.

[0016] In this technical solution, by directly analyzing the height difference of adjacent point clouds, the complex surface constraint of the intersection of the signboard and the ground is accurately positioned as a single-dimensional point feature, realizing intelligent dimension reduction and automatic extraction of constraints, and simplifying the process. The jump detection mechanism based on the beam sequence excludes the interference of non-edge regions, making the feature extraction process robust to environmental noise, ensuring the purity of key data and the accuracy and stability of the external parameter calibration result.

[0017] In some embodiments, before controlling the laser radar to scan the pre-constructed calibration site to obtain the source point cloud data in the laser radar coordinate system, the method for calibrating the laser radar comprises: obtaining the installation position of the laser radar, and establishing the laser radar coordinate system based on the installation position.

[0018] In this embodiment, before controlling the laser radar to scan the pre-constructed calibration site to obtain the source point cloud data in the laser radar coordinate system, the method for calibrating the laser radar comprises: obtaining the installation position of the laser radar, and establishing the laser radar coordinate system based on the installation position. This provides an exact mathematical expression basis for the subsequently obtained source point cloud data, so that all point cloud coordinates can be accurately expressed and subsequently processed in a unified reference system, avoids data processing confusion caused by ambiguous coordinate systems, and ensures the accuracy and stability of the external parameter calibration result.

[0019] In some embodiments, the laser point cloud key point set is determined according to the height jump edge points, comprising: fitting and purifying the height jump edge points by using a random sample consensus algorithm to form the laser point cloud key point set.

[0020] In this embodiment, the laser point cloud key point set is determined according to the height jump edge points, comprising: fitting and purifying the height jump edge points by using a random sample consensus algorithm to form the laser point cloud key point set. In this way, the abnormal jump points are filtered out while the real boundary features of the signboard and the ground are retained, so that the external parameter calculation result not only retains the statistical optimization characteristics of the least square method, but also has strong robustness against actual complex environmental interference.

[0021] In some embodiments, the calibration parameters of the laser radar are obtained according to the preset distance threshold, the rotation matrix and the translation vector, comprising: converting the laser point cloud key points according to the rotation matrix and the translation vector, and calculating the residual error between the laser point cloud key points and the real corner points; screening the residual error according to the preset distance threshold, and recalculating the rotation matrix and the translation vector based on the screened points to obtain the calibration parameters of the laser radar.

[0022] In the technical solution, the laser point cloud key points are converted according to the rotation matrix and the translation vector, and the residual error between the laser point cloud key points and the corresponding real corner points is calculated; the residual error is filtered according to the preset distance threshold, and the rotation matrix and the translation vector are recalculated based on the filtered point pairs to obtain the calibration parameters of the laser radar. The laser point cloud key points are converted using the rotation matrix and the translation vector obtained by preliminary solving, the quantitative evaluation of the calibration quality is realized by calculating the residual error between the laser point cloud key points and the real corner points, the abnormal matching point pairs are intelligently identified and removed according to the preset distance threshold, the noise points, the error matching and the local deformation and other interference factors not found in the early processing are eliminated, the external parameter is recalculated based on the high-quality point set after purification, the negative influence of the abnormal value on the overall solution is reduced while the effective data statistical characteristics are retained, the calibration precision and the robustness are cooperatively improved, and a complete calibration process with self-verification and self-optimization capabilities is formed.

[0023] In some technical solutions, optionally, the spatial pose distribution of the signboard in the calibration site includes different positions in the horizontal plane, different heights in the vertical direction, and rotation angles around different coordinate axes.

[0024] In the technical solution, the spatial pose distribution of the signboard in the calibration site includes different positions in the horizontal plane, different heights in the vertical direction, and rotation angles around different coordinate axes. The signboards with different heights and rotation angles provide rich pitch angle, roll angle and yaw angle constraint information for the calibration system, and significantly enhance the independent constraint ability of each degree of freedom in the rotation matrix.

[0025] In some technical solutions, optionally, the construction of the redundant corresponding relationship satisfies that the number of laser point cloud key points is greater than the number of real corner points in the real corner point coordinate information.

[0026] In the technical solution, by constructing the redundant corresponding relationship with the number of laser point cloud key points being significantly greater than the number of real corner points, the calibration problem is converted into an over-constrained mathematical system, so that the statistical characteristics of the least square method are used to realize effective suppression and averaging processing of random measurement noise, and the obtained rotation matrix and translation vector can reflect the overall optimal solution of all observation data, and the robustness and engineering practicability are enhanced.

[0027] According to a second aspect of the present application, a calibration device of a laser radar is provided. The laser radar is arranged on a vehicle. The calibration device of the laser radar comprises a first acquisition module, a second acquisition module, a third acquisition module, a first determination module, a fourth acquisition module, a second determination module, a third determination module and a fourth determination module. The first acquisition module is configured to control the laser radar to scan a pre-constructed calibration site to obtain source point cloud data in a laser radar coordinate system, wherein the calibration site comprises a plurality of markers distributed in different spatial poses. The second acquisition module is configured to obtain marker region information based on the source point cloud data, and extract ground point cloud in the calibration site. The third acquisition module is configured to obtain height jump edge points in the calibration site according to the marker region information and the ground point cloud. The first determination module is configured to determine a laser point cloud key point set according to the height jump edge points. The fourth acquisition module is configured to obtain a plurality of real corner point coordinate information of the markers in a vehicle coordinate system. The second determination module is configured to determine a redundant correspondence relationship between the laser point cloud key points and the real corner points according to the laser point cloud key point set and the plurality of real corner point coordinate information. The third determination module is configured to convert laser point cloud coordinates in the laser point cloud key point set to a rotation matrix and a translation vector required for coordinate conversion of the laser radar to the vehicle coordinate system according to the redundant correspondence relationship. The fourth determination module is configured to obtain calibration parameters of the laser radar according to a preset distance threshold, the rotation matrix and the translation vector.

[0028] The application provides a laser radar calibration device. The laser radar is arranged on a vehicle. The laser radar calibration device comprises a first acquisition module, a second acquisition module, a third acquisition module, a first determination module, a fourth acquisition module, a second determination module, a third determination module and a fourth determination module. The first acquisition module is configured to control the laser radar to scan a pre-constructed calibration site to obtain source point cloud data in a laser radar coordinate system. The calibration site comprises a plurality of signboards arranged in different spatial poses. The second acquisition module is configured to obtain signboard region information based on the source point cloud data and extract ground point cloud in the calibration site. The third acquisition module is configured to obtain height jump edge points in the calibration site according to the signboard region information and the ground point cloud. The first determination module is configured to determine a laser point cloud key point set according to the height jump edge points. The fourth acquisition module is configured to obtain a plurality of real corner point coordinate information of the signboard in a vehicle coordinate system. The second determination module is configured to determine a redundant corresponding relationship between the laser point cloud key points and the real corner points according to the laser point cloud key point set and the plurality of real corner point coordinate information. The third determination module is configured to convert laser point cloud coordinates in the laser point cloud key point set to a rotation matrix and a translation vector required for coordinate conversion of the laser radar to the vehicle coordinate system according to the redundant corresponding relationship. The fourth determination module is configured to obtain calibration parameters of the laser radar according to a preset distance threshold, the rotation matrix and the translation vector. A rotation matrix R and a translation vector T are found by the least square method, so that the overall error between all laser point cloud key points after transformation and the real corner points associated with them is minimized.

[0029] According to a third aspect of the application, a laser radar calibration device is provided, comprising a processor and a memory, the memory storing programs or instructions, and the processor implementing the steps of the laser radar calibration method according to any one of the above technical solutions when executing the programs or instructions in the memory. Therefore, the laser radar calibration device has all the beneficial effects of the laser radar calibration method according to any one of the above technical solutions.

[0030] According to a fourth aspect of the application, a readable storage medium is provided, the readable storage medium storing programs or instructions, and the programs or instructions being executed by a processor to implement the steps of the laser radar calibration method according to any one of the above technical solutions. Therefore, the readable storage medium has all the beneficial effects of the laser radar calibration method according to any one of the above technical solutions.

[0031] Additional aspects and advantages of the application will become apparent in the light of the following description and by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0032] The above and / or additional aspects and advantages of the application will become apparent and be readily understood by considering the following detailed description, from which the above aspects and advantages will become apparent and be readily understood in view of the accompanying drawings. Figure 1 A flow chart of a calibration method of a lidar provided in some embodiments of the present application is shown; Figure 2 A structural block diagram of a calibration device of a lidar provided in some embodiments of the present application is shown; Figure 3 A structural block diagram of a calibration device of a lidar provided in some embodiments of the present application is shown; Figure 4 A schematic diagram of a lidar coordinate system of a calibration method of a lidar provided in some embodiments of the present application is shown; Figure 5 A schematic diagram of a calibration site in some embodiments of the present application is shown; Figure 6 A schematic diagram of a calibration site in some embodiments of the present application is shown. DETAILED DESCRIPTION

[0033] In order to enable a clearer understanding of the above-mentioned objects, features and advantages of the present application, the present application will be further described below with reference to the drawings and specific embodiments. It should be noted that the features in the embodiments and examples can be combined with each other as long as they do not conflict with each other.

[0034] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, however, the present application can also be implemented in other ways different from those described herein, therefore, the scope of protection of the present application is not limited by the specific embodiments disclosed below.

[0035] The following refers to Figures 1 to 6 A calibration method, device and readable storage medium of a lidar according to some embodiments of the present application are described.

[0036] As Figure 1 shown, a calibration method of a lidar provided in embodiments of the present application, the lidar is arranged on a vehicle, the steps of the calibration method of the lidar include: Step 102, control the lidar to scan the pre-constructed calibration site, and obtain the source point cloud data under the lidar coordinate system, wherein the calibration site contains a plurality of identification boards distributed in different spatial poses; Step 104, based on the source point cloud data, obtain the identification board region information, and extract the ground point cloud in the calibration site; Step 106, according to the identification board region information and the ground point cloud, obtain the height jump edge point in the calibration site; Step 108, according to the height jump edge point, determine the key point set of the laser point cloud; In step 110, the multiple real corner point coordinate information of the signboard in the vehicle coordinate system is obtained. In step 112, the redundant correspondence relationship between the laser point cloud key points and the real corner points is determined according to the laser point cloud key point set and the multiple real corner point coordinate information. In step 114, the rotation matrix and the translation vector required for converting the laser point cloud coordinates in the laser point cloud key point set to the vehicle coordinate system are obtained according to the redundant correspondence relationship. In step 116, the calibration parameters of the laser radar are obtained according to the preset distance threshold, the rotation matrix and the translation vector.

[0037] The laser radar calibration method provided in the application, the laser radar is arranged on a vehicle, and the laser radar calibration method comprises the following steps: controlling the laser radar to scan a pre-constructed calibration site to obtain source point cloud data in a laser radar coordinate system, wherein the calibration site contains multiple signs distributed in different spatial poses, the construction of the calibration site is completed, multiple signs are pre-set in the calibration site, the multiple signs are installed at different positions in front of the laser radar, and the installation heights of the multiple signs are also different. The laser radar can be calibrated by increasing the rotatable multi-dimensional signs, and the calibration accuracy can be further improved.

[0038] Based on the source point cloud data, the signboard region information is obtained, and the ground point cloud in the calibration site is extracted, that is, the point cloud clusters belonging to each signboard are separated out by a point cloud segmentation algorithm to form the "signboard region information". At the same time, the ground points in the point cloud are extracted by using a plane fitting or filtering algorithm. The point cloud segmentation algorithm can be a clustering algorithm or a region growing algorithm. By clearly defining the point cloud regions of the signboard and the ground, the edge points at the intersection of the two can be accurately positioned. The plane fitting algorithm can be RANSAC (random sample consensus).

[0039] According to the signboard region information and the ground point cloud, the height jump edge points in the calibration site are obtained. According to the characteristics of the line bundle scanning of the laser radar, the algorithm analyzes the point cloud sequence line by line. When a line bundle scans from the facade of the signboard to the ground in front of it, the Y coordinates (height values) of the adjacent laser points will have a sharp and discontinuous jump (from the signboard height y1 to the ground height y2). Detecting these jump points is to locate the lower edge of the signboard and its projection edge on the ground. Further, the face constraint of the abstract "signboard plane" intersecting with the "ground plane" is converted into specific and detectable point features, i.e., jump edge points, and the algorithm complexity is reduced.

[0040] According to the height jump edge points, a set of laser point cloud key points is determined. The initially obtained jump edge points may contain noise, such as hitting weeds and uneven ground. In this step, a robust algorithm such as RANSAC is used to purify these points. RANSAC distinguishes correct edge points from false noise by iteratively fitting an ideal geometric model, thereby forming a reliable set of laser point cloud key points. This can effectively eliminate interference points caused by environmental noise and measurement errors, ensuring the quality of data used for calibration calculation.

[0041] A plurality of real corner point coordinate information of the signboard in the vehicle coordinate system is obtained. The three-dimensional coordinates of specific corner points on each signboard are accurately measured in the vehicle coordinate system by a high-precision external measurement device. These coordinates serve as the "ground truth" for calibration, providing an accurate reference system for the entire calibration process. According to the laser point cloud key point set and the plurality of real corner point coordinate information, a redundant correspondence relationship between the laser point cloud key points and the real corner points is determined. The least squares method can average out the influence of random measurement noise when solving the over-constrained system, thereby obtaining an optimal solution in a statistical sense and significantly improving the calibration accuracy.

[0042] According to the redundant correspondence relationship, the rotation matrix and translation vector required to convert the laser point cloud coordinates in the laser point cloud key point set to the vehicle coordinate system are obtained. According to the pre-set distance threshold, rotation matrix, and translation vector, the calibration parameters of the laser radar are obtained. Using the redundant correspondence relationship, a set of rotation matrix R and translation vector T is determined by the least squares method (such as SVD decomposition) to minimize the overall error between all laser point cloud key points and the associated real corner points after transformation by P_vehicle = R x P_lidar + T. This realizes the automatic calculation from data to external parameter, and obtains the accurate transformation parameters required to convert the laser radar point cloud to the vehicle coordinate system. Using the initially solved R and T to convert all laser key points and calculate the 3D residual error of each point pair. By setting a distance threshold, those abnormal point pairs with too large residual error are removed. Then, using the cleaner point set after removal, R and T are recalculated. Through post-screening and iterative calculation, the accuracy and reliability of the final calibration parameters are further improved.

[0043] Specifically, the three-dimensional coordinates of specific corner points on each signboard are accurately measured in the vehicle coordinate system by a high-precision external measurement device. The actual distance from the signboard to the vehicle can be obtained by GPS positioning.

[0044] Specifically, the signboard can be a fixed-size signboard.

[0045] Specifically, a large number of automatically extracted laser key points (n in number) are associated with a small number of manually measured real corner points (m in number, such as 24), forming a correspondence relationship with n much greater than m, and a system of over-constraints is constructed.

[0046] In some embodiments, optionally, according to the signboard region information and the ground point cloud, the height jump edge points in the calibration site are obtained, including: based on the laser radar beam structure information, the height jump edge points are obtained from the adjacent place of the signboard region and the ground point cloud by analyzing the height jump of adjacent points.

[0047] In this embodiment, by directly analyzing the height difference of adjacent point clouds, the complex surface intersection of the signboard and the ground is accurately positioned as a single-dimensional point feature, realizing intelligent dimension reduction and automatic extraction of constraints, and simplifying the process. The jump detection mechanism based on the beam sequence excludes the interference of non-edge regions, making the feature extraction process have strong robustness to environmental noise, ensuring the purity of the key data and the accuracy and stability of the external parameter calibration result.

[0048] In some embodiments, optionally, before controlling the laser radar to scan the pre-constructed calibration site to obtain the source point cloud data in the laser radar coordinate system, the calibration method of the laser radar comprises: obtaining the installation position of the laser radar, and establishing the laser radar coordinate system based on the installation position.

[0049] In this embodiment, before controlling the laser radar to scan the pre-constructed calibration site to obtain the source point cloud data in the laser radar coordinate system, the calibration method of the laser radar comprises: obtaining the installation position of the laser radar, and establishing the laser radar coordinate system based on the installation position. The exact mathematical expression basis is provided for the subsequently obtained source point cloud data, so that all point cloud coordinates can be accurately expressed and subsequently processed in a unified reference system, avoiding data processing confusion caused by ambiguous coordinate systems, and ensuring the accuracy and stability of the external parameter calibration result.

[0050] In some embodiments, optionally, according to the height jump edge points, the laser point cloud key point set is determined, including: fitting and purifying the height jump edge points by using a random sample consensus algorithm to form the laser point cloud key point set.

[0051] In this embodiment, according to the height jump edge points, the laser point cloud key point set is determined, including: fitting and purifying the height jump edge points by using a random sample consensus algorithm to form the laser point cloud key point set. Thus, while retaining the real intersection features of the signboard and the ground, abnormal jump points are filtered out, ensuring that the external parameter calculation result not only retains the statistical optimization characteristics of the least squares method, but also has strong robustness against actual complex environmental interference.

[0052] In some embodiments, optionally, the obtaining of the calibration parameters of the laser radar according to the preset distance threshold, the rotation matrix and the translation vector comprises: converting the laser point cloud key points according to the rotation matrix and the translation vector, and calculating the residual error between the laser point cloud key points and the corresponding real corner points; screening the residual error according to the preset distance threshold, and recalculating the rotation matrix and the translation vector based on the screened point pairs to obtain the calibration parameters of the laser radar.

[0053] In this embodiment, the laser point cloud key points are converted according to the rotation matrix and the translation vector, and the residual error between the laser point cloud key points and the corresponding real corner points is calculated; the residual error is screened according to the preset distance threshold, and the rotation matrix and the translation vector are recalculated based on the screened point pairs to obtain the calibration parameters of the laser radar. The rotation matrix and the translation vector obtained by preliminary solving are used for coordinate conversion of the laser point cloud key points, the residual error between the laser point cloud key points and the real corner points is calculated to realize quantitative evaluation of the calibration quality, the abnormal matching point pairs are intelligently identified and removed according to the preset distance threshold, the noise points, error matching and local deformation and other interference factors not found in the early processing are eliminated, the external parameter parameters are recalculated based on the purified high-quality point set, the effective data statistical characteristics are retained, the negative influence of abnormal values on the overall solution is reduced, the calibration precision and the robustness are cooperatively improved, and a complete calibration process with self-verification and self-optimization capabilities is formed.

[0054] In some embodiments, optionally, the spatial pose distribution of the signboard in the calibration site comprises: different positions in the horizontal plane, different heights in the vertical direction, and rotation angles around different coordinate axes.

[0055] In this embodiment, the spatial pose distribution of the signboard in the calibration site comprises: different positions in the horizontal plane, different heights in the vertical direction, and rotation angles around different coordinate axes. The signboards of different heights and rotation angles provide rich pitch angle, roll angle and yaw angle constraint information for the calibration system, and significantly enhance the independent constraint ability of each degree of freedom in the rotation matrix.

[0056] In some embodiments, optionally, the construction of the redundant corresponding relationship satisfies that the number of the laser point cloud key points is greater than the number of the real corner points in the real corner point coordinate information.

[0057] In this embodiment, by constructing the redundant corresponding relationship with the number of laser point cloud key points being significantly greater than the number of real corner points, the calibration problem is converted into an over-constrained mathematical system, so that the statistical characteristics of the least square method are used to realize effective suppression and averaging processing of random measurement noise, so that the obtained rotation matrix and translation vector can reflect the overall optimal solution of all observation data, and the robustness and engineering practicability are enhanced.

[0058] As Figure 2As shown, the embodiment of the present application provides a calibration device 200 of a laser radar, the laser radar is arranged on a vehicle, and the calibration device 200 of the laser radar comprises a first acquisition module 210, a second acquisition module 220, a third acquisition module 230, a first determination module 240, a fourth acquisition module 250, a second determination module 260, a third determination module 270 and a fourth determination module 280. The first acquisition module 210 is configured to control the laser radar to scan a pre-constructed calibration site, and acquire source point cloud data in a laser radar coordinate system, wherein the calibration site comprises a plurality of identification boards distributed in different spatial poses; the second acquisition module 220 is configured to acquire identification board region information based on the source point cloud data, and extract ground point cloud in the calibration site; the third acquisition module 230 is configured to acquire height jump edge points in the calibration site according to the identification board region information and the ground point cloud; the first determination module 240 is configured to determine a laser point cloud key point set according to the height jump edge points; the fourth acquisition module 250 is configured to acquire a plurality of real corner point coordinate information of the identification board in a vehicle coordinate system; the second determination module 260 is configured to determine a redundant corresponding relationship between the laser point cloud key points and the real corner points according to the laser point cloud key point set and the plurality of real corner point coordinate information; the third determination module 270 is configured to convert laser point cloud coordinates in the laser point cloud key point set to a rotation matrix and a translation vector required by the vehicle coordinate system according to the redundant corresponding relationship; and the fourth determination module 280 is configured to acquire calibration parameters of the laser radar according to a preset distance threshold, the rotation matrix and the translation vector.

[0059] The application provides a laser radar calibration device 200, the laser radar is arranged on a vehicle, and the laser radar calibration device 200 comprises a first acquisition module 210, a second acquisition module 220, a third acquisition module 230, a first determination module 240, a fourth acquisition module 250, a second determination module 260, a third determination module 270 and a fourth determination module 280. The first acquisition module 210 is used for controlling the laser radar to scan a pre-constructed calibration site, and acquiring source point cloud data in a laser radar coordinate system, wherein the calibration site comprises a plurality of signboards distributed in different spatial poses; the second acquisition module 220 is used for acquiring signboard region information based on the source point cloud data, and extracting ground point cloud in the calibration site; the third acquisition module 230 is used for acquiring height jump edge points in the calibration site according to the signboard region information and the ground point cloud; the first determination module 240 is used for determining a laser point cloud key point set according to the height jump edge points; the fourth acquisition module 250 is used for acquiring a plurality of real corner point coordinate information of the signboard in a vehicle coordinate system; the second determination module 260 is used for determining a redundant corresponding relationship between the laser point cloud key points and the real corner points according to the laser point cloud key point set and the plurality of real corner point coordinate information; the third determination module 270 is used for converting laser point cloud coordinates in the laser point cloud key point set to a rotation matrix and a translation vector required by the laser radar calibration device 200 according to the redundant corresponding relationship; and the fourth determination module 280 is used for acquiring the calibration parameters of the laser radar according to a preset distance threshold, the rotation matrix and the translation vector. A rotation matrix R and a translation vector T are found by using the least square method, so that the overall error between all the laser point cloud key points after transformation and the real corner points associated with the laser point cloud key points is minimized.

[0060] As shown in Figure 3 The embodiment of the application provides a laser radar calibration device 300, which comprises a processor 310 and a memory 320, the memory 320 stores programs or instructions, and the processor 310 implements the steps of the laser radar calibration method of any one of the above-mentioned embodiments when executing the programs or instructions in the memory 320. Therefore, the laser radar calibration device 300 has all the beneficial effects of the laser radar calibration method of any one of the above-mentioned embodiments.

[0061] The embodiment of the application provides a readable storage medium, the readable storage medium stores programs or instructions, and the programs or instructions are executed by a processor to implement the steps of the laser radar calibration method of any one of the above-mentioned embodiments. Therefore, the readable storage medium has all the beneficial effects of the laser radar calibration method of any one of the above-mentioned embodiments.

[0062] As shown in Figure 4 A coordinate system is established by the laser radar installed at the front of the vehicle, the X axis points to the right side of the vehicle head, the Y axis points to the sky, and the Z axis points to the direction of the vehicle head.

[0063] The vehicle can be a wide-body vehicle, a mining truck, an agricultural machine, and a terrain exploration robot.

[0064] As shown in Figure 5 The installation heights of the A signboard, the B signboard, the C signboard, and the D signboard are different, that is, the heights are z1, z2, z3, and z4 respectively.

[0065] As shown in Figure 6 E is the ground projection point of the F signboard on the ground.

[0066] It should be noted that in the claims, the specification and the drawings of the present application, the term "multiple" refers to two or more, unless otherwise explicitly limited, and the terms "upper", "lower", and the like indicate relative positions or orientation relationships based on the positions or orientation relationships shown in the drawings, and are only for the convenience of describing the present application and making the description process more simple, and are not intended to indicate or imply that the devices or elements referred to must have the specific orientation described, and therefore these descriptions cannot be understood as limitations on the present application; the terms "connection", "installation", "fixation" and the like should be understood broadly, for example, "connection" can be a fixed connection between objects, or a detachable connection between objects, or an integral connection; it can be a direct connection between objects, or an indirect connection between objects through an intermediate medium. For those skilled in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0067] In the claims, the specification and the drawings of the present application, the description of the terms "one embodiment", "some embodiments", "a specific embodiment" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the claims, the specification and the drawings of the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0068] The above is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of calibrating a lidar, the method comprising: The laser radar is arranged on a vehicle, and a calibration method of the laser radar comprises the following steps: controlling the laser radar to scan a pre-constructed calibration site to obtain source point cloud data in a laser radar coordinate system, wherein the calibration site comprises a plurality of markers distributed in different spatial poses; based on the source point cloud data, obtaining marker region information and extracting ground point cloud in the calibration site; based on the marker region information and the ground point cloud, obtaining height jump edge points in the calibration site; based on the height jump edge points, determining a laser point cloud key point set; obtaining a plurality of real corner point coordinate information of the markers in a vehicle coordinate system; based on the laser point cloud key point set and the plurality of real corner point coordinate information, determining a redundant correspondence relationship between the laser point cloud key points and the real corner points; based on the redundant correspondence relationship, converting laser point cloud coordinates in the laser point cloud key point set to a rotation matrix and a translation vector required by the vehicle coordinate system; based on a preset distance threshold, the rotation matrix and the translation vector, obtaining calibration parameters of the laser radar.

2. The calibration method of a lidar according to claim 1, wherein, The method comprises the following steps: based on laser radar line bundle structure information, the height jump edge points are obtained from the adjoining part of the marker region and the ground point cloud by analyzing the height jump of adjacent points.

3. The method of calibrating a lidar according to claim 1, wherein, Before the step of controlling the laser radar to scan the pre-constructed calibration site to obtain the source point cloud data in the laser radar coordinate system, the calibration method of the laser radar comprises the following steps: obtaining an installation position of the laser radar, and establishing a laser radar coordinate system based on the installation position.

4. The method of calibrating a lidar according to claim 1, wherein, The method comprises the following steps: the height jump edge points are fitted and purified by using a random sample consensus algorithm to form the laser point cloud key point set.

5. The method of calibrating a lidar according to claim 1, wherein, The method comprises the following steps: the laser point cloud key points are converted according to the rotation matrix and the translation vector, and the residual error between the laser point cloud key points and the real corner points is calculated; the residual error is screened according to the preset distance threshold, and the rotation matrix and the translation vector are recalculated based on the screened point pairs to obtain the calibration parameters of the laser radar.

6. The method of calibrating a lidar according to claim 1, wherein, The spatial pose distribution of the markers in the calibration site comprises different positions on a horizontal plane, different heights in a vertical direction and rotation angles around different coordinate axes.

7. The method of calibrating a lidar according to claim 1, wherein, The construction of the redundant correspondence relationship satisfies that the number of the laser point cloud key points is greater than the number of the real corner points in the real corner point coordinate information.

8. A calibration device for a lidar, characterized in that The laser radar is arranged on a vehicle, and a calibration device of the laser radar comprises the following steps: a first obtaining module is configured to control the laser radar to scan a pre-constructed calibration site to obtain source point cloud data in a laser radar coordinate system, wherein the calibration site comprises a plurality of markers distributed in different spatial poses; A second acquisition module is configured to acquire signboard region information based on the source point cloud data, and extract ground point cloud in a calibration site; A third acquisition module is configured to acquire height jump edge points in the calibration site according to the signboard region information and the ground point cloud; A first determination module is configured to determine a laser point cloud key point set according to the height jump edge points; A fourth acquisition module is configured to acquire a plurality of real corner point coordinate information of the signboard in a vehicle coordinate system; A second determination module is configured to determine a redundant correspondence relationship between laser point cloud key points and real corner points according to the laser point cloud key point set and the plurality of real corner point coordinate information; A third determination module is configured to convert laser point cloud coordinates in the laser point cloud key point set to a rotation matrix and a translation vector required by the vehicle coordinate system according to the redundant correspondence relationship; A fourth determination module is configured to acquire calibration parameters of the laser radar according to a preset distance threshold, the rotation matrix and the translation vector.

9. A calibration device for a lidar, characterized in that comprising: a processor; a memory, the memory storing programs or instructions, the processor executing the programs or instructions in the memory to implement the steps of the laser radar calibration method according to any one of claims 1 to 7.

10. A readable storage medium, characterized by, The readable storage medium stores programs or instructions, which are executed by the processor to implement the steps of the laser radar calibration method according to any one of claims 1 to 7.

Images