Vehicle-mounted laser radar and RTK external parameter calibration method and related device

Abstract

The invention provides a vehicle-mounted laser radar and RTK external parameter calibration method and a related device, and relates to the technical field of positioning, the vehicle-mounted laser radar and RTK external parameter calibration method is used for a vehicle, the vehicle carries a laser radar and a positioning navigation system, and the calibration method comprises the following steps: constructing a calibration field and a reference map, the calibration field comprises a retaining wall with a slope structure; when the to-be-calibrated vehicle is located in the calibration field, acquiring original point cloud data according to the laser radar, and acquiring RTK pose data according to the positioning navigation system; uniformly converting the original point cloud data and the RTK pose data into a coordinate system which is the same as a reference map; and on the basis of the original point cloud data and the RTK pose data, inversely solving spatial transformation parameters of the laser radar relative to the positioning navigation system. According to the method, high-precision and repeatable automatic calibration of the vehicle-mounted laser radar and the RTK external parameters is realized, and the method is particularly suitable for industrial environments with sparse characteristics, such as mining areas and ports.

Description

Technical Field

[0001] This application relates to the field of positioning technology, and more specifically, to a calibration method and related apparatus for vehicle-mounted lidar and RTK extrinsic parameters. Background Technology

[0002] In the fields of autonomous driving and high-precision positioning, the integrated application of vehicle-mounted LiDAR and real-time dynamic differential positioning systems is becoming increasingly widespread. A fundamental prerequisite for achieving accurate fusion of multi-sensor data is to accurately determine the relative spatial position and attitude relationship between the LiDAR and the RTK (Real-Time Kinematic) system, i.e., to perform extrinsic parameter calibration.

[0003] Currently, traditional external parameter calibration methods have many limitations. For example, manual calibration relies on specific targets such as calibration boards or checkerboard patterns, requiring on-site measurements and calculations by professionals. This method is cumbersome, inefficient, and susceptible to human error, making it difficult to meet the needs of rapid vehicle deployment or periodic calibration. Another approach is to use natural scenes (such as urban roads and buildings) for calibration, but these methods require a high degree of feature richness and stability. In scenes with sparse features or high repetition (such as flat ground or structured factory buildings), calibration accuracy and success rate will significantly decrease.

[0004] Therefore, how to propose a method that can achieve automated and high-precision external parameter calibration without relying on human intervention, without demanding complex natural environments, and in specific sites has become an urgent technical problem to be solved. Summary of the Invention

[0005] This application aims to address at least one of the technical problems existing in external parameter calibration methods, namely, cumbersome process, low efficiency, susceptibility to human factors, and significant decrease in calibration accuracy and success rate in scenarios with sparse features or high repeatability.

[0006] Therefore, the first aspect of this application provides a method for calibrating vehicle-mounted lidar and RTK extrinsic parameters.

[0007] The second aspect of this application provides a calibration device for vehicle-mounted lidar and RTK extrinsic parameters.

[0008] A third aspect of this application provides a readable storage medium.

[0009] The fourth aspect of this application provides an electronic device.

[0010] In view of this, the first aspect of this application proposes a calibration method for vehicle-mounted LiDAR and RTK extrinsic parameters, for use in vehicles equipped with LiDAR and a positioning and navigation system. The calibration method includes: constructing a calibration field and a reference map, wherein the calibration field includes retaining walls with sloping structures; when the vehicle to be calibrated is located in the calibration field, acquiring raw point cloud data based on the LiDAR and obtaining RTK pose data based on the positioning and navigation system; uniformly converting the raw point cloud data and RTK pose data to the same coordinate system as the reference map; and based on the raw point cloud data and RTK pose data, inversely solving the spatial transformation parameters of the LiDAR relative to the positioning and navigation system.

[0011] The calibration method for vehicle-mounted LiDAR and RTK extrinsic parameters proposed in this application is used in vehicles equipped with LiDAR and a positioning and navigation system. First, a calibration field and a reference map are constructed. The key facility in the calibration field is a retaining wall with a sloping structure. This sloping structure provides the LiDAR with a continuously changing and distinctive three-dimensional surface, avoiding the feature degradation problem caused by flat walls. During the calibration process, the vehicle enters the calibration field, and the LiDAR on the vehicle begins to collect raw point cloud data of the surrounding environment. Simultaneously, the vehicle's positioning and navigation system collects high-precision pose data of the vehicle itself. The positioning and navigation system includes an RTK-GNSS / INS (Real-time kinematic - Global Navigation Satellite System / Inertial Navigation System) integrated navigation system. Subsequently, all data is uniformly transformed to a common coordinate system, which is consistent with the coordinate system of the reference map pre-constructed using high-precision measurement methods. Finally, through algorithmic processing, the point cloud collected by the LiDAR is precisely registered with the reference map. By comparing the RTK pose data provided by the vehicle positioning and navigation system with the LiDAR pose data in the reference map coordinate system obtained after point cloud registration optimization, the spatial transformation parameters of the LiDAR relative to the positioning and navigation system can be deduced. This allows for the high-precision deduction of the spatial transformation parameters of the LiDAR relative to the positioning and navigation system, thus calculating the required extrinsic parameters for calibration. This application utilizes artificially designed retaining walls with stable geometric features, reducing dependence on the natural environment and improving the automation level and robustness of the calibration process. Simultaneously, it achieves high-precision, repeatable, automated calibration of vehicle-mounted LiDAR and RTK extrinsic parameters without relying on manual on-site intervention or complex natural scenes, making it particularly suitable for feature-sparse industrial environments such as mining areas and ports.

[0012] Optionally, in some embodiments, the steps of constructing the calibration field and the reference map include: constructing the calibration field based on the retaining wall; determining a geographic reference point within the calibration field and establishing a northeast-sky coordinate system with the geographic reference point as the origin; collecting point cloud data of the calibration field through a data acquisition carrier; and converting the point cloud data of the calibration field into the northeast-sky coordinate system to generate a reference map of the calibration field.

[0013] In this embodiment, a physical calibration field is first constructed using a retaining wall with a sloping structure. This artificial structure provides the lidar with a stable and distinctive 3D scanning target, effectively solving the problem of obtaining effective point cloud data in flat mining areas with sparse natural features. Subsequently, within the constructed calibration field, a geographic reference point is determined using high-precision measurement methods. For example, long-term static observation post-processing GNSS (Global Navigation Satellite System) technology is used to determine the geographic reference point, and this point is used as the origin to establish a northeast-sky coordinate system. The X-axis of the northeast-sky coordinate system points east, the Y-axis points north, and the Z-axis points vertically upward, serving as a unified spatial reference. Next, a specialized acquisition carrier scans the entire calibration field, acquiring the raw point cloud data. Finally, the acquired calibration field point cloud data is unified into the previously established northeast-sky coordinate system through coordinate transformation, thereby generating a high-precision reference map with absolute geographic coordinates. As can be seen, this application ensures the quality of lidar scanning data and the strength of geometric constraints through artificially designed structured retaining walls, avoiding the risk of calibration failure in open areas. Simultaneously, by establishing a unified northeast-southeast coordinate system and generating a high-precision reference map, a reliable and consistent truth benchmark is provided for the subsequent accurate fusion and automated calibration of multi-sensor data, greatly improving the robustness of the calibration process and the repeatability of the results.

[0014] Optionally, in some embodiments, the step of collecting point cloud data of the calibration field by means of a data acquisition vehicle includes: collecting point cloud data of the calibration field by means of a professional mapping vehicle; and / or collecting point cloud data of the calibration field by means of UAV aerial survey; and / or collecting point cloud data of the calibration field by means of a handheld scanner; and / or collecting point cloud data of the calibration field by means of a fixed ground laser scanning method.

[0015] In this embodiment, point cloud data of the calibration field is collected using at least one of the following methods: a specialized mapping vehicle, UAV aerial surveying, a handheld scanner, and a fixed ground-based laser scanner. The specialized mapping vehicle is typically equipped with a high-line-count lidar and a high-precision GNSS / INS (Global Navigation Satellite System / Inertial Navigation System) integrated navigation system. The vehicle travels along a preset path within the calibration field to perform a comprehensive and continuous scan of the retaining wall and its surrounding environment. The UAV aerial surveying system uses a UAV equipped with lidar or an oblique photography camera to fly over the calibration field, quickly acquiring large-area, high-density 3D point cloud data from an aerial perspective, particularly suitable for tall retaining walls or areas difficult for vehicles to access. The handheld or backpack mobile scanner is carried by an operator, a lightweight device integrating lidar and an inertial measurement unit, moving within the calibration field to collect detailed information about the retaining wall surface at close range and high resolution. Fixed ground laser scanners perform static scanning at different locations within the site, then precisely stitch together point cloud data from multiple stations to obtain highly accurate and low-noise point cloud data. These acquisition carriers all use pre-determined geographical reference points as absolute coordinate benchmarks, ensuring that the acquired point clouds can be accurately converted to a unified northeast-southeast coordinate system. Therefore, this application provides diverse reference map construction methods adaptable to different application scenarios and cost budgets. It can obtain high-precision maps using existing professional equipment and can also be quickly deployed through more flexible solutions, ensuring the universality and feasibility of this method under different mining area conditions, ultimately providing a reliable true benchmark for automatic external parameter calibration.

[0016] Optionally, in some embodiments, the step of inversely solving the spatial transformation parameters of the lidar relative to the positioning and navigation system based on the original point cloud data and RTK pose data includes: converting the RTK pose data into a homogeneous transformation matrix in the northeast-northeast coordinate system; setting initial extrinsic parameters, and projecting the original point cloud data onto the northeast-northeast coordinate system according to the initial extrinsic parameters to obtain initial projected point cloud data; spatially aligning the initial projected point cloud data with a reference map using a point cloud registration algorithm to obtain a fine registration transformation matrix; and inversely solving the spatial transformation parameters of the lidar relative to the positioning and navigation system based on the fine registration transformation matrix and the RTK pose data.

[0017] In this embodiment, the RTK pose data acquired by the positioning and navigation system, including latitude, longitude, altitude coordinates, and roll, pitch, and yaw angles, is first converted into a homogeneous transformation matrix in the northeast-northeast coordinate system. This homogeneous transformation matrix fully describes the vehicle's position and attitude in the reference map coordinate system. Next, an initial extrinsic parameter estimate can be set based on the installation relationship between the LiDAR and the positioning and navigation system. Using this initial extrinsic parameter estimate, the raw point cloud data collected by the LiDAR is converted to the northeast-northeast coordinate system to obtain initial projected point cloud data. Then, a high-precision point cloud registration algorithm is used to perform precise spatial alignment calculations between the initial projected point cloud data and a pre-constructed high-precision reference map, solving for a fine registration transformation matrix that best matches the initial projected point cloud onto the reference map. Finally, based on the fine registration transformation matrix and the known RTK pose data, the precise spatial transformation parameters of the LiDAR relative to the positioning and navigation system, i.e., the extrinsic parameters to be calibrated, are calculated through mathematical inverse kinematics. As can be seen, this application compares sensor observation data with a high-precision ground truth map through coordinate transformation and iterative optimization registration, thereby accurately separating and solving the relative pose relationship between sensors, realizing fully automatic and high-precision extrinsic parameter calibration, avoiding the dependence on manual measurement in traditional methods, and ensuring the reliability and repeatability of calibration results.

[0018] Optionally, in some embodiments, when the vehicle to be calibrated is located in the calibration field, the steps of collecting raw point cloud data based on the lidar and obtaining RTK pose data based on the positioning and navigation system include: when the vehicle to be calibrated is located in the calibration field, parking the vehicle to be calibrated sequentially at multiple pre-planned stationary points in the calibration field; at each stationary point, collecting raw point cloud data based on the lidar and obtaining RTK pose data based on the positioning and navigation system; wherein both the raw point cloud data and the RTK pose data carry a timestamp.

[0019] In this embodiment, after the vehicle to be calibrated (such as a mining dump truck) enters the pre-constructed calibration field, the operator moves the vehicle sequentially and precisely stops it at multiple pre-planned stationary points within the calibration field according to a preset plan. At each stationary point, the vehicle needs to come to a complete stop before data recording begins. The collected data includes two types: first, raw point cloud data obtained by the vehicle-mounted LiDAR over several seconds; and second, high-precision RTK pose data output by the vehicle's positioning and navigation system. The RTK pose data includes latitude, longitude, and altitude coordinates, as well as attitude information such as roll angle, pitch angle, and heading angle. Throughout the data acquisition process, the RTK pose data must remain in a fixed solution state. In this state, by calculating the correction signal from the ground base station, centimeter-level ultra-high precision positioning data can be obtained to ensure the high reliability of the pose data. Furthermore, both the raw point cloud data collected by the LiDAR and the RTK pose data acquired by the positioning and navigation system must carry precise timestamps, providing a necessary foundation for subsequent data synchronization and the establishment of frame correspondence. By employing a multi-pose static acquisition strategy, the observability of each degree of freedom of the extrinsic parameters can be effectively enhanced, avoiding instantaneous errors that may be introduced by vehicle movement during dynamic acquisition. Simultaneously, discrete static point data facilitates independent quality checks and filtering. If the RTK signal at a particular point is lost or the point cloud quality is poor, it can be removed individually without affecting the overall process, thereby improving the robustness of the calibration process and the accuracy of the final results. The static acquisition mode is particularly suitable for scenarios requiring high initial calibration accuracy or needing to verify dynamic calibration results.

[0020] Optionally, in some embodiments, when the vehicle to be calibrated is located in the calibration field, the steps of collecting raw point cloud data based on the lidar and obtaining RTK pose data based on the positioning and navigation system include: when the vehicle to be calibrated is located in the calibration field, causing the vehicle to be calibrated to travel along a preset trajectory; during the travel, the lidar and the positioning and navigation system continuously collect raw data; and filtering the raw point cloud data and RTK pose data from the continuously collected raw data; wherein both the raw point cloud data and the RTK pose data carry a timestamp.

[0021] In this embodiment, after the vehicle to be calibrated (such as a mining dump truck) enters the pre-constructed calibration field, the operator controls the vehicle to travel at a constant or variable speed along a preset trajectory, for example, circling around the calibration field's circular road or other paths that cover different viewpoints. Throughout the vehicle's journey, the onboard LiDAR and positioning and navigation system work synchronously, continuously collecting raw data. The LiDAR continuously scans the surrounding environment, generating a continuous point cloud sequence containing structures such as the calibration field retaining walls; the positioning and navigation system continuously outputs the vehicle's RTK pose data, which includes latitude, longitude, elevation coordinates, roll angle, pitch angle, and heading angle. Data frames that meet the requirements are then selected based on preset conditions. These preset conditions may require the LiDAR point cloud to clearly scan a sufficiently large area of ​​the retaining wall surface. Both the selected raw point cloud data and the corresponding RTK pose data carry precise timestamps to ensure synchronization and correspondence between the data. This application's dynamic acquisition mode can quickly acquire massive amounts of data from vehicles at different positions and attitudes, offering high efficiency and more closely reflecting the actual operating conditions of the vehicle. By filtering valid frames from continuous data, segments with instantaneous RTK signal loss or poor point cloud quality can be avoided, and point cloud data and RTK pose data for calibration can be selected. This improves acquisition efficiency while ensuring data quality, and is particularly suitable for scenarios that require rapid and batch calibration of vehicle fleets.

[0022] Optionally, static acquisition mode and dynamic acquisition mode are two complementary data acquisition methods provided by the present invention. Users can flexibly choose a single mode or a combination of modes to perform calibration according to actual conditions, efficiency requirements and accuracy targets.

[0023] Optionally, in some embodiments, the number of original point cloud data and RTK pose data are both multiple, and the calibration method further includes: calculating the spatial transformation parameters of the lidar relative to the positioning and navigation system at multiple independent observation times; solving for the optimal spatial transformation parameters by the least squares method, evaluating the quality of the calibration result of the optimal spatial transformation parameters, and outputting the extrinsic parameter matrix when the quality index meets the preset threshold.

[0024] In this embodiment, during data acquisition, raw point cloud data and corresponding RTK pose data at multiple independent observation times are acquired, such as at different stationary positions or different time points during dynamic movement of the mining truck within the calibration field. First, based on each frame of independent data—that is, the point cloud data and corresponding RTK pose data at a single moment—a spatial transformation parameter of the lidar relative to the positioning and navigation system is calculated according to the registration and inverse kinematics process. Then, the least squares method is used to jointly optimize all independently calculated spatial transformation parameters, aiming to find an optimal spatial transformation parameter that better fits all observation data, thereby effectively suppressing potential random errors in single-frame data and obtaining statistically more accurate and stable optimal extrinsic parameters. Finally, the calibration results of the optimal spatial transformation parameter are evaluated for quality, with quality indicators including the consistency of parameter estimates across multiple frames and the accuracy of point cloud registration. The consistency of multi-frame parameter estimates includes the standard deviations of translation and rotation components. Point cloud registration accuracy includes calculating the average residual between the finely registered point cloud and the reference map, i.e., the average distance from a point to the corresponding model surface. These are compared with preset thresholds based on quality indicators; for example, the standard deviation of translation is required to be less than 5 cm, and the standard deviation of rotation less than 0.5 degrees. Only when all quality indicators meet the preset threshold requirements is the calibration considered successful, and the optimized extrinsic parameter matrix is ​​finally output. If any indicator fails to meet the requirements, a calibration failure is indicated, and data checking or re-acquisition is recommended. This application significantly improves the accuracy and robustness of extrinsic parameter calibration results through joint optimization of multi-frame data and a rigorous quality control process. It effectively avoids calibration errors caused by poor single-frame data quality, such as those caused by instantaneous RTK signal fluctuations or partial point cloud occlusion. It effectively prevents the application of defective or inaccurate calibration results to actual vehicles, thereby ensuring the positioning accuracy and driving safety of the multi-sensor fusion system. It is particularly suitable for batch calibration scenarios of unmanned mining trucks with extremely high reliability requirements.

[0025] According to a second aspect of this application, a calibration device for vehicle-mounted LiDAR and RTK extrinsic parameters is also proposed for use in a vehicle equipped with a LiDAR and a positioning and navigation system. The calibration device includes: a construction unit for constructing a calibration field and a reference map, wherein the calibration field includes a retaining wall with a sloping structure; an acquisition unit for acquiring raw point cloud data based on the LiDAR and obtaining RTK pose data based on the positioning and navigation system when the vehicle to be calibrated is located in the calibration field; a transformation unit for uniformly transforming the raw point cloud data and RTK pose data to the same coordinate system as the reference map; and a calculation unit for inversely solving the spatial transformation parameters of the LiDAR relative to the positioning and navigation system based on the raw point cloud data and RTK pose data.

[0026] The calibration device for vehicle-mounted LiDAR and RTK extrinsic parameters proposed in this application is used in a vehicle equipped with a LiDAR and a positioning and navigation system. First, a calibration field and reference map are constructed by a construction unit. The key facility of the calibration field is a retaining wall with a sloping structure. The sloping structure provides a continuously changing and distinctive three-dimensional curved surface for LiDAR scanning, avoiding the feature degradation problem caused by flat walls. During the calibration process, the vehicle to be calibrated enters the calibration field, and the acquisition unit begins to collect raw point cloud data of the surrounding environment through the vehicle's LiDAR. Simultaneously, the acquisition unit collects high-precision pose data of the vehicle itself through the vehicle's positioning and navigation system. The positioning and navigation system includes an RTK-GNSS / INS integrated navigation system. Subsequently, a conversion unit converts all data to a common coordinate system, which is consistent with the coordinate system of the reference map pre-constructed using high-precision measurement methods. Finally, the point cloud collected by the LiDAR is precisely registered with the reference map. The computing unit compares the RTK pose data provided by the vehicle positioning and navigation system with the LiDAR pose data in the reference map coordinate system obtained after point cloud registration optimization to inversely solve the spatial transformation parameters of the LiDAR relative to the positioning and navigation system. This allows for the high-precision inverse solution of the spatial transformation parameters of the LiDAR relative to the positioning and navigation system, thus calculating the required extrinsic parameters for calibration. This application utilizes artificially designed retaining walls with stable geometric features, reducing dependence on the natural environment and improving the automation level and robustness of the calibration process. Simultaneously, it achieves high-precision, repeatable, automated calibration of vehicle-mounted LiDAR and RTK extrinsic parameters without relying on manual on-site intervention or complex natural scenes, making it particularly suitable for sparsely characterized industrial environments such as mining areas and ports.

[0027] According to a third aspect of this application, a readable storage medium is also proposed, on which a program or instructions are stored, which, when executed by a processor, implement the steps of the calibration method as proposed in the first aspect. Therefore, it possesses all the beneficial effects of the calibration method for vehicle-mounted LiDAR and RTK extrinsic parameters.

[0028] According to a fourth aspect of this application, an electronic device is also proposed, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the calibration method as proposed in the first aspect. Therefore, it possesses all the beneficial effects of the calibration method for vehicle-mounted LiDAR and RTK extrinsic parameters.

[0029] Additional aspects and advantages of this application will become apparent in the following description or may be learned by practice of this application. Attached Figure Description

[0030] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0031] Figure 1 This document illustrates one of the flowcharts illustrating the calibration method for vehicle-mounted lidar and RTK extrinsic parameters in one embodiment of this application.

[0032] Figure 2 This is a second schematic flowchart illustrating a calibration method for vehicle-mounted lidar and RTK extrinsic parameters in one embodiment of this application;

[0033] Figure 3 A structural block diagram of a calibration device according to one embodiment of this application is shown;

[0034] Figure 4 A structural block diagram of an electronic device according to one embodiment of this application is shown;

[0035] Figure 5 A schematic diagram of the calibration field structure is shown in one embodiment of this application;

[0036] Figure 6 A schematic diagram of the retaining wall structure in one embodiment of this application is shown;

[0037] Figure 7 The third schematic diagram shows a method for calibrating vehicle-mounted lidar and RTK extrinsic parameters in one embodiment of this application. Detailed Implementation

[0038] To better understand the above-mentioned objectives, features, and advantages of this application, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0039] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of this application is not limited to the specific embodiments disclosed below.

[0040] The following reference Figures 1 to 7 This application describes a calibration method and related apparatus for vehicle-mounted lidar and RTK extrinsic parameters according to some embodiments.

[0041] like Figure 1 As shown, embodiments of this application provide a calibration method for vehicle-mounted LiDAR and RTK extrinsic parameters, used in vehicles equipped with LiDAR and a positioning and navigation system. The calibration method includes:

[0042] S101: Construct a calibration field and reference map. The calibration field includes retaining walls with sloping structures.

[0043] S103: When the vehicle to be calibrated is located in the calibration field, the raw point cloud data is collected by the lidar and the RTK pose data is obtained by the positioning and navigation system.

[0044] S105: Convert the raw point cloud data and RTK pose data to the same coordinate system as the reference map;

[0045] S107: Based on the original point cloud data and RTK pose data, the spatial transformation parameters of the lidar relative to the positioning and navigation system are solved.

[0046] The calibration method for vehicle-mounted LiDAR and RTK extrinsic parameters proposed in this application is used in vehicles equipped with LiDAR and a positioning and navigation system. First, a calibration field and a reference map are constructed. The key facility in the calibration field is a retaining wall with a sloping structure. This sloping structure provides the LiDAR with a continuously changing and distinctive three-dimensional surface, avoiding the feature degradation problem caused by flat walls. During the calibration process, the vehicle enters the calibration field, and the vehicle's LiDAR begins to collect raw point cloud data of the surrounding environment. Simultaneously, the vehicle's positioning and navigation system collects high-precision pose data of the vehicle itself. The positioning and navigation system includes an RTK-GNSS / INS integrated navigation system. Subsequently, all data is uniformly transformed to a common coordinate system, which is consistent with the coordinate system of the reference map pre-constructed using high-precision measurement methods. Finally, through algorithmic processing, the point cloud collected by the LiDAR is precisely registered with the reference map. By comparing the RTK pose data provided by the vehicle positioning and navigation system with the LiDAR pose data in the reference map coordinate system obtained after point cloud registration optimization, the spatial transformation parameters of the LiDAR relative to the positioning and navigation system can be deduced. This allows for the high-precision deduction of the spatial transformation parameters of the LiDAR relative to the positioning and navigation system, thus calculating the required extrinsic parameters for calibration. This application utilizes artificially designed retaining walls with stable geometric features, reducing dependence on the natural environment and improving the automation level and robustness of the calibration process. Simultaneously, it achieves high-precision, repeatable, automated calibration of vehicle-mounted LiDAR and RTK extrinsic parameters without relying on manual on-site intervention or complex natural scenes, making it particularly suitable for feature-sparse industrial environments such as mining areas and ports.

[0047] Optionally, in some embodiments, the steps of constructing the calibration field and the reference map include: constructing the calibration field based on the retaining wall; determining a geographic reference point within the calibration field and establishing a northeast-sky coordinate system with the geographic reference point as the origin; collecting point cloud data of the calibration field through a data acquisition carrier; and converting the point cloud data of the calibration field into the northeast-sky coordinate system to generate a reference map of the calibration field.

[0048] In this embodiment, a physical calibration field is first constructed using a retaining wall with a sloping structure. This artificial structure provides the lidar with a stable and distinctive 3D scanning target, effectively solving the problem of obtaining effective point cloud data in flat mining areas with sparse natural features. Subsequently, within the constructed calibration field, a geographic reference point is determined using high-precision measurement methods. For example, a post-processing GNSS technique based on long-term static observation is used to determine the geographic reference point, and this point is used as the origin to establish a northeast-sky coordinate system. The X-axis of the northeast-sky coordinate system points east, the Y-axis points north, and the Z-axis points vertically upward, serving as a unified spatial reference. Next, a specialized acquisition carrier scans the entire calibration field, acquiring the raw point cloud data. Finally, the acquired calibration field point cloud data is unified to the previously established northeast-sky coordinate system using coordinate transformation methods, thereby generating a high-precision reference map with absolute geographic coordinates. It is evident that this application, through artificially designed structured retaining walls, ensures the quality of lidar scanning data and the strength of geometric constraints, avoiding the risk of calibration failure in open areas. Meanwhile, by establishing a unified northeast-sky coordinate system and generating a high-precision reference map, a reliable and consistent truth benchmark is provided for the accurate fusion and automated calibration of subsequent multi-sensor data, which greatly improves the robustness of the calibration process and the repeatability of the results.

[0049] Optionally, in some embodiments, the step of collecting point cloud data of the calibration field by means of a data acquisition vehicle includes: collecting point cloud data of the calibration field by means of a professional mapping vehicle; and / or collecting point cloud data of the calibration field by means of UAV aerial survey; and / or collecting point cloud data of the calibration field by means of a handheld scanner; and / or collecting point cloud data of the calibration field by means of a fixed ground laser scanning method.

[0050] In this embodiment, point cloud data of the calibration field is collected using at least one of the following methods: a professional mapping vehicle, UAV aerial surveying, a handheld scanner, and a fixed ground laser scanner. The professional mapping vehicle is typically equipped with a high-line-count lidar and a high-precision GNSS / INS integrated navigation system. By traveling along a preset path within the calibration field, it performs a comprehensive and continuous scan of the retaining wall and its surrounding environment. The UAV aerial surveying system uses a UAV equipped with lidar or an oblique photography camera to fly over the calibration field, quickly acquiring large-area, high-density 3D point cloud data from an aerial perspective, particularly suitable for tall retaining walls or areas difficult for vehicles to access. The handheld or backpack mobile scanner, carried by an operator, is a lightweight device integrating lidar and an inertial measurement unit. It moves within the calibration field to collect detailed information about the retaining wall surface at close range and high resolution. The fixed ground laser scanner performs static scanning at different locations within the field and then precisely stitches together the point cloud data from multiple stations to obtain point cloud data with extremely high accuracy and extremely low noise. These data acquisition carriers all use pre-determined geographical reference points as absolute coordinate benchmarks to ensure that the acquired point clouds can be accurately converted to a unified northeast-northeast coordinate system. Therefore, this application provides diverse reference map construction methods adapted to different application scenarios and cost budgets. It can obtain high-precision maps using existing professional equipment and can also be quickly deployed through more flexible solutions, ensuring the universality and feasibility of this method under different mining area conditions, ultimately providing a reliable true benchmark for automatic external parameter calibration.

[0051] Optionally, in some embodiments, the step of inversely solving the spatial transformation parameters of the lidar relative to the positioning and navigation system based on the original point cloud data and RTK pose data includes: converting the RTK pose data into a homogeneous transformation matrix in the northeast-northeast coordinate system; setting initial extrinsic parameters, and projecting the original point cloud data onto the northeast-northeast coordinate system according to the initial extrinsic parameters to obtain initial projected point cloud data; spatially aligning the initial projected point cloud data with a reference map using a point cloud registration algorithm to obtain a fine registration transformation matrix; and inversely solving the spatial transformation parameters of the lidar relative to the positioning and navigation system based on the fine registration transformation matrix and the RTK pose data.

[0052] In this embodiment, the RTK pose data acquired by the positioning and navigation system, including latitude, longitude, altitude coordinates, and roll, pitch, and yaw angles, is first converted into a homogeneous transformation matrix in the northeast-northeast coordinate system. This homogeneous transformation matrix fully describes the vehicle's position and attitude in the reference map coordinate system. Next, an initial extrinsic parameter estimate can be set based on the installation relationship between the LiDAR and the positioning and navigation system. Using this initial extrinsic parameter estimate, the raw point cloud data collected by the LiDAR is converted to the northeast-northeast coordinate system to obtain initial projected point cloud data. Then, a high-precision point cloud registration algorithm is used to perform precise spatial alignment calculations between the initial projected point cloud data and a pre-constructed high-precision reference map, solving for a fine registration transformation matrix that best matches the initial projected point cloud onto the reference map. Finally, based on the fine registration transformation matrix and the known RTK pose data, the precise spatial transformation parameters of the LiDAR relative to the positioning and navigation system, i.e., the extrinsic parameters to be calibrated, are calculated through mathematical inverse kinematics. As can be seen, this application compares sensor observation data with a high-precision ground truth map through coordinate transformation and iterative optimization registration, thereby accurately separating and solving the relative pose relationship between sensors, realizing fully automatic and high-precision extrinsic parameter calibration, avoiding the dependence on manual measurement in traditional methods, and ensuring the reliability and repeatability of calibration results.

[0053] Optionally, in some embodiments, when the vehicle to be calibrated is located in the calibration field, the steps of collecting raw point cloud data based on the lidar and obtaining RTK pose data based on the positioning and navigation system include: when the vehicle to be calibrated is located in the calibration field, parking the vehicle to be calibrated sequentially at multiple pre-planned stationary points in the calibration field; at each stationary point, collecting raw point cloud data based on the lidar and obtaining RTK pose data based on the positioning and navigation system; wherein both the raw point cloud data and the RTK pose data carry a timestamp.

[0054] In this embodiment, after the vehicle to be calibrated (such as a mining dump truck) enters the pre-constructed calibration field, the operator moves the vehicle sequentially and precisely stops it at multiple pre-planned stationary points within the calibration field according to a preset plan. At each stationary point, the vehicle needs to come to a complete stop before data recording begins. The collected data includes two types: first, raw point cloud data obtained by the vehicle-mounted LiDAR over several seconds; and second, high-precision RTK pose data output by the vehicle's positioning and navigation system. The RTK pose data includes latitude, longitude, and altitude coordinates, as well as attitude information such as roll angle, pitch angle, and heading angle. Throughout the data acquisition process, the RTK pose data must remain in a fixed solution state. In this state, by calculating the correction signal from the ground base station, centimeter-level ultra-high precision positioning data can be obtained to ensure the high reliability of the pose data. Furthermore, both the raw point cloud data collected by the LiDAR and the RTK pose data acquired by the positioning and navigation system must carry precise timestamps, providing a necessary foundation for subsequent data synchronization and the establishment of frame correspondence. By employing a multi-pose static acquisition strategy, the observability of each degree of freedom of the extrinsic parameters can be effectively enhanced, avoiding instantaneous errors that may be introduced by vehicle movement during dynamic acquisition. Simultaneously, discrete static point data facilitates independent quality checks and filtering. If the RTK signal at a particular point is lost or the point cloud quality is poor, it can be removed individually without affecting the overall process, thereby improving the robustness of the calibration process and the accuracy of the final results. The static acquisition mode is particularly suitable for scenarios requiring high initial calibration accuracy or needing to verify dynamic calibration results.

[0055] Optionally, in some embodiments, when the vehicle to be calibrated is located in the calibration field, the steps of collecting raw point cloud data based on the lidar and obtaining RTK pose data based on the positioning and navigation system include: when the vehicle to be calibrated is located in the calibration field, causing the vehicle to be calibrated to travel along a preset trajectory; during the travel, the lidar and the positioning and navigation system continuously collect raw data; and filtering the raw point cloud data and RTK pose data from the continuously collected raw data; wherein both the raw point cloud data and the RTK pose data carry a timestamp.

[0056] In this embodiment, after the vehicle to be calibrated (such as a mining dump truck) enters the pre-constructed calibration field, the operator controls the vehicle to travel at a constant or variable speed along a preset trajectory, for example, circling around the calibration field's circular road or other paths that cover different viewpoints. Throughout the vehicle's journey, the onboard LiDAR and positioning and navigation system work synchronously, continuously collecting raw data. The LiDAR continuously scans the surrounding environment, generating a continuous point cloud sequence containing structures such as the calibration field retaining walls; the positioning and navigation system continuously outputs the vehicle's RTK pose data, which includes latitude, longitude, elevation coordinates, roll angle, pitch angle, and heading angle. Data frames that meet the requirements are then selected based on preset conditions. These preset conditions may require the LiDAR point cloud to clearly scan a sufficiently large area of ​​the retaining wall surface. Both the selected raw point cloud data and the corresponding RTK pose data carry precise timestamps to ensure synchronization and correspondence between the data. This application's dynamic acquisition mode can quickly acquire massive amounts of data from vehicles at different positions and attitudes, offering high efficiency and more closely reflecting the actual operating conditions of the vehicle. By filtering valid frames from continuous data, segments with instantaneous RTK signal loss or poor point cloud quality can be avoided, and point cloud data and RTK pose data for calibration can be selected. This improves acquisition efficiency while ensuring data quality, and is particularly suitable for scenarios that require rapid and batch calibration of vehicle fleets.

[0057] Optionally, static acquisition mode and dynamic acquisition mode are two complementary data acquisition methods provided by the present invention. Users can flexibly choose a single mode or a combination of modes to perform calibration according to actual conditions, efficiency requirements and accuracy targets.

[0058] Optionally, in some embodiments, the number of original point cloud data and RTK pose data are both multiple, and the calibration method further includes: calculating the spatial transformation parameters of the lidar relative to the positioning and navigation system at multiple independent observation times; solving for the optimal spatial transformation parameters by the least squares method, evaluating the quality of the calibration result of the optimal spatial transformation parameters, and outputting the extrinsic parameter matrix when the quality index meets the preset threshold.

[0059] In this embodiment, during data acquisition, raw point cloud data and corresponding RTK pose data at multiple independent observation times are acquired, such as at different stationary positions or different time points during dynamic movement of the mining truck within the calibration field. First, based on each frame of independent data—that is, the point cloud data and corresponding RTK pose data at a single moment—a spatial transformation parameter of the lidar relative to the positioning and navigation system is calculated according to the registration and inverse kinematics process. Then, the least squares method is used to jointly optimize all independently calculated spatial transformation parameters, aiming to find an optimal spatial transformation parameter that better fits all observation data, thereby effectively suppressing potential random errors in single-frame data and obtaining statistically more accurate and stable optimal extrinsic parameters. Finally, the calibration results of the optimal spatial transformation parameter are evaluated for quality, with quality indicators including the consistency of parameter estimates across multiple frames and the accuracy of point cloud registration. The consistency of multi-frame parameter estimates includes the standard deviations of translation and rotation components. Point cloud registration accuracy includes calculating the average residual between the finely registered point cloud and the reference map, i.e., the average distance from a point to the corresponding model surface. These are compared with preset thresholds based on quality indicators; for example, the standard deviation of translation is required to be less than 5 cm, and the standard deviation of rotation less than 0.5 degrees. Only when all quality indicators meet the preset threshold requirements is the calibration considered successful, and the optimized extrinsic parameter matrix is ​​finally output. If any indicator fails to meet the requirements, a calibration failure is indicated, and data checking or re-acquisition is recommended. This application significantly improves the accuracy and robustness of extrinsic parameter calibration results through joint optimization of multi-frame data and a rigorous quality control process. It effectively avoids calibration errors caused by poor single-frame data quality, such as those caused by instantaneous RTK signal fluctuations or partial point cloud occlusion. It effectively prevents the application of defective or inaccurate calibration results to actual vehicles, thereby ensuring the positioning accuracy and driving safety of the multi-sensor fusion system. It is particularly suitable for batch calibration scenarios of unmanned mining trucks with extremely high reliability requirements.

[0060] like Figure 2 As shown, embodiments of this application provide a detailed method for calibrating vehicle-mounted LiDAR and RTK extrinsic parameters, including:

[0061] S201: Construct a calibration field based on the retaining wall, determine a geographical reference point within the calibration field, and establish a northeast-sky coordinate system with the geographical reference point as the origin;

[0062] S203: Collect point cloud data of the calibration field through professional mapping vehicles, UAV aerial surveying, handheld scanners and / or fixed ground laser scanning;

[0063] S205: Convert the point cloud data of the calibration field to the northeast-sky coordinate system to generate a reference map of the calibration field;

[0064] S207: When the vehicle to be calibrated is located in the calibration field, raw point cloud data is collected based on the lidar, and RTK pose data is obtained based on the positioning and navigation system;

[0065] S209: Convert the raw point cloud data and RTK pose data to the same coordinate system as the reference map;

[0066] S211: Convert RTK pose data into a homogeneous transformation matrix in the northeast-north sky coordinate system;

[0067] S213: Set initial extrinsic parameters, and project the original point cloud data onto the northeast-sky coordinate system according to the initial extrinsic parameters to obtain the initial projected point cloud data;

[0068] S215: Spatially align the initial projected point cloud data with the reference map using a point cloud registration algorithm to obtain a fine registration transformation matrix;

[0069] S217: Based on the fine registration transformation matrix and RTK pose data, the spatial transformation parameters of the lidar relative to the positioning and navigation system are solved inversely;

[0070] S219: Calculate the spatial transformation parameters of the lidar relative to the positioning and navigation system at multiple independent observation times;

[0071] S221: Solve for the optimal spatial transformation parameters using the least squares method, evaluate the quality of the calibration results of the optimal spatial transformation parameters, and output the extrinsic parameter matrix if the quality index meets the preset threshold.

[0072] like Figure 3 As shown, according to a second aspect of this application, a calibration device 100 for vehicle-mounted LiDAR and RTK extrinsic parameters is also proposed for use in a vehicle equipped with a LiDAR and a positioning and navigation system. The calibration device 100 includes: a construction unit 110 for constructing a calibration field and a reference map, the calibration field including a retaining wall with a sloping structure; an acquisition unit 120 for acquiring raw point cloud data from the LiDAR and obtaining RTK pose data from the positioning and navigation system when the vehicle to be calibrated is located in the calibration field; a transformation unit 130 for uniformly transforming the raw point cloud data and RTK pose data to the same coordinate system as the reference map; and a calculation unit 140 for inversely solving the spatial transformation parameters of the LiDAR relative to the positioning and navigation system based on the raw point cloud data and RTK pose data.

[0073] The calibration device 100 for vehicle-mounted LiDAR and RTK extrinsic parameters proposed in this application is used in a vehicle equipped with a LiDAR and a positioning and navigation system. First, a calibration field and reference map are constructed by a construction unit 110. The key facility of the calibration field is a retaining wall with a sloping structure. The sloping structure provides a continuously changing and distinctive three-dimensional curved surface for LiDAR scanning, avoiding the feature degradation problem caused by flat walls. During the calibration process, when the vehicle to be calibrated enters the calibration field, the acquisition unit 120 begins to collect raw point cloud data of the surrounding environment through the LiDAR on the vehicle. Simultaneously, the acquisition unit 120 collects high-precision pose data of the vehicle itself through the vehicle's positioning and navigation system. The positioning and navigation system includes an RTK-GNSS / INS integrated navigation system. Subsequently, the conversion unit 130 converts all data to a common coordinate system, which is consistent with the coordinate system of the reference map pre-constructed using high-precision measurement methods. Finally, the point cloud collected by the LiDAR is precisely registered with the reference map. The computing unit 140 compares the RTK pose data provided by the vehicle positioning and navigation system with the LiDAR pose data in the reference map coordinate system obtained after point cloud registration optimization to inversely solve the spatial transformation parameters of the LiDAR relative to the positioning and navigation system. This allows for the high-precision inverse solution of the spatial transformation parameters of the LiDAR relative to the positioning and navigation system, thus calculating the required extrinsic parameters for calibration. This application utilizes artificially designed retaining walls with stable geometric features, reducing dependence on the natural environment and improving the automation level and robustness of the calibration process. Simultaneously, it achieves high-precision, repeatable, automated calibration of vehicle-mounted LiDAR and RTK extrinsic parameters without relying on manual on-site intervention or complex natural scenes, making it particularly suitable for industrial environments with sparse features, such as mining areas and ports.

[0074] According to a third aspect of this application, a readable storage medium is also proposed, on which a program or instructions are stored, which, when executed by a processor, implement the steps of the calibration method as proposed in the first aspect. Therefore, it possesses all the beneficial effects of the calibration method for vehicle-mounted LiDAR and RTK extrinsic parameters.

[0075] like Figure 4 As shown, according to a fourth aspect of this application, an electronic device 200 is also proposed, including a memory 210, a processor 220, and a computer program stored in the memory 210 and executable on the processor 220. When the processor 220 executes the computer program, it implements the steps of the calibration method as proposed in the first aspect. Therefore, it possesses all the beneficial effects of the calibration method for vehicle-mounted lidar and RTK extrinsic parameters.

[0076] In a specific application, the specific implementation of this invention includes four core steps: calibration field construction, reference map generation, multimodal data acquisition, and automatic extrinsic parameter solution. These steps work collaboratively to form a closed-loop, repeatable, and highly robust LiDAR-RTK extrinsic parameter calibration process. These steps are detailed below:

[0077] (1) Calibration Field Construction: First, a dedicated calibration field is planned and constructed in a flat area of ​​the mining area. The calibration field can adopt a rectangular, circular, or other layout with obvious geometric features, and retaining walls are generally constructed using local materials. The retaining walls need to have a certain height to fully cover the scanning field of the vehicle-mounted lidar. The retaining walls can be designed as special structures, such as slopes, so that the lidar can form continuous, clear, and distinctive three-dimensional geometric information during scanning, thereby effectively improving the stability and uniqueness of point cloud registration.

[0078] like Figure 5 As shown, the calibration field 1 is enclosed by a retaining wall 11, and the calibration field 1 includes a vehicle parking area 13 for vehicle calibration. The cross-section of the retaining wall 11 is shown below. Figure 6 As shown, the retaining wall 11 with a sloping structure enables the lidar to form continuous, clear, and distinctive three-dimensional geometric information during scanning, thereby effectively improving the stability and uniqueness of point cloud registration.

[0079] A geographic reference point, M_0 = {lat, lon, h_0}, is selected within the calibration field, where M_0 represents a dot or reference point, lat represents latitude, lon represents longitude, and h_0 represents altitude. The geographic reference point is obtained through high-precision static global navigation satellite system measurements, for example, using a 24-hour static observation post-processing method to ensure its accurate coordinates. This point will serve as the origin for all subsequent Northeast-Sky coordinate systems. The X-axis of the Northeast-Sky coordinate system points east, the Y-axis points north, and the Z-axis points vertically upward. All subsequently generated maps and positioning attitude data are uniformly converted to the Northeast-Sky coordinate system to ensure data consistency in time and space. Other coordinate systems besides the Northeast-Sky coordinate system can also be used without affecting the final result.

[0080] (2) Construction of a high-precision reference map: This invention does not limit the construction method of the calibration field reference map, but the key point is that the final map must meet the requirements of geometric integrity, coordinate uniformity, and absolute accuracy. Therefore, one or more of the following methods can be flexibly selected for data collection based on the site conditions, equipment availability, and cost budget:

[0081] Professional mapping vehicle mode: This mode uses a vehicle equipped with a global satellite navigation system / inertial navigation system and multi-line lidar to scan along a predetermined route. It is suitable for mature mining areas with good road conditions and open spaces.

[0082] Unmanned Aerial Vehicle (UAV) Aerial Survey Mode: This mode utilizes a real-time dynamic / post-differential positioning UAV equipped with a lidar or high-resolution camera to conduct low-altitude aerial surveys of the entire calibration field at an altitude of 10 to 30 meters. Point clouds or 3D models with geographic coordinates are generated through airborne positioning systems or ground control point constraints. This method is particularly suitable for scenarios with high retaining walls where vehicles cannot easily approach for scanning, and it enables rapid full-field coverage.

[0083] Handheld / backpack-style mobile scanning mode: Operators carry backpack devices or handheld scanners with integrated positioning and attitude determination systems and LiDAR, walking around a retaining wall to acquire high-density point clouds at close range. In areas with limited satellite signals, visual-inertial assistance or loop traversal optimization can be used to ensure trajectory accuracy. This method is flexible and lightweight, suitable for rapid deployment in confined spaces or temporary calibration fields.

[0084] Fixed ground laser scanning mode: Two to four scanning stations are set up around the calibration field, and a high-precision static laser scanner is used to perform a full-range scan of the retaining wall. The points from each station are stitched together using target spheres or control points, and the overall point cloud is converted to the northeast-central coordinate system using the control points with known coordinates. This method yields a point cloud with extremely high density and low noise, making it suitable as a "gold standard" reference map.

[0085] In addition to the methods mentioned above, users can also customize the map construction method according to their actual situation. Regardless of the acquisition method used, the following common requirements must be met: Geographic coordinate alignment: All point clouds must be reliably converted to the Northeast Celestial Coordinate System with M_0 = {lat, lon, h_0} as the origin; Structural integrity: It must completely cover the slopes on both sides of the retaining wall and the flat areas inside the site, without large areas of obstruction or missing parts.

[0086] The final generated reference map M serves as the unique truth benchmark, which is shared by all subsequent mining card calibration tasks.

[0087] (3) Data collection for mining trucks to be calibrated: For each mining dump truck to be calibrated, after the installation of the target lidar and real-time dynamic positioning equipment is completed, one of the following two types of data collection must be performed:

[0088] Method 1: Static Data Acquisition. Park the mining truck sequentially at one or more pre-planned static locations. At each location, ensure the vehicle is completely stationary (handbrake engaged, wheel chocks engaged), activate the data recording system, and continuously acquire data for several seconds. The system simultaneously records the following information: raw lidar point cloud data, latitude, longitude, and altitude coordinates output by the real-time dynamic positioning system, along with 6 degrees of freedom attitude angles (roll, pitch, and yaw) and a precise timestamp (based on the GPS week second or the Precision Time Protocol master clock). The yaw angle is defined with true north as 0° and clockwise rotation as the positive direction. The real-time dynamic positioning system must maintain a "fixed solution" state throughout the entire acquisition process.

[0089] The second method is dynamic data acquisition. The mining truck moves continuously along a planned path within the calibration field (e.g., circling a retaining wall) or other preset trajectories, simultaneously acquiring and saving continuous data. From the continuous motion data, data segments located at different positions and times are selected. The selected data segments should meet the following conditions: the lidar point cloud can scan as much of the retaining wall area as possible, thus providing good geometric constraints for subsequent calibration. The data recording requirements of the real-time dynamic positioning system are consistent with those of the static acquisition method.

[0090] (4) Automatic solution algorithm for external parameters: After data acquisition is completed, the offline calibration stage is entered.

[0091] Step 1: RTK pose to N-H sky coordinate transformation matrix. For each frame of RTK data, first use an ellipsoidal model to convert latitude, longitude, and altitude {lat, Lon, h} into a N-H sky coordinate translation vector relative to M_0. , , ).in, This represents the eastward offset from the origin M_0. This represents the offset northward from the origin M_0. This represents the offset of the origin M_0 vertically upwards from the horizontal ground. Simultaneously, the attitude angles (including roll, pitch, and yaw angles) are converted into rotation matrices in the ZYX Euler angle order. Therefore, the homogeneous transformation matrix is ​​constructed as follows:

[0092] ;

[0093] Step 2: Initial projection of the lidar point cloud. Assume initial extrinsic parameters. Based on the structural design drawings, reasonable initial values ​​were set, and the initial pose of the lidar in the northeast-northeast coordinate system was calculated. :

[0094] ;

[0095] in, Representing the true pose of the RTK in the northeast-northeast coordinate system, the lidar point cloud L at this moment is projected onto the northeast-northeast coordinate system through matrix transformation to obtain L. enu，init .

[0096] Step 3: Point Cloud Fine Registration. Using a point cloud registration algorithm, L... enu，init The map is registered with the reference map M. The registration result is a fine registration transformation matrix. , making Align with M.

[0097] Step 4: Inverse kinematics to determine the true pose and extrinsic parameters of the lidar. Calculate the true pose of the lidar in the northeast-northeast coordinate system:

[0098] ;

[0099] in, This represents the true pose of the lidar in the northeast-northeast coordinate system. Represents the fine registration transformation matrix. This indicates the initial pose of the lidar in the northeast-northeast coordinate system.

[0100] Furthermore, the extrinsic parameters of the lidar relative to the RTK at this moment can be obtained by inverse solving:

[0101] ;

[0102] in, This represents the extrinsic parameters of the lidar relative to the RTK. This represents the true pose of the RTK in the northeast-sky coordinate system. This indicates the true pose of the lidar in the northeast-northeast coordinate system.

[0103] Step 5: Multi-frame joint optimization. For all stationary points and multiple keyframes in the dynamic trajectory, calculate... Then construct the least squares problem:

[0104] ;

[0105] in, The square root of the sum of the squares of all elements of a matrix. This represents the true pose of the lidar in the northeast-northeast coordinate system at the i-th time or i-th position. This represents the true pose of the RTK in the northeast-northeast coordinate system at the i-th time or i-th position. This represents the extrinsic parameters of the lidar relative to the RTK. The optimal extrinsic parameters are solved using a nonlinear optimization library, and the final result, including the rotation matrix and translation vector, is output.

[0106] Step Six: Quality Assessment and Output. The system automatically calculates the registration residuals (such as the average point-to-surface distance) and extrinsic parameter consistency (the standard deviation of the solution results for each frame). If the preset thresholds are met, such as the translation standard deviation being less than 5 cm and the rotation standard deviation being less than 0.5 degrees, the calibration is considered successful, and the extrinsic parameters are saved to the configuration file; otherwise, the user is prompted to check the data quality or re-acquire the data.

[0107] like Figure 7 As shown, this application provides a schematic diagram of a calibration process, comprising two parts: reference map construction and extrinsic parameter calibration. The reference map construction first involves data acquisition through various methods, including vehicle-mounted data acquisition, UAV aerial surveying, handheld / backpack scanning, fixed ground scanning, and other acquisition methods. The acquired data is then processed, involving coordinate transformation, multi-source data fusion, data denoising and filtering, and other related processing.

[0108] After map construction is completed, extrinsic parameter calibration is performed. First, vehicle sensor data is collected, including vehicle parking and data acquisition. Then, the collected data is preprocessed, including time synchronization verification and coordinate transformation. Next, the point cloud is initially projected, including initial extrinsic parameter assumptions and projecting the point cloud onto the northeast-northeast coordinate system. Finally, multi-frame joint optimization is performed, using a joint nonlinear optimization method to further optimize the data and project it onto the northeast-northeast coordinate system, thus completing the calibration process.

[0109] In this application, the term "multiple" refers to two or more unless otherwise expressly defined. The terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; "linking" can be a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0110] In the description of this specification, the terms "one embodiment," "some embodiments," "specific embodiment," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0111] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for calibrating vehicle-mounted lidar and RTK extrinsic parameters, characterized in that, For use in a vehicle equipped with a lidar and a positioning and navigation system, the calibration method includes: Construct a calibration field and a reference map, wherein the calibration field includes a retaining wall with a sloping structure; When the vehicle to be calibrated is located in the calibration field, raw point cloud data is collected based on the lidar, and RTK pose data is obtained based on the positioning and navigation system. The original point cloud data and the RTK pose data are uniformly converted to the same coordinate system as the reference map; Based on the original point cloud data and the RTK pose data, the spatial transformation parameters of the lidar relative to the positioning and navigation system are deduced.

2. The calibration method for vehicle-mounted lidar and RTK extrinsic parameters according to claim 1, characterized in that, The steps of constructing the calibration field and reference map include: The calibration field is constructed based on the retaining wall; A geographical reference point is determined within the calibration field, and a northeast-sky coordinate system is established using the geographical reference point as the origin. Point cloud data of the calibration field are collected using a data acquisition carrier; The point cloud data of the calibration field is converted into the northeast-sky coordinate system to generate a reference map of the calibration field.

3. The calibration method for vehicle-mounted lidar and RTK extrinsic parameters according to claim 2, characterized in that, The step of acquiring point cloud data of the calibration field using a data acquisition carrier includes: Point cloud data of the calibration field is collected using a specialized mapping vehicle; and / or Point cloud data of the calibration field were collected via UAV aerial survey; and / or Point cloud data of the calibration field are acquired using a handheld scanner; and / or Point cloud data of the calibration field are collected using a fixed ground-based laser scanning method.

4. The calibration method for vehicle-mounted lidar and RTK extrinsic parameters according to claim 1, characterized in that, The step of reversing the spatial transformation parameters of the lidar relative to the positioning and navigation system based on the original point cloud data and the RTK pose data includes: The RTK pose data is converted into a homogeneous transformation matrix in the northeast-north sky coordinate system. Set initial extrinsic parameters, and project the original point cloud data onto the northeast-sky coordinate system according to the initial extrinsic parameters to obtain initial projected point cloud data; The initial projected point cloud data is spatially aligned with the reference map using a point cloud registration algorithm to obtain a fine registration transformation matrix; Based on the fine registration transformation matrix and the RTK pose data, the spatial transformation parameters of the lidar relative to the positioning and navigation system are solved.

5. The calibration method for vehicle-mounted lidar and RTK extrinsic parameters according to claim 1, characterized in that, The steps of acquiring raw point cloud data based on the lidar and obtaining RTK pose data based on the positioning and navigation system when the vehicle to be calibrated is located in the calibration field include: When the vehicle to be calibrated is located in the calibration field, the vehicle to be calibrated is parked in sequence at multiple pre-planned stationary points in the calibration field; At each stationary point, the raw point cloud data is collected by the lidar, and the RTK pose data is obtained by the positioning and navigation system. Both the original point cloud data and the RTK pose data carry timestamps.

6. The calibration method for vehicle-mounted lidar and RTK extrinsic parameters according to claim 1, characterized in that, The steps of acquiring raw point cloud data based on the lidar and obtaining RTK pose data based on the positioning and navigation system when the vehicle to be calibrated is located in the calibration field include: When the vehicle to be calibrated is located in the calibration field, the vehicle to be calibrated is made to travel along a preset trajectory; During the driving process, the lidar and the positioning and navigation system continuously collect raw data; The raw point cloud data and the RTK pose data are filtered out from the continuously collected raw data; Both the original point cloud data and the RTK pose data carry timestamps.

7. The calibration method for vehicle-mounted lidar and RTK extrinsic parameters according to any one of claims 1 to 6, characterized in that, The number of original point cloud data and the number of RTK pose data are both multiple, and the calibration method further includes: Calculate the spatial transformation parameters of the lidar relative to the positioning and navigation system at multiple independent observation times; The optimal spatial transformation parameters are solved by the least squares method, and the quality of the calibration results of the optimal spatial transformation parameters is evaluated. If the quality index meets the preset threshold, the extrinsic parameter matrix is ​​output.

8. A calibration device for vehicle-mounted lidar and RTK extrinsic parameters, characterized in that, For use in a vehicle equipped with a lidar and positioning navigation system, the calibration device includes: A construction unit is used to construct a calibration field and a reference map, wherein the calibration field includes a retaining wall with a sloping structure; The acquisition unit is used to acquire raw point cloud data based on the lidar and obtain RTK pose data based on the positioning and navigation system when the vehicle to be calibrated is located in the calibration field. A conversion unit is used to uniformly convert the original point cloud data and the RTK pose data to the same coordinate system as the reference map; The computing unit is used to inversely solve the spatial transformation parameters of the lidar relative to the positioning and navigation system based on the original point cloud data and the RTK pose data.

9. A readable storage medium, characterized in that, It stores a program or instructions that, when executed by a processor, implement the steps of the calibration method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the calibration method as described in any one of claims 1 to 7.

Images