Pose estimation method based on laser radar and related equipment
By employing screening and normal distribution transformation verification methods, combined with inertial measurement unit data, the non-fully automatic and noise interference problems of lidar pose estimation are solved, achieving automated and accurate pose estimation, which is applicable to fields such as autonomous driving and mobile robots.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-26
- Publication Date
- 2026-03-27
Smart Images
Figure CN121741691A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of positioning technology, and in particular to a pose estimation method and related equipment based on lidar. Background Technology
[0002] Currently, LiDAR-based map matching and localization technology is widely used in autonomous driving, mobile robots, warehousing and logistics, and other fields. Mainstream solutions typically employ point cloud registration algorithms such as NDT (Normal Distributions Transform) or GICP (Generalized Iterative Closest Point) for pose estimation. However, these methods have the following significant drawbacks: they are not fully automatic, requiring manual specification of the initial position, making them unsuitable for autonomous in unattended scenarios; they suffer from orientation ambiguity, resulting in a lack of precise orientation in the output, which can easily lead to registration failures; they fail to integrate motion information to construct a high-quality query scenario, and single-frame point clouds are susceptible to noise interference; ultimately, this can lead to significant errors in the pose estimation results. Summary of the Invention
[0003] This application provides a pose estimation method and related equipment based on lidar, which can solve the problem of large errors in pose estimation results in related technologies.
[0004] To solve the above-mentioned technical problems, this application is implemented as follows: In a first aspect, embodiments of this application provide a pose estimation method based on lidar, comprising: Acquire the first frame of point cloud data collected by the lidar; The first frame of point cloud data is matched with the data information of the map to determine K candidate locations in the map that correspond to the point cloud data, where K is a positive integer; Extract local information corresponding to the K candidate locations from the map; The first frame point cloud data and the point cloud data corresponding to the local information are subjected to normal distribution transformation for verification to obtain the verification result. The verification result is used to indicate the candidate positions that have passed the verification among the K candidate positions, and the verification score of the candidate positions that have passed the verification. Based on the first candidate position, a normal distribution transformation is verified for tracking for N consecutive frames. If the first candidate position passes the verification in all N frames, the pose corresponding to the first candidate position is set as the initial pose. If the first candidate position fails the verification in any of the N frames, a second candidate position is selected for verification for tracking for N consecutive frames until the initial pose is set. The first candidate position is the candidate position with the highest verification score among the K candidate positions.
[0005] Optionally, the step of performing a normal distribution transformation verification on the first frame point cloud data and the point cloud data corresponding to the local information includes: Based on the first frame point cloud data, perform multi-angle transformation to obtain point cloud data corresponding to the first frame point cloud data at multiple angles; The point cloud data of the first frame at multiple angles and the point cloud data corresponding to the local information of the target candidate position are verified by performing a multi-angle normal distribution transformation. The point cloud data corresponding to the local information of the target candidate position includes point cloud data at multiple angles. The target candidate position is any one of the K candidate positions.
[0006] Optionally, the first frame of point cloud data acquired by the lidar includes any of the following: A single frame of point cloud data collected by a lidar within a single acquisition cycle; The point cloud data set is formed by superimposing and fusing multiple frames of point cloud data continuously collected by lidar in multiple acquisition cycles after being unified to the same coordinate system through pose transformation. The structured sub-graph data is generated by registering and optimizing multiple frames of point cloud data continuously acquired by the lidar in multiple acquisition cycles. The structured sub-graph data includes: two-dimensional probabilistic grids, three-dimensional point cloud sub-graphs, or truncated symbolic distance function voxels.
[0007] Optionally, acquiring the first frame of point cloud data collected by the lidar includes: Acquire the first frame of point cloud data collected by the lidar; The motion data collected by the inertial measurement unit and the relative pose of the lidar and the coordinate system of the inertial measurement unit are acquired. The motion data includes angular velocity data and acceleration data. The pose increment of the inertial measurement unit is calculated based on the motion data; Based on the first frame point cloud data, the relative pose of the coordinate systems of the lidar and the inertial measurement unit, and the pose increment, the first frame point cloud data is transformed to calculate the distortion-corrected first frame point cloud data.
[0008] Secondly, the present invention provides a pose estimation device based on lidar, comprising: The first acquisition module is used to acquire the first frame of point cloud data collected by the lidar. The filtering module is used to perform similarity matching between the first frame point cloud data and the map data information to determine K candidate locations in the map that correspond to the point cloud data, where K is a positive integer. The second acquisition module is used to extract local information corresponding to the K candidate locations from the map; The calculation module is used to perform normal distribution transformation verification on the first frame point cloud data and the point cloud data corresponding to the local information to obtain the verification result. The verification result is used to indicate the candidate positions that have passed the verification among the K candidate positions, and the verification score of the candidate positions that have passed the verification. The verification module is used to perform continuous N-frame tracking normal distribution transformation verification based on the first candidate position. If the first candidate position passes the verification in all N frames, the pose corresponding to the first candidate position is set as the initial pose. If the first candidate position fails the verification in any of the N frames, a second candidate position is selected to perform continuous N-frame tracking normal distribution transformation verification until the initial pose is set. The first candidate position is the candidate position with the highest verification score among the K candidate positions.
[0009] Thirdly, the present invention provides an electronic device comprising: 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 lidar-based pose estimation method as described in any one of the first aspects.
[0010] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the lidar-based pose estimation method as described in any one of the first aspects.
[0011] Fifthly, the present invention provides a computer program product including computer instructions that, when executed by a processor, implement the steps in the lidar-based pose estimation method as described in any one of the first aspects.
[0012] In this invention, a first frame of point cloud data collected by a lidar is acquired; the first frame of point cloud data is matched with map data to determine K candidate locations on the map corresponding to the point cloud data, where K is a positive integer; local information corresponding to the K candidate locations is extracted from the map; the first frame of point cloud data and the point cloud data corresponding to the local information are subjected to normal distribution transformation verification to obtain verification results, which indicate the candidate locations that have passed verification among the K candidate locations and their verification scores; finally, based on the first candidate location, N consecutive frames of tracking normal distribution transformation verification are performed. If the first candidate location passes verification in all N frames, the pose corresponding to the first candidate location is set as the initial pose; if the first candidate location fails verification in any of the N frames, a second candidate location is selected for N consecutive frames of tracking normal distribution transformation verification until the initial pose is set; wherein, the first candidate location is the candidate location with the highest verification score among the K candidate locations. Thus, by filtering map locations and performing N consecutive frames of tracking effectiveness verification on candidate locations, the pose estimation results can be made more accurate. Attached Figure Description
[0013] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0014] Figure 1 A schematic flowchart illustrating the pose estimation method based on lidar provided in this application embodiment; Figure 2 A schematic flowchart illustrating another lidar-based pose estimation method provided in this application embodiment; Figure 3 A schematic diagram of the structure of a lidar-based pose estimation device provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0016] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and are not used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same class, and the number of objects is not limited; for example, the first object can be one or more.
[0017] In this embodiment of the invention, a pose estimation method and related equipment based on lidar are proposed to solve the problems of non-fully automatic, azimuth ambiguity and large noise interference in related technologies.
[0018] refer to Figure 1 , Figure 1 This is a flowchart illustrating a pose estimation method provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes the following steps: Step 101: Acquire the first frame of point cloud data collected by the lidar; In this embodiment of the invention, there is no need to rely on manual specification of the initial pose. The system can automatically enter the initialization process at startup, collect point cloud data for pose estimation, which greatly improves the user experience and the automation level of the system, and realizes the true "positioning upon startup" function.
[0019] Specifically, refer to Figure 2 When the system starts up, it first determines whether the initial pose is known. If the initial pose is known, it directly enters the tracking mode. If the initial pose is unknown, it automatically enters the initialization process provided in the embodiment of the present invention. This reduces the dependence on professional operators, lowers the threshold for system use and maintenance costs, and is suitable for large-scale deployment and long-term operation.
[0020] Optionally, the first frame of point cloud data acquired by the lidar includes any of the following: A single frame of point cloud data collected by a lidar within a single acquisition cycle; The point cloud data set is formed by superimposing and fusing multiple frames of point cloud data collected by lidar in multiple consecutive acquisition cycles after being unified to the same coordinate system through pose transformation. The structured subgraph data generated by the lidar after acquisition and registration in multiple consecutive acquisition cycles includes: two-dimensional probabilistic grids, three-dimensional point cloud subgraphs, or truncated symbolic distance function voxels.
[0021] In this embodiment of the invention, not only single-frame point cloud input is supported, but also multi-frame point cloud input is supported. Based on the multi-frame point cloud, subsequent global Scan Context retrieval can be performed, which can increase the quality of the query point cloud, reduce noise interference, and improve retrieval efficiency and the accuracy, reliability and robustness of the retrieval results.
[0022] Furthermore, the embodiments of the present invention are also compatible with submap input methods, which makes the present invention applicable to different SLAM (Simultaneous Localization and Mapping) front-end architectures and has broad application prospects.
[0023] Optionally, acquiring the first frame of point cloud data collected by the lidar includes: Acquire the first frame of point cloud data collected by the lidar; The motion data collected by the inertial measurement unit and the relative pose of the lidar and the coordinate system of the inertial measurement unit are acquired. The motion data includes angular velocity data and acceleration data. The pose increment of the inertial measurement unit is calculated based on the motion data; Based on the first frame point cloud data, the relative pose of the coordinate systems of the lidar and the inertial measurement unit, and the pose increment, the first frame point cloud data is transformed to calculate the distortion-corrected first frame point cloud data.
[0024] Since lidar cannot instantly collect all points in a frame, but rather collects them point by point or line by line over time, the collected point cloud data may have motion distortion. As an optional implementation method, the attitude and trajectory of the carrier can be calculated using high-frequency motion data collected by an IMU (Inertial Measurement Unit). Therefore, by combining the data from the inertial measurement unit, the point clouds collected by lidar at different times within a frame can be uniformly aligned to the same reference time, such as the middle time or the end of a frame, to eliminate the distortion and stretching of the point cloud caused by the movement and / or rotation of the carrier, and to ensure that the point cloud is consistent with the real scene. The high-frequency motion data is the motion data output at a sampling frequency much higher than that of lidar.
[0025] Step 102: Perform similarity matching between the first frame point cloud data and the map data information to determine K candidate locations in the map that correspond to the point cloud data, where K is a positive integer; In this embodiment of the invention, the system uses a global retrieval algorithm to select K possible locations from the entire pre-built map that are most similar to the first frame point cloud data, so as to perform similarity comparison calculations in the subsequent process. Instead of directly comparing the first frame point cloud data with all map locations, this method can reduce the amount of computation while ensuring the accuracy of pose estimation, thus effectively improving computational efficiency. Selecting K possible locations instead of just the most similar one also provides a high-quality, limited set of candidate locations for the subsequent verification stage, reducing the risk of errors and minimizing the impact of similar structures and noise.
[0026] Specifically, refer to Figure 2 The point cloud data is used to perform a global Scan Context search for Top-K candidate positions. Furthermore, the data can be sorted by similarity score to facilitate rapid comparison later.
[0027] Step 103: Extract local information corresponding to the K candidate locations from the map; Step 104: Perform normal distribution transformation verification on the first frame point cloud data and the point cloud data corresponding to the local information to obtain the verification result. The verification result is used to indicate the candidate positions that have passed the verification among the K candidate positions, and the verification score of the candidate positions that have passed the verification. In this embodiment of the invention, the point cloud data corresponding to the local information is modeled as a multidimensional normal probability field. The target position is determined by finding the transformation with the highest probability density in the first frame of point cloud data in this probability field. The candidate position with the highest score is selected from the candidate positions with converged verification results. If a converged verification result cannot be obtained, the candidate position is discarded. If all candidate positions do not meet the conditions, the initialization is restarted from acquiring point cloud data. The convergence of the verification result means that a stable solution that no longer changes significantly can be obtained by iteratively optimizing the pose transformation parameters.
[0028] Specifically, if there are frames that fail verification, a second frame of point cloud data is re-acquired, and the second frame of point cloud data is matched with the data information of the map to determine L candidate locations in the map that correspond to the point cloud data; then, local information corresponding to the L candidate locations is extracted from the map; the second frame of point cloud data and the point cloud data corresponding to the local information are subjected to normal distribution transformation for verification to obtain a high-scoring verification result.
[0029] See Figure 2After performing normal distribution transformation registration verification, the pose corresponding to the converged result with the highest fitness score (registration fit score) that is greater than the preset threshold is selected as the initial pose; if there is no high-scoring and converged registration result, the process automatically returns to the initial stage, re-acquires point cloud for pose estimation, and re-selects the corresponding new candidate position until a high-scoring and converged registration result appears, and then proceeds to the next step.
[0030] Optionally, the step of performing a normal distribution transformation verification on the first frame point cloud data and the point cloud data corresponding to the local information includes: Based on the first frame point cloud data, perform multi-angle transformation to obtain point cloud data corresponding to the first frame point cloud data at multiple angles; The point cloud data of the first frame at multiple angles and the point cloud data corresponding to the local information of the target candidate position are verified by performing a multi-angle normal distribution transformation. The point cloud data corresponding to the local information of the target candidate position includes point cloud data at multiple angles. The target candidate position is any one of the K candidate positions.
[0031] As an optional embodiment, by rotating the first frame of point cloud data to different angles, it can be optimally matched with the point cloud data of the candidate target location in the map. Since the map used is pre-built, the map contains information about each location rotating 360° along the horizontal, vertical, and longitudinal axes, respectively or together. Specifically, it contains field-of-view data of different roll angles, pitch angles, and yaw angles of each location. That is, the different attitude data corresponding to each location in the map are known. Therefore, the attitude of the lidar can be deduced from the matrix of the point cloud data of the pose corresponding to the candidate target location and the first frame of point cloud data during the rotation transformation process, so as to accurately correct the orientation deviation and improve the accuracy of the initial pose estimation.
[0032] Specifically, refer to Figure 2 For each candidate location of Scan Context, the point cloud data is rotated in 30° increments within the range of [-90°, +90°] for NDT registration to correct orientation bias and obtain accurate orientation.
[0033] Step 105: Perform continuous N-frame tracking normal distribution transformation verification based on the first candidate position. If the first candidate position passes verification in all N frames, the pose corresponding to the first candidate position is set as the initial pose. If the first candidate position fails verification in any of the N frames, a second candidate position is selected for continuous N-frame tracking normal distribution transformation verification until the initial pose is set. Wherein, the first candidate position is the candidate position with the highest verification score among the K candidate positions.
[0034] In this embodiment of the invention, after setting the initial pose, the tracking validity is verified through N consecutive frames. Only after successful matching within N consecutive frames is the initialization confirmed as valid and the pose estimation initialization completed. Otherwise, if there are frames that fail verification, point cloud data is re-acquired, pose estimation is performed again, and corresponding new candidate positions are selected until a valid initial pose is found. This avoids misjudgment due to accidental factors, enhances the robustness and reliability of the system, and enables the system to have self-recovery capabilities. The system can detect failures and automatically retry, ensuring stable operation even in complex or changing environments and preventing service interruption due to a single failure.
[0035] Specifically, if there are frames that fail verification, a second frame of point cloud data is re-acquired, and the second frame of point cloud data is matched with the data information of the map to determine J candidate positions in the map corresponding to the point cloud data; then, local information corresponding to the J candidate positions is extracted from the map; the second frame of point cloud data and the point cloud data corresponding to the local information are subjected to normal distribution transformation verification to obtain the verification result; based on the second candidate position, the tracking normal distribution transformation verification is performed again for N consecutive frames. If the second candidate position passes the verification in all N frames, the pose corresponding to the second candidate position is set as the initial pose; if the second candidate position fails the verification in any of the N frames, a third candidate position is selected for tracking normal distribution transformation verification for N consecutive frames until the initial pose is set; wherein, the second candidate position is the candidate position that converges and has the highest verification score among the J candidate positions.
[0036] In one possible implementation, even if the first frame of point cloud data and the point cloud data of similar locations in the map converge and achieve high scores after normal distribution transformation registration verification, the system may mistakenly set the similar scene as the initial pose. However, since the location is not actually the same, the system will obtain an unqualified result in the tracking effectiveness verification process of the subsequent N consecutive frames. As a result, the pose will be excluded due to the low matching quality, and point cloud data will be re-acquired to determine the corresponding new candidate location for pose estimation.
[0037] refer to Figure 2 , Figure 2 The method introduces a TRACKING verification state to verify the effectiveness of the pose tracking process. N can be set to 5, meaning that initialization is confirmed to be valid only after 5 consecutive frames of successful tracking. This can prevent the spread of mismatches. If any step fails, it will start over and supports unlimited retries. It can realize a fully automated closed-loop process of detection-construction-retrieval-verification-release.
[0038] It should be understood that the present invention is highly feasible and easy to integrate. It has been verified in the ROS 2 environment (Robot Operating System 2 environment, which is a second-generation open-source software framework runtime environment for robot development), proving its feasibility and stability. At the same time, the present invention adopts a modular design, which facilitates integration with other components.
[0039] This invention is user-friendly. In specific embodiments, it can provide detailed log records and performance statistics, which facilitates developers in debugging and optimizing system performance. At the same time, it can set clear error messages to quickly locate the root cause of the problem.
[0040] Please refer to Figure 3 This invention provides a pose estimation device 30 based on lidar, comprising: The first acquisition module 301 is used to acquire the first frame of point cloud data collected by the lidar. Optionally, the first acquisition module 301 includes any one of the following: The first acquisition unit is used to acquire single-frame point cloud data collected by the lidar within a single acquisition cycle; The second acquisition unit is used to acquire a set of point cloud data formed by superimposing and fusing multiple frames of point cloud data collected by the lidar in multiple consecutive acquisition cycles after being unified to the same coordinate system through pose transformation. The third acquisition unit is used to acquire structured sub-graph data generated by the lidar after acquisition and registration in multiple consecutive acquisition cycles. The structured sub-graph data includes: two-dimensional probabilistic grid, three-dimensional point cloud sub-graph, or truncated symbolic distance function voxel.
[0041] Optionally, the first acquisition module 301 includes: The fourth acquisition unit is used to acquire the first frame of point cloud data collected by the lidar; The fifth acquisition unit is used to acquire motion data collected by the inertial measurement unit and the relative pose of the coordinate system between the lidar and the inertial measurement unit. The motion data includes angular velocity data and acceleration data. The first calculation unit is used to calculate the pose increment of the inertial measurement unit based on the motion data; The second calculation unit is used to transform the first frame point cloud data based on the first frame point cloud data, the relative pose of the coordinate systems of the lidar and the inertial measurement unit, and the pose increment, and calculate the distortion-corrected first frame point cloud data.
[0042] The filtering module 302 is used to perform similarity matching between the first frame point cloud data and the data information of the map to determine K candidate positions in the map that correspond to the point cloud data, where K is a positive integer; The second acquisition module 303 is used to extract local information corresponding to the K candidate locations from the map; The calculation module 304 is used to perform normal distribution transformation verification on the first frame point cloud data and the point cloud data corresponding to the local information to obtain a verification result. The verification result is used to indicate the candidate positions that have passed the verification among the K candidate positions, and the verification score of the candidate positions that have passed the verification. Optionally, the computing module 304 includes: The third calculation unit is used to perform multi-angle transformation based on the first frame point cloud data to obtain point cloud data corresponding to the first frame point cloud data at multiple angles. The fourth calculation unit is used to perform multi-angle normal distribution transformation verification on the point cloud data corresponding to multiple angles of the first frame point cloud data and the point cloud data corresponding to the local information of the target candidate position. The point cloud data corresponding to the local information of the target candidate position includes point cloud data from multiple angles. The target candidate position is any candidate position among the K candidate positions.
[0043] The verification module 305 is used to perform continuous N-frame tracking normal distribution transformation verification based on the first candidate position. If the first candidate position passes the verification in all N frames, the pose corresponding to the first candidate position is set as the initial pose. If the first candidate position fails the verification in any of the N frames, a second candidate position is selected to perform continuous N-frame tracking normal distribution transformation verification until the initial pose is set. The first candidate position is the candidate position with the highest verification score among the K candidate positions.
[0044] The lidar-based pose estimation device provided in this embodiment of the invention is capable of executing the lidar-based pose estimation method described above. Therefore, all implementation methods in the lidar-based pose estimation method embodiments described above are applicable to this device and can achieve the same or similar beneficial effects. To avoid repetition, this embodiment will not elaborate further.
[0045] Please refer to Figure 4 This invention provides an electronic device 40, including a processor 401 and a memory 402. The memory 402 stores a program or instructions that can run on the processor 401. When the program or instructions are executed by the processor 401, they implement the various steps of the above-described lidar-based pose estimation method embodiment and achieve the same technical effect. To avoid repetition, they will not be described again here.
[0046] This application provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described lidar-based pose estimation method embodiment and achieve the same technical effect. To avoid repetition, they will not be described again here.
[0047] This invention provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, they implement the various processes of the above-described lidar-based pose estimation method embodiments and achieve the same technical effects. To avoid repetition, these will not be described again here.
[0048] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0049] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.
[0050] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of the present invention.
Claims
1. A pose estimation method based on lidar, characterized in that, include: Acquire the first frame of point cloud data collected by the lidar; The first frame of point cloud data is matched with the data information of the map to determine K candidate locations in the map that correspond to the point cloud data, where K is a positive integer; Extract local information corresponding to the K candidate locations from the map; The first frame point cloud data and the point cloud data corresponding to the local information are subjected to normal distribution transformation for verification to obtain the verification result. The verification result is used to indicate the candidate positions that have passed the verification among the K candidate positions, and the verification score of the candidate positions that have passed the verification. Based on the first candidate position, the normal distribution transformation is verified by tracking for N consecutive frames. If the first candidate position passes the verification in all N frames, the pose corresponding to the first candidate position is set as the initial pose. If the first candidate position fails to pass verification in any of the N frames, then the second candidate position is selected for continuous N-frame tracking normal distribution transformation verification until the initial pose is set; wherein, the first candidate position is the candidate position with the highest verification score among the K candidate positions.
2. The method according to claim 1, characterized in that, The step of performing a normal distribution transformation verification on the first frame point cloud data and the point cloud data corresponding to the local information includes: Based on the first frame point cloud data, perform multi-angle transformation to obtain point cloud data corresponding to the first frame point cloud data at multiple angles; The point cloud data of the first frame at multiple angles and the point cloud data corresponding to the local information of the target candidate position are verified by performing a multi-angle normal distribution transformation. The point cloud data corresponding to the local information of the target candidate position includes point cloud data at multiple angles. The target candidate position is any one of the K candidate positions.
3. The method according to claim 1, characterized in that, The first frame of point cloud data acquired by the lidar includes any one of the following: A single frame of point cloud data collected by a lidar within a single acquisition cycle; The point cloud data set is formed by superimposing and fusing multiple frames of point cloud data continuously collected by lidar in multiple acquisition cycles after being unified to the same coordinate system through pose transformation. The structured sub-graph data is generated by registering and optimizing multiple frames of point cloud data continuously acquired by the lidar in multiple acquisition cycles. The structured sub-graph data includes: two-dimensional probabilistic grids, three-dimensional point cloud sub-graphs, or truncated symbolic distance function voxels.
4. The method according to claim 1, characterized in that, The acquisition of the first frame of point cloud data collected by the lidar includes: Acquire the first frame of point cloud data collected by the lidar; The motion data collected by the inertial measurement unit and the relative pose of the lidar and the coordinate system of the inertial measurement unit are acquired. The motion data includes angular velocity data and acceleration data. The pose increment of the inertial measurement unit is calculated based on the motion data; Based on the first frame point cloud data, the relative pose of the coordinate systems of the lidar and the inertial measurement unit, and the pose increment, the first frame point cloud data is transformed to calculate the distortion-corrected first frame point cloud data.
5. A pose estimation device based on lidar, characterized in that, include: The first acquisition module is used to acquire the first frame of point cloud data collected by the lidar. The filtering module is used to perform similarity matching between the first frame point cloud data and the map data information to determine K candidate locations in the map that correspond to the point cloud data, where K is a positive integer. The second acquisition module is used to extract local information corresponding to the K candidate locations from the map; The calculation module is used to perform normal distribution transformation verification on the first frame point cloud data and the point cloud data corresponding to the local information to obtain the verification result. The verification result is used to indicate the candidate positions that have passed the verification among the K candidate positions, and the verification score of the candidate positions that have passed the verification. The verification module is used to perform continuous N-frame tracking normal distribution transformation verification based on the first candidate position. If the first candidate position passes the verification in all N frames, the pose corresponding to the first candidate position is set as the initial pose. If the first candidate position fails to pass verification in any of the N frames, then the second candidate position is selected for continuous N-frame tracking normal distribution transformation verification until the initial pose is set; wherein, the first candidate position is the candidate position with the highest verification score among the K candidate positions.
6. The apparatus according to claim 5, characterized in that, The first acquisition module includes any one of the following: The first acquisition unit is used to acquire single-frame point cloud data collected by the lidar within a single acquisition cycle; The second acquisition unit is used to acquire a set of point cloud data formed by superimposing and fusing multiple frames of point cloud data collected by the lidar in multiple consecutive acquisition cycles after being unified to the same coordinate system through pose transformation. The third acquisition unit is used to acquire structured sub-graph data generated by the lidar after acquisition and registration in multiple consecutive acquisition cycles. The structured sub-graph data includes: two-dimensional probabilistic grid, three-dimensional point cloud sub-graph, or truncated symbolic distance function voxel.
7. The apparatus according to claim 5, characterized in that, The computing module includes: The first computing unit is used to perform multi-angle transformation based on the first frame point cloud data to obtain point cloud data corresponding to the first frame point cloud data at multiple angles. The second calculation unit is used to perform multi-angle normal distribution transformation verification on the point cloud data corresponding to multiple angles of the first frame point cloud data and the point cloud data corresponding to the local information corresponding to the target candidate position. The point cloud data corresponding to the local information corresponding to the target candidate position includes point cloud data from multiple angles. The target candidate position is any candidate position among the K candidate positions.
8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps in the lidar-based pose estimation method as described in any one of claims 1 to 4.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps in the lidar-based pose estimation method as described in any one of claims 1 to 4.
10. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps in the lidar-based pose estimation method as described in any one of claims 1 to 4.