Lightweight spherical map representation method, device and equipment for point cloud data

Abstract

The invention belongs to the technical field of laser SLAM (Simultaneous Localization and Mapping), and particularly relates to a lightweight spherical map representation method, device and equipment for point cloud data. The method comprises the following steps: dividing point cloud data into at least one local point cloud according to a preset standard; projecting the local point cloud to a unit spherical surface, sampling in a preset mode based on scene attributes of the local point cloud to generate a local spherical map, establishing a globally unique index for sampling points based on a K-dimensional tree, segmenting the local point cloud into a plurality of point cloud blocks based on the sampling points, storing the structural features of the fitting plane of the point cloud block in the local spherical map; and forming lightweight spherical map representation of the point cloud data by all the local spherical maps and the corresponding poses. The invention provides a map representation method for lightweight point cloud data.

Description

Technical Field

[0001] The present disclosure belongs to the field of laser SLAM technology, and in particular relates to a lightweight spherical map representation method, apparatus, and device for point cloud data. Background Art

[0002] Laser radar (LiDAR), due to its high precision and robustness against light interference, has become a key sensor in fields such as autonomous driving, robotic navigation, and augmented reality. Simultaneous Localization and Mapping (SLAM) technology based on LiDAR, capable of building a real-time map of the environment and estimating the vehicle's position, is a core capability for autonomous operation of intelligent unmanned systems.

[0003] LiDAR SLAM systems generate massive amounts of point cloud data during large-scale mapping tasks. This point cloud data is primarily expressed in the form of point cloud maps. Traditional point cloud maps grow exponentially as the scene expands. For example, a complete point cloud map in the KITTI dataset can reach several GB, resulting in significant storage overhead. Furthermore, the disordered nature of point clouds makes efficient retrieval and updating difficult, leading to high computational costs.

[0004] To alleviate the above problems, existing studies have proposed a variety of alternative map representation methods, such as surfel expression, mesh expression, voxel grid expression, etc., but there is no essential reduction in storage and computational costs. Summary of the Invention

[0005] The disclosed embodiments propose a lightweight spherical map representation solution for point cloud data to reduce the storage and computational costs of current map representation solutions for point cloud data.

[0006] A first aspect of an embodiment of the present disclosure provides a lightweight spherical map representation method for point cloud data, comprising:

[0007] Dividing the point cloud data into at least one local point cloud according to a preset standard;

[0008] Projecting the local point cloud onto a unit sphere, sampling in a preset manner based on scene attributes of the local point cloud to generate a local spherical map, establishing a globally unique index for the sampling points based on a K-dimensional tree, segmenting the local point cloud into a plurality of point cloud blocks based on the sampling points, and storing structural features of the fitting planes of the point cloud blocks in the local spherical map;

[0009] A lightweight spherical map representation of the point cloud data is composed of all the local spherical maps and corresponding poses.

[0010] In some embodiments of the present disclosure, dividing the point cloud data into at least one local point cloud according to a preset standard includes:

[0011] In the point cloud data, a key point cloud frame is selected based on a preset standard, and a local point cloud is formed by the current key point cloud frame and all point cloud frames between the current key point cloud frame and the next key point cloud frame and the corresponding postures;

[0012] The coordinates of the local point cloud are transformed into the coordinate system of the current key point cloud frame.

[0013] In some embodiments of the present disclosure, selecting a key point cloud frame based on a preset standard includes:

[0014] If the distance between adjacent key point cloud frames of the current point cloud frame exceeds a first preset threshold, or the scene overlap ratio between the current point cloud frame and the previous key point cloud frame is lower than a second preset threshold, the current point cloud frame is used as a key point cloud frame.

[0015] In some embodiments of the present disclosure, projecting the local point cloud onto a unit sphere includes:

[0016] The local point cloud is normalized to a unit sphere so that all point cloud points are located on the unit sphere.

[0017] In some embodiments of the present disclosure, sampling the scene attributes based on the local point cloud in a preset manner includes:

[0018] Generate candidate sampling points uniformly distributed on the unit sphere using a Fibonacci grid;

[0019] The density of performing local sampling at the candidate sampling points is determined based on scene attributes of the local point cloud.

[0020] In some embodiments of the present disclosure, determining the density of performing local sampling at the candidate sampling points based on the scene attributes of the local point cloud includes:

[0021] In an autonomous driving scenario, the number of local sampling performed on the candidate sampling points distributed in the lower half of the unit sphere is reduced.

[0022] In some embodiments of the present disclosure, segmenting the local point cloud into a plurality of point cloud blocks based on sampling points includes:

[0023] Determine, based on the K-dimensional tree, the sampling point closest to the projection point of the point cloud point in the local point cloud on the unit sphere, and establish a correspondence between the point cloud point and the sampling point;

[0024] A point cloud block is composed of all the point cloud points corresponding to one sampling point.

[0025] In some embodiments of the present disclosure, the structural features of the fitting plane storing the point cloud block include:

[0026] Fitting a plane to the point cloud block using RANSAC to determine a plane index and structural parameters of a valid plane, wherein the valid plane refers to a plane fitted by the point cloud block in which the distance difference between each point cloud point does not exceed a third preset threshold, and the structural parameters include at least a normal vector and a distance, wherein the distance refers to the distance from the plane to the origin of the unit sphere;

[0027] The plane index and structure parameters of the valid plane are stored in the local spherical map.

[0028] A second aspect of an embodiment of the present disclosure provides a lightweight spherical map representation device for point cloud data, comprising:

[0029] a division module, configured to divide the point cloud data into at least one local point cloud according to a preset standard;

[0030] a sampling module, configured to project the local point cloud onto a unit sphere, generate a local spherical map by sampling in a preset manner based on scene attributes of the local point cloud, establish a globally unique index for the sampling points based on a K-dimensional tree, segment the local point cloud into a plurality of point cloud blocks based on the sampling points, and store structural features of the fitted planes of the point cloud blocks in the local spherical map;

[0031] A mapping module is used to form a lightweight spherical map representation of the point cloud data by combining all the local spherical maps and corresponding postures.

[0032] A third aspect of the present disclosure provides a lightweight spherical map representation device for point cloud data, including a memory and a processor:

[0033] The memory is used to store computer programs;

[0034] The processor is configured to implement the lightweight spherical map representation method for point cloud data according to the first aspect of the embodiment of the present disclosure when executing the computer program.

[0035] In summary, the lightweight spherical map representation method, device and equipment for point cloud data provided by the embodiments of the present disclosure realize lightweight representation of the map by projecting the local point cloud onto the unit sphere and extracting the plane parameters using a preset segmentation method. Since the spherical segmentation rules are preset, there is no need to store the point cloud coordinates, and only the plane parameters need to be saved to restore the geometric structure, so the required storage space is greatly reduced (only 60MB is required to store the KITTI dataset, which is 0.7% of the original point cloud). At the same time, establishing a globally unique index based on the K-dimensional tree avoids the computational overhead of repeatedly constructing the index during the retrieval process. Therefore, this method significantly reduces the storage and computing costs while ensuring geometric accuracy, and is suitable for large-scale long-term mapping tasks. In addition, the present disclosure adaptively determines the generation frequency and sampling point density of the local map based on scene attributes, further reducing the storage and computing overhead of point cloud data in autonomous driving scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The features and advantages of the present disclosure will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present disclosure in any way. In the accompanying drawings:

[0037] Figure 1 is a schematic diagram of a computer system to which the present disclosure is applicable;

[0038] Figure 2 is a flowchart of a lightweight spherical map representation method for point cloud data according to some embodiments of the present disclosure;

[0039] Figure 3 is an example of a local point cloud constructed using an arbitrary laser odometry in one embodiment of the present disclosure;

[0040] Figure 4 is an example of uniformly distributed candidate sampling points generated by sampling Fibonacci grid points on a unit sphere in some embodiments of the present disclosure;

[0041] Figure 5 This is an example of grouping of local point clouds generated by segmenting the local point cloud according to spherical sampling points in one embodiment of the present disclosure;

[0042] Figure 6 This is an example of extracting the normal vector of each point cloud set in one embodiment of the present disclosure;

[0043] Figure 7 This is a comparison between a BALLmap constructed on the KITTI 00 sequence and a point cloud map according to an embodiment of the present disclosure;

[0044] Figure 8 Comparison of storage space required by BALLmap and other map representation methods on KITTI 00, 05, and 08 sequences;

[0045] Figure 9 is a schematic diagram of a lightweight spherical map representation device for point cloud data according to some embodiments of the present disclosure;

[0046] Figure 10 This is a schematic diagram of a lightweight spherical map representation device for point cloud data according to some embodiments of the present disclosure. DETAILED DESCRIPTION

[0047] In the detailed description that follows, many specific details of the present disclosure are set forth by way of example in order to provide a thorough understanding of the relevant disclosure. However, it will be apparent to one of ordinary skill in the art that the present disclosure can be implemented without these details. It should be understood that the use of the terms "system," "device," "unit," and / or "module" in the present disclosure is a method for distinguishing between different parts, elements, parts, or assemblies at different levels in a sequential arrangement. However, these terms may be replaced by other expressions if they can achieve the same purpose.

[0048] It should be understood that when a device, unit, or module is referred to as being "on," "connected to," or "coupled to" another device, unit, or module, it may be directly on, connected to, coupled to, or in communication with the other device, unit, or module, or there may be intervening devices, units, or modules, unless the context clearly indicates an exception. For example, the term "and / or" as used in this disclosure includes any and all combinations of one or more of the associated listed items.

[0049] The terms used in this disclosure are only for describing specific embodiments and are not intended to limit the scope of this disclosure. As shown in the specification and claims of this disclosure, unless the context clearly indicates an exception, the words "a", "an", "a kind" and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of clearly identified features, wholes, steps, operations, elements and / or components, and such expressions do not constitute an exclusive list, and other features, wholes, steps, operations, elements and / or components may also be included.

[0050] These and other features and characteristics of the present disclosure, as well as the methods of operation, the functions of the related elements of the structure, the combination of parts, and the economy of manufacture may be better understood with reference to the following description and accompanying drawings, which form a part of this specification. However, it is to be expressly understood that the drawings are for illustration and description purposes only and are not intended to limit the scope of protection of the present disclosure. It is to be understood that the drawings are not drawn to scale.

[0051] Various structural diagrams are used in this disclosure to illustrate various variations of the embodiments of the present disclosure. It should be understood that the preceding or following structures are not intended to limit the present disclosure. The scope of protection of the present disclosure is subject to the claims.

[0052] Figure 1 is a schematic diagram of a computer system to which the present disclosure is applicable. Figure 1 The computer system shown includes a laser radar that performs SLAM tasks on a target area and a mapping server connected to the laser radar. The mapping server generates a lightweight spherical map representation of the point cloud data based on the point cloud data generated by the laser radar on the target area.

[0053] Laser radar (LiDAR) generates point cloud data by emitting laser pulses and receiving the target's reflected signal (based on the Time of Flight principle). Due to its high precision and resistance to light interference, LiDAR has become a key sensor in fields such as autonomous driving, robotic navigation, and augmented reality. This disclosure does not limit the type of LiDAR; any LiDAR based on the Time of Flight principle that can generate point cloud data for a target area is applicable to this disclosure.

[0054] The mapping server acquires the point cloud data generated by the LiDAR during SLAM tasks and stores this data in a structured manner to form a map representation of the point cloud data. The mapping server must also support searching the point cloud data within the map representation. The mapping server can be a standalone server, a cluster, or a distributed server.

[0055] SLAM (Simultaneous Localization and Mapping) refers to the simultaneous positioning and construction of an environmental map. LiDAR-based SLAM technology can construct environmental maps and estimate its own position in real time, making it a core capability for autonomous operation of intelligent unmanned systems. In recent years, with the improvement of LiDAR hardware performance and the optimization of SLAM algorithms, laser SLAM has achieved centimeter-level short-term positioning accuracy.

[0056] The traditional map representation of point cloud data is a point cloud map, which stores the geometric information of point cloud data in three-dimensional space, including coordinates and spatial distribution characteristics. As SLAM accuracy improves, LiDAR SLAM systems based on point cloud maps face the following problems in large-scale mapping tasks:

[0057] 1. Huge storage overhead: Traditional point cloud maps grow exponentially as the scene expands. For example, a complete point cloud map of the KITTI dataset can reach several GB, making it difficult to store and transmit on resource-constrained devices.

[0058] 2. Computational efficiency bottleneck: Operations such as point cloud registration in SLAM tasks rely on frequent nearest neighbor searches. However, the disorder of point clouds in point cloud maps makes it difficult to retrieve and update them efficiently, resulting in high computational costs.

[0059] To alleviate these issues, existing research has proposed a variety of alternative map representation methods, but these still have significant limitations. For example, while surfels and meshes are more compact than point clouds, they still require storing the position and normal vector of each patch, resulting in limited storage compression (for example, SUMA still requires hundreds of MB to store a KITTI map). Implicit neural representations (such as NeRF-SLAM) can store maps with high compression rates, but they rely on extensive training computations and lack the accuracy to recover geometric details, making them difficult to meet high-precision positioning requirements. Voxel grids reduce storage through downsampling, but this can result in a loss of geometric detail.

[0060] Therefore, in order to perform large-scale mapping tasks with high precision, the LiDAR SLAM system requires a map representation method that can greatly reduce storage and computational costs while ensuring geometric accuracy.

[0061] To this end, this paper proposes a lightweight spherical map representation method (BALLmap) for LiDAR positioning. This method uses a lightweight spherical unit (BALL) structure to achieve efficient point cloud data storage and fast retrieval, making it suitable for SLAM mapping and localization tasks in large-scale environments. The core idea of ​​BALLmap is to decompose the global point cloud map into multiple lightweight local units (BALLs), each of which stores the local environment geometry information at its location.

[0062] Figure 2 is a flowchart of a method for representing a light-weight spherical map of point cloud data according to some embodiments of the present disclosure. In some embodiments, the light-weight spherical map representation method of point cloud data is composed of Figure 1 The mapping server shown is executed, and the method includes the following steps:

[0063] S210: Divide the point cloud data into at least one local point cloud according to a preset standard.

[0064] First, key point cloud frames are selected from the point cloud data.

[0065] Specifically, the point cloud data described in this disclosure refers to the continuous point cloud frames S obtained by any LiDAR odometer. i and its corresponding pose P i The present disclosure adaptively selects key point cloud frames from the point cloud data.

[0066] In some embodiments of the present disclosure, the strategy for selecting key point cloud frames is:

[0067] When the distance between adjacent key point cloud frames exceeds the set distance threshold (such as 6 meters), or the scene overlap ratio between the current point cloud frame and the previous key point cloud frame is less than 70% (for example, when turning a corner or entering or exiting a room), a new key point cloud frame will be automatically generated and the creation of a new BALL will be triggered.

[0068] The strategy of adaptively creating new BALLs balances the map coverage density and storage efficiency, avoiding the generation of redundant BALLs.

[0069] Then, the local point cloud for generating the BALL is determined based on the key point cloud frame.

[0070] The local point cloud is composed of the current key point cloud frame and all point cloud frames between the current key point cloud frame and the next key point cloud frame and their corresponding poses. The formal expression is:

[0071] For the key point cloud frame k i , according to the pose P j Accumulate its subsequent frame point cloud S j , until the next key point cloud frame k i+1 , forming a local point cloud Right now where k i ≤j≤k i+1 .

[0072] Then the local point cloud Convert to In the coordinate system, it is recorded as Right now

[0073] Figure 3 This is an example of a local point cloud constructed using an arbitrary laser odometry in one embodiment of the present disclosure.

[0074] S220, projecting the local point cloud onto a unit sphere, sampling in a preset manner based on the scene attributes of the local point cloud to generate a local spherical map, establishing a globally unique index for the sampling points based on a K-dimensional tree, dividing the local point cloud into multiple point cloud blocks based on the sampling points, and storing the structural features of the fitting planes of the point cloud blocks in the local spherical map.

[0075] First, project the local point cloud onto the unit sphere. Specifically:

[0076] Will Normalize to the unit sphere so that all points lie on the unit sphere.

[0077] Then perform spherical sampling on the local point cloud in a preset way. Specifically:

[0078] 1. Use the Fibonacci lattice to sample the unit sphere and generate uniformly distributed candidate sampling points, such as Figure 4 shown.

[0079] 2. Perform actual sampling on the candidate sampling points to generate a spherical point cloud In some embodiments of the present disclosure, the sampling density at candidate sampling points is adaptively determined based on scene attributes of the local point cloud. For example, in autonomous driving scenarios, because the distance from the ground to the LiDAR is closer than the distance from street buildings to the LiDAR, the number of samples in the lower half of the sphere can be reduced to balance the number of points in the local point cloud corresponding to each spherical point.

[0080] 3. Construct a globally unique K-dimensional tree Tree for the spherical point cloud F F .

[0081] 4. For each point in the tree, F Find the nearest neighbor spherical point p F , and press p F Grouping to form N sp A point cloud set G j . Figure 5 In one embodiment of the present disclosure, An example of grouping of the generated local point clouds after segmentation.

[0082] Then the spatial distribution features of the local point cloud are extracted and stored to form a local spherical map. Specifically:

[0083] For each group G j , use RANSAC to fit the plane and calculate its normal vector n j and distance d j (The distance from the plane to the origin of the unit sphere). If the distance difference between the point clouds in the group is too large (such as the background and foreground are mixed), it is determined to be an invalid plane and removed. By removing the invalid plane, the storage efficiency is further improved. Finally, a BALL representing the local point cloud map is formed. It stores the parameters of all valid planes (n j ,d j ), and the corresponding spherical index.

[0084] Figure 6 This is an example of extracting the normal vector of each set of point clouds in one embodiment of the present disclosure.

[0085] S230, forming a lightweight spherical map representation of the point cloud data by including all the local spherical maps and corresponding postures.

[0086] Finally, all BALLs and their positions A complete lightweight spherical map representation BALLmap.

[0087] Figure 7 Figure 1 is a comparison of a ballmap and a point cloud map constructed on the KITTI 00 sequence according to an embodiment of the present disclosure. (b) and (d) are point cloud maps, (a) is the ballmap corresponding to (b), and (c) is the ballmap corresponding to (d).

[0088] Since BALL only stores geometric parameters rather than the original point cloud, BALLmap is a "lightweight" map representation of point cloud data, which greatly reduces the storage space required (storing the KITTI dataset only requires about 60MB). One embodiment of the present disclosure compares the storage space required by BALLmap with other map representation methods on the KITTI 00, 05, and 08 sequences. The results are as follows: Figure 8 shown. Figure 8 The unit is MB. Figure 8 The storage space required by BALLmap is much smaller than that required by other map representation methods.

[0089] At the same time, the globally unique K-dimensional tree also provides a fast retrieval mechanism, avoiding the computational overhead of repeated index construction and improving the registration efficiency of SLAM tasks.

[0090] Figure 9 Schematic diagram of a lightweight spherical map representation device for point cloud data according to some embodiments of the present disclosure. Figure 9 As shown, the light spherical map representation device 900 for point cloud data includes a partitioning module 910, a sampling module 920, and a mapping module 930. In some embodiments of the present disclosure, the light spherical map representation function of the point cloud data is provided by Figure 1 The mapping server is executed as shown.

[0091] A division module 910 is configured to divide the point cloud data into at least one local point cloud according to a preset standard;

[0092] Sampling module 920 is configured to project the local point cloud onto a unit sphere, generate a local spherical map by sampling in a preset manner based on scene attributes of the local point cloud, establish a globally unique index for the sampling points based on a K-dimensional tree, segment the local point cloud into a plurality of point cloud blocks based on the sampling points, and store structural features of the fitted planes of the point cloud blocks in the local spherical map;

[0093] The mapping module 930 is configured to construct a lightweight spherical map representation of the point cloud data using all the local spherical maps and corresponding poses.

[0094] Figure 10 FIG. 1 is a schematic diagram of a lightweight spherical map representation device for point cloud data according to some embodiments of the present disclosure. Figure 10 As shown, the lightweight spherical map representation device 1000 of point cloud data includes a memory 1020 and a processor 1010, wherein the memory 1020 is used to store a computer program; the processor 1010 is used to implement when executing the computer program Figure 2 The lightweight spherical map representation method of point cloud data described in S210-S230.

[0095] In summary, the lightweight spherical map representation method, device and equipment for point cloud data provided by the embodiments of the present disclosure realize lightweight representation of the map by projecting the local point cloud onto the unit sphere and extracting the plane parameters using a preset segmentation method. Since the spherical segmentation rules are preset, there is no need to store the point cloud coordinates, and only the plane parameters need to be saved to restore the geometric structure, so the required storage space is greatly reduced (only 60MB is required to store the KITTI dataset, which is 0.7% of the original point cloud). At the same time, establishing a globally unique index based on the K-dimensional tree avoids the computational overhead of repeatedly constructing the index during the retrieval process. Therefore, this method significantly reduces storage and computing costs while ensuring geometric accuracy, and is suitable for large-scale long-term mapping tasks. In addition, some embodiments of the present disclosure adaptively determine the generation frequency and sampling point density of local maps based on scene attributes, further reducing the storage and computing overhead of point cloud data in autonomous driving scenarios.

[0096] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding descriptions in the aforementioned device embodiments and will not be repeated here.

[0097] Although the subject matter described herein is provided in the general context of being executed in conjunction with the execution of an operating system and application programs on a computer system, those skilled in the art will recognize that other implementations may also be performed in conjunction with other types of program modules. Generally speaking, program modules include routines, programs, components, data structures, and other types of structures that perform specific tasks or implement specific abstract data types. Those skilled in the art will appreciate that the subject matter described herein may be practiced using other computer system configurations, including handheld devices, multiprocessor systems, microprocessor-based or programmable consumer electronics, minicomputers, mainframe computers, and the like, and may also be used in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

[0098] Those skilled in the art will appreciate that the units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0099] It should be understood that the above-described specific embodiments of the present disclosure are merely illustrative of or explanation of the principles of the present disclosure and do not constitute limitations on the present disclosure. Therefore, any modifications, equivalent substitutions, improvements, etc. made without departing from the spirit and scope of the present disclosure shall be included within the scope of protection of the present disclosure. In addition, the claims appended to the present disclosure are intended to cover all variations and modifications that fall within the scope and metes and bounds of the appended claims, or equivalents of such scope and metes and bounds.

Claims

1. A lightweight spherical map representation method for point cloud data, characterized in that: include: Dividing the point cloud data into at least one local point cloud according to a preset standard; Projecting the local point cloud onto a unit sphere, sampling in a preset manner based on scene attributes of the local point cloud to generate a local spherical map, establishing a globally unique index for the sampling points based on a K-dimensional tree, segmenting the local point cloud into a plurality of point cloud blocks based on the sampling points, and storing structural features of the fitting planes of the point cloud blocks in the local spherical map; A lightweight spherical map representation of the point cloud data is composed of all the local spherical maps and corresponding poses.

2. The method according to claim 1, characterized in that The dividing the point cloud data into at least one local point cloud according to a preset standard comprises: In the point cloud data, a key point cloud frame is selected based on a preset standard, and a local point cloud is formed by the current key point cloud frame and all point cloud frames between the current key point cloud frame and the next key point cloud frame and the corresponding postures; The coordinates of the local point cloud are transformed into the coordinate system of the current key point cloud frame.

3. The method according to claim 2, characterized in that The selecting of key point cloud frames based on preset standards includes: If the distance between adjacent key point cloud frames of the current point cloud frame exceeds a first preset threshold, or the scene overlap ratio between the current point cloud frame and the previous key point cloud frame is lower than a second preset threshold, the current point cloud frame is used as a key point cloud frame.

4. The method according to claim 3, characterized in that Projecting the local point cloud onto a unit sphere includes: The local point cloud is normalized to a unit sphere so that all point cloud points are located on the unit sphere.

5. The method according to claim 4, characterized in that: The sampling of scene attributes based on the local point cloud in a preset manner includes: Generate candidate sampling points uniformly distributed on the unit sphere using a Fibonacci grid; The density of performing local sampling at the candidate sampling points is determined based on scene attributes of the local point cloud.

6. The method according to claim 5, characterized in that The determining of the density of performing local sampling at the candidate sampling points based on the scene attributes of the local point cloud includes: In an autonomous driving scenario, the number of local sampling performed on the candidate sampling points distributed in the lower half of the unit sphere is reduced.

7. The method according to claim 5, characterized in that The dividing the local point cloud into a plurality of point cloud blocks based on the sampling points comprises: Determine, based on the K-dimensional tree, the sampling point closest to the projection point of the point cloud point in the local point cloud on the unit sphere, and establish a correspondence between the point cloud point and the sampling point; A point cloud block is composed of all the point cloud points corresponding to one sampling point.

8. The method according to claim 7, characterized in that: The structural features of the fitting plane storing the point cloud block include: Fitting a plane to the point cloud block using RANSAC to determine a plane index and structural parameters of a valid plane, wherein the valid plane refers to a plane fitted by the point cloud block in which the distance difference between each point cloud point does not exceed a third preset threshold, and the structural parameters include at least a normal vector and a distance, wherein the distance refers to the distance from the plane to the origin of the unit sphere; The plane index and structure parameters of the valid plane are stored in the local spherical map.

9. A lightweight spherical map representation device for point cloud data, characterized in that: include: a division module, configured to divide the point cloud data into at least one local point cloud according to a preset standard; a sampling module, configured to project the local point cloud onto a unit sphere, generate a local spherical map by sampling in a preset manner based on scene attributes of the local point cloud, establish a globally unique index for the sampling points based on a K-dimensional tree, segment the local point cloud into a plurality of point cloud blocks based on the sampling points, and store structural features of the fitted planes of the point cloud blocks in the local spherical map; A mapping module is used to form a lightweight spherical map representation of the point cloud data by combining all the local spherical maps and corresponding postures.

10. A lightweight spherical map representation device for point cloud data, characterized in that: Including memory and processor: The memory is used to store computer programs; The processor is configured to implement the lightweight spherical map representation method for point cloud data according to any one of claims 1 to 8 when executing the computer program.

Images