Point cloud mapping method, computer device and computer readable storage medium

Abstract

The present application relates to the technical field of automatic driving, and particularly provides a point cloud mapping method, a computer device and a computer readable storage medium, aiming at solving the problem of how to efficiently and low-costly map specified areas such as parking lots and energy supplement sites. To this end, the method provided by the present application comprises: acquiring laser radar point cloud key frames collected after an authorized vehicle enters a specified area; establishing a new map of the specified area according to the laser radar point cloud key frames; acquiring a regional base map of the specified area; respectively updating each second map tile on the regional base map according to each first map tile on the new map to obtain an updated regional base map; and the authorized vehicle is a vehicle initiating mapping authorization of the specified area. Through the above implementation, without professional map data acquisition vehicles, the ordinary vehicle initiating mapping authorization can be used to dynamically, efficiently, accurately and low-costly acquire mapping data of each specified area, and complete map establishment.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, specifically to a point cloud mapping method, a computer device, and a computer-readable storage medium. Background Technology

[0002] When controlling autonomous vehicles, maps are typically used for vehicle localization and / or trajectory planning to ensure safe and reliable autonomous driving on roads. Since vehicles spend most of their time in high-speed scenarios (high-speed scenarios) such as urban roads and highways, and only spend a small portion of their time in low-speed scenarios (low-speed scenarios) such as parking lots, charging stations, and highway service areas, and given the large number and diverse structures of low-speed scenarios, current maps for autonomous driving typically only cover these high-speed scenarios due to cost and efficiency considerations. When vehicles enter low-speed scenarios such as parking lots, maps cannot be used to continue safe and reliable autonomous driving control.

[0003] Accordingly, a new technical solution is needed in this field to solve the above problems. Summary of the Invention

[0004] To overcome the above-mentioned shortcomings, this invention is proposed to provide a point cloud mapping method, computer equipment, and computer-readable storage medium that solves, or at least partially solves, the technical problem of how to efficiently and cost-effectively map low-speed scenes such as parking lots and refueling sites.

[0005] In a first aspect, a point cloud mapping method is provided, the method comprising:

[0006] Obtain key frames of lidar point cloud collected after an authorized vehicle enters a designated area;

[0007] A new map of the specified area is created based on the keyframes of the lidar point cloud.

[0008] Obtain the base map of the specified area;

[0009] Based on each first map tile on the new map, update the map for each second map tile on the regional base map to obtain the updated regional base map;

[0010] The authorized vehicle is the vehicle that initiates the authorization for mapping the designated area, and each first map tile corresponds one-to-one with each second map tile.

[0011] In one technical solution of the above-mentioned point cloud mapping method, before the step of "updating the map of each second map tile on the regional base map according to each first map tile on the new map to obtain the updated regional base map", the method further includes:

[0012] Align the new map with the base map of the region.

[0013] In one technical solution of the above point cloud mapping method, the step of "aligning the new map with the regional base map" specifically includes aligning each keyframe of the LiDAR point cloud on the new map with the regional base map in the following ways:

[0014] Based on the pose of the current lidar point cloud keyframe on the new map, search for the nearest neighbor lidar point cloud keyframe from the lidar point cloud keyframes on the regional base map.

[0015] Obtain a local map from the base map of the region, including the keyframes of the nearest neighbor lidar point cloud;

[0016] Point cloud matching is performed between the current lidar point cloud keyframe and the local map to determine the relative pose between the current lidar point cloud keyframe and the local map.

[0017] The pose transformation of the current lidar point cloud keyframe is performed based on the relative pose so that the current lidar point cloud keyframe is aligned with the region base map.

[0018] In one technical solution of the above point cloud mapping method, the step of "obtaining a local map from the regional base map, including the keyframes of the nearest neighbor lidar point cloud" specifically includes:

[0019] Select multiple lidar point cloud keyframes from the base map of the region that are located before and after the nearest neighbor lidar point cloud keyframe.

[0020] The local map is formed by overlaying multiple key frames of the lidar point cloud before and after the aforementioned key frames, as well as the key frames of the nearest-neighbor lidar point cloud.

[0021] In one technical solution of the above point cloud mapping method, the step of "updating the map for each second map tile on the regional base map based on each first map tile on the new map to obtain the updated regional base map" specifically includes:

[0022] Obtain the first map tile and the second map tile with the same tile position;

[0023] Map difference is performed on the first map tile and the second map tile at the same tile position, and the second map tile is updated based on the result of the map difference.

[0024] In one technical solution of the above point cloud mapping method, the step of "updating each second map tile on the regional base map according to each first map tile on the new map to obtain the updated regional base map" further includes:

[0025] Output the first map tile of the new map and the second map tile of the region base map;

[0026] In response to a received manual update instruction for the second map tile, the second map tile is updated according to the manual update instruction.

[0027] In one technical solution of the above point cloud mapping method, the step of "creating a new map of the specified area based on the keyframes of the lidar point cloud" specifically includes:

[0028] Acquire key frames of lidar point cloud collected after the authorized vehicle has entered the designated area multiple times;

[0029] Based on the keyframes of the lidar point cloud collected each time the authorized vehicle enters the designated area, a single map is created.

[0030] A factor graph is created based on the single map, the factor graph including factor nodes and relative pose constraints;

[0031] Based on the factor map, the pose of each key frame of the LiDAR point cloud on each single map is optimized to obtain the optimized pose of each key frame of the LiDAR point cloud.

[0032] Based on the optimized pose of each LiDAR point cloud keyframe, each LiDAR point cloud keyframe is superimposed to form a new map of the specified area.

[0033] Each of the factor nodes corresponds one-to-one with each of the LiDAR point cloud keyframes, and the relative pose constraint is the constraint on the relative pose between two LiDAR point cloud keyframes that can form a loop on different single maps.

[0034] In one technical solution of the above point cloud mapping method, the method further includes creating the relative pose constraint term in the following manner:

[0035] Acquire two keyframes of LiDAR point clouds that can form a loop on different single maps;

[0036] From the single map containing the first LiDAR point cloud keyframe in the two LiDAR point cloud keyframes, obtain a first local map including the first LiDAR point cloud keyframe.

[0037] From the single map containing the second LiDAR point cloud keyframe in the two LiDAR point cloud keyframes, obtain a second local map including the second LiDAR point cloud keyframe.

[0038] Point cloud matching is performed on the first local map and the second local map to determine the matching relative pose between the first local map and the second local map, and the matching relative pose is used as the optimal relative pose between the two lidar point cloud keyframes.

[0039] Based on the optimal relative pose, constraints are established for the relative pose between the two lidar point cloud keyframes.

[0040] In one technical solution of the above-mentioned point cloud mapping method, before the step of "performing point cloud matching between the first local map and the second local map to determine the matching relative pose between the first local map and the second local map", the method further includes:

[0041] The authorized vehicle acquires two external images captured by the same image acquisition device on the authorized vehicle at the same time as the first and second lidar point cloud keyframes are acquired.

[0042] Image matching is performed on the two vehicle exterior images, and the initial relative pose between the first lidar point cloud keyframe and the second lidar point cloud keyframe is determined based on the image matching result.

[0043] The step of "performing point cloud matching between the first local map and the second local map to determine the relative pose of the matching between the first local map and the second local map" specifically includes:

[0044] The initial pose is used as the initial value of the relative pose between the first local map and the second local map. Point cloud matching is performed on the first local map and the second local map based on the initial value to determine the matching relative pose between the first local map and the second local map.

[0045] In one technical solution of the above point cloud mapping method, the step of "obtaining a first local map including the first LiDAR point cloud keyframe from the single map where the first LiDAR point cloud keyframe is located among the two LiDAR point cloud keyframes" specifically includes:

[0046] Select multiple lidar point cloud keyframes from the single map, located before and after the first lidar point cloud keyframe.

[0047] The first local map is formed by overlaying multiple keyframes of the lidar point cloud before and after the first lidar point cloud keyframe.

[0048] In one technical solution of the above point cloud mapping method, the step of "obtaining a second local map including the second LiDAR point cloud keyframe from the single map where the second LiDAR point cloud keyframe is located" specifically includes:

[0049] Select multiple lidar point cloud keyframes from the single map, located before and after the second lidar point cloud keyframe.

[0050] The second local map is formed by overlaying multiple key frames of the LiDAR point cloud before and after the first LiDAR point cloud and the second LiDAR point cloud key frame.

[0051] In one technical solution of the above point cloud mapping method, the step of "obtaining two key frames of LiDAR point clouds that can form a loop on different single maps" specifically includes:

[0052] Based on the absolute pose of each lidar point cloud keyframe on each single map, determine two lidar point cloud keyframes on different single maps that can form a loop.

[0053] In one technical solution of the above point cloud mapping method, the step of "obtaining two key frames of lidar point clouds that can form a loop on different single maps" also includes:

[0054] The external image of the vehicle is acquired by the same image acquisition device on the authorized vehicle while the authorized vehicle is acquiring each key frame of the lidar point cloud;

[0055] Similarity analysis is performed on the vehicle exterior images corresponding to each keyframe of the LiDAR point cloud on different single maps to obtain at least one similar image group, wherein the similar image group includes two similar vehicle exterior images that correspond to the keyframes of the LiDAR point cloud on different single maps respectively.

[0056] For each group of similar images, the key frames of the LiDAR point cloud corresponding to the two vehicle exterior images in the group of similar images are taken as the two key frames of the LiDAR point cloud that can form a loop on different single maps.

[0057] In a second aspect, a computer device is provided, comprising a processor and a storage device, the storage device being adapted to store a plurality of program codes, the program codes being adapted to be loaded and run by the processor to perform the point cloud mapping method described in any of the above-described technical solutions.

[0058] In a third aspect, a computer-readable storage medium is provided, wherein a plurality of program codes are stored therein, the program codes being adapted to be loaded and run by a processor to perform the point cloud mapping method described in any of the above-described technical solutions.

[0059] Solution 1. A point cloud mapping method, characterized in that the method includes:

[0060] Obtain key frames of lidar point cloud collected after an authorized vehicle enters a designated area;

[0061] A new map of the specified area is created based on the keyframes of the lidar point cloud.

[0062] Obtain the base map of the specified area;

[0063] Based on each first map tile on the new map, update the map for each second map tile on the regional base map to obtain the updated regional base map;

[0064] The authorized vehicle is the vehicle that initiates the authorization for mapping the designated area, and each first map tile corresponds one-to-one with each second map tile.

[0065] Solution 2. The point cloud mapping method according to Solution 1, characterized in that, before the step of "updating the map of each second map tile on the regional base map according to each first map tile on the new map, to obtain the updated regional base map", the method further includes:

[0066] Align the new map with the base map of the region.

[0067] Solution 3. The point cloud mapping method according to Solution 2, characterized in that the step of "aligning the new map with the regional base map" specifically includes aligning each keyframe of the lidar point cloud on the new map with the regional base map in the following ways:

[0068] Based on the pose of the current lidar point cloud keyframe on the new map, search for the nearest neighbor lidar point cloud keyframe from the lidar point cloud keyframes on the regional base map.

[0069] Obtain a local map from the base map of the region, including the keyframes of the nearest neighbor lidar point cloud;

[0070] Point cloud matching is performed between the current lidar point cloud keyframe and the local map to determine the relative pose between the current lidar point cloud keyframe and the local map.

[0071] The pose transformation of the current lidar point cloud keyframe is performed based on the relative pose so that the current lidar point cloud keyframe is aligned with the region base map.

[0072] Solution 4. The point cloud mapping method according to Solution 3, characterized in that the step of "obtaining a local map from the regional base map including the keyframes of the nearest neighbor lidar point cloud" specifically includes:

[0073] Select multiple lidar point cloud keyframes from the base map of the region that are located before and after the nearest neighbor lidar point cloud keyframe.

[0074] The local map is formed by overlaying multiple key frames of the lidar point cloud before and after the aforementioned key frames, as well as the key frames of the nearest-neighbor lidar point cloud.

[0075] Solution 5. The point cloud mapping method according to Solution 1, characterized in that the step of "updating the map of each second map tile on the regional base map according to each first map tile on the new map, to obtain the updated regional base map" specifically includes:

[0076] Obtain the first map tile and the second map tile with the same tile position;

[0077] Map difference is performed on the first map tile and the second map tile at the same tile position, and the second map tile is updated based on the result of the map difference.

[0078] Solution 6. The point cloud mapping method according to Solution 1, characterized in that the step of "updating the map of each second map tile on the regional base map according to each first map tile on the new map to obtain the updated regional base map" further includes:

[0079] Output the first map tile of the new map and the second map tile of the region base map;

[0080] In response to a received manual update instruction for the second map tile, the second map tile is updated according to the manual update instruction.

[0081] Solution 7. The point cloud mapping method according to Solution 1, characterized in that the step of "establishing a new map of the specified area based on the key frames of the lidar point cloud" specifically includes:

[0082] Acquire key frames of lidar point cloud collected after the authorized vehicle has entered the designated area multiple times;

[0083] Based on the keyframes of the lidar point cloud collected each time the authorized vehicle enters the designated area, a single map is created.

[0084] A factor graph is created based on the single map, the factor graph including factor nodes and relative pose constraints;

[0085] Based on the factor map, the pose of each key frame of the LiDAR point cloud on each single map is optimized to obtain the optimized pose of each key frame of the LiDAR point cloud.

[0086] Based on the optimized pose of each LiDAR point cloud keyframe, each LiDAR point cloud keyframe is superimposed to form a new map of the specified area.

[0087] Each of the factor nodes corresponds one-to-one with each of the LiDAR point cloud keyframes, and the relative pose constraint is the constraint on the relative pose between two LiDAR point cloud keyframes that can form a loop on different single maps.

[0088] Solution 8. The point cloud mapping method according to Solution 7, characterized in that the method further includes creating the relative pose constraint term in the following manner:

[0089] Acquire two keyframes of LiDAR point clouds that can form a loop on different single maps;

[0090] From the single map containing the first LiDAR point cloud keyframe in the two LiDAR point cloud keyframes, obtain a first local map including the first LiDAR point cloud keyframe.

[0091] From the single map containing the second LiDAR point cloud keyframe in the two LiDAR point cloud keyframes, obtain a second local map including the second LiDAR point cloud keyframe.

[0092] Point cloud matching is performed on the first local map and the second local map to determine the matching relative pose between the first local map and the second local map, and the matching relative pose is used as the optimal relative pose between the two lidar point cloud keyframes.

[0093] Based on the optimal relative pose, constraints are established for the relative pose between the two lidar point cloud keyframes.

[0094] Solution 9. The point cloud mapping method according to Solution 8, characterized in that, before the step of "performing point cloud matching between the first local map and the second local map to determine the matching relative pose between the first local map and the second local map", the method further includes:

[0095] The authorized vehicle acquires two external images captured by the same image acquisition device on the authorized vehicle at the same time as the first and second lidar point cloud keyframes are acquired.

[0096] Image matching is performed on the two vehicle exterior images, and the initial relative pose between the first lidar point cloud keyframe and the second lidar point cloud keyframe is determined based on the image matching result.

[0097] The step of "performing point cloud matching between the first local map and the second local map to determine the relative pose of the matching between the first local map and the second local map" specifically includes:

[0098] The initial pose is used as the initial value of the relative pose between the first local map and the second local map. Point cloud matching is performed on the first local map and the second local map based on the initial value to determine the matching relative pose between the first local map and the second local map.

[0099] Solution 10. The point cloud mapping method according to Solution 8, characterized in that the step of "obtaining a first local map including the first LiDAR point cloud keyframe from the single map where the first LiDAR point cloud keyframe is located among the two LiDAR point cloud keyframes" specifically includes:

[0100] Select multiple lidar point cloud keyframes from the single map, located before and after the first lidar point cloud keyframe.

[0101] The first local map is formed by overlaying multiple keyframes of the lidar point cloud before and after the first lidar point cloud keyframe.

[0102] Solution 11. The point cloud mapping method according to Solution 8, characterized in that the step of "obtaining a second local map including the second lidar point cloud keyframe from the single map where the second lidar point cloud keyframe is located among the two lidar point cloud keyframes" specifically includes:

[0103] Select multiple lidar point cloud keyframes from the single map, located before and after the second lidar point cloud keyframe.

[0104] The second local map is formed by overlaying multiple key frames of the LiDAR point cloud before and after the first LiDAR point cloud and the second LiDAR point cloud key frame.

[0105] Solution 12. The point cloud mapping method according to Solution 8, characterized in that the step of "obtaining two key frames of lidar point clouds that can form a loop on different single maps" specifically includes:

[0106] Based on the absolute pose of each LiDAR point cloud keyframe on each single map, determine two LiDAR point cloud keyframes on different single maps that can form a loop.

[0107] Scheme 13. The point cloud mapping method according to Scheme 8, characterized in that the step of "obtaining two key frames of lidar point clouds that can form a loop on different single maps" further includes:

[0108] The external image of the vehicle is acquired by the same image acquisition device on the authorized vehicle while the authorized vehicle is acquiring each key frame of the lidar point cloud;

[0109] Similarity analysis is performed on the vehicle exterior images corresponding to each keyframe of the LiDAR point cloud on different single maps to obtain at least one similar image group, wherein the similar image group includes two similar vehicle exterior images that correspond to the keyframes of the LiDAR point cloud on different single maps respectively.

[0110] For each group of similar images, the key frames of the LiDAR point cloud corresponding to the two vehicle exterior images in the group of similar images are taken as the two key frames of the LiDAR point cloud that can form a loop on different single maps.

[0111] Scheme 14. A computer device comprising a processor and a storage device, the storage device being adapted to store a plurality of program codes, characterized in that the program codes are adapted to be loaded and run by the processor to perform the point cloud mapping method as described in any one of Schemes 1 to 13.

[0112] Scheme 15. A computer-readable storage medium storing a plurality of program codes, characterized in that the program codes are adapted to be loaded and run by a processor to perform the point cloud mapping method described in any one of Schemes 1 to 13.

[0113] The above-described technical solutions of the present invention have at least one or more of the following beneficial effects:

[0114] In implementing the technical solution of this invention, key frames of LiDAR point clouds collected after an authorized vehicle enters a designated area can be obtained. Then, a new map of the designated area can be built based on the key frames of the LiDAR point clouds. Finally, a regional base map of the designated area can be obtained. Based on each first map tile on the new map, each second map tile on the regional base map is updated to obtain the updated regional base map. Here, the authorized vehicle is the vehicle that initiates the authorization for map building in the designated area, and each first map tile corresponds one-to-one with each second map tile.

[0115] Through the above implementation method, no specialized map data collection vehicles are required. Even ordinary vehicles can acquire the necessary LiDAR point cloud keyframes for a designated area by initiating a mapping authorization, thereby completing map creation. The designated area can be either a high-speed area such as urban roads and highways (high-speed scenario) or a low-speed area such as parking lots, refueling stations, and highway service areas (low-speed scenario). Therefore, by utilizing this implementation method, a large number of ordinary vehicles can efficiently, accurately, and cost-effectively acquire LiDAR point cloud keyframes for various designated areas to complete map creation, overcoming the limitation of existing technologies where conventional maps can only cover high-speed scenarios due to considerations of mapping cost and efficiency.

[0116] Furthermore, the above-described implementation method allows for dynamic and modular updates of the regional base map of a specified area, further improving the map accuracy and update flexibility of the specified area. Attached Figure Description

[0117] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Wherein:

[0118] Figure 1 This is a schematic diagram of the main steps of a point cloud mapping method according to an embodiment of the present invention;

[0119] Figure 2 This is a schematic flowchart of the main steps of a method for creating a new map according to an embodiment of the present invention;

[0120] Figure 3 This is a schematic flowchart of the main steps of a method for aligning a new map with a regional base map according to an embodiment of the present invention.

[0121] Figure 4 This is a schematic diagram of the main structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0122] Some embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0123] In the description of this invention, "processor" can include hardware, software, or a combination of both. A processor can be a central processing unit, microprocessor, image processor, digital signal processor, or any other suitable processor. The processor has data and / or signal processing capabilities. The processor can be implemented in software, in hardware, or a combination of both. Computer-readable storage media includes any suitable medium capable of storing program code, such as magnetic disks, hard disks, optical disks, flash memory, read-only memory, random access memory, etc.

[0124] Here we will first explain some of the terms involved in this invention.

[0125] An authorized vehicle refers to a vehicle that initiates authorization for mapping a designated area. For example, a vehicle that initiates mapping authorization to the server allows the server to obtain LiDAR point cloud and image data collected by the authorized vehicle after it enters the designated area, and to use this data to build a map of that designated area. The designated area can be any designated area in high-speed scenarios such as urban roads and highways (high-speed scenarios), or any designated area in low-speed scenarios such as parking lots, refueling stations, and highway service areas (low-speed scenarios). Specifically, in some implementations, the authorized vehicle can be a vehicle whose owner initiates mapping authorization for the designated area to the server via a terminal device, such as a mobile phone or tablet.

[0126] A keyframe of a LiDAR point cloud refers to a LiDAR point cloud frame in which the change in pose is greater than a preset threshold compared to a previous LiDAR point cloud frame. Those skilled in the art can flexibly set the specific value of the preset threshold according to actual needs; this embodiment of the invention does not impose specific limitations on it.

[0127] The following describes an embodiment of the point cloud mapping method provided by the present invention.

[0128] See appendix Figure 1 , Figure 1 This is a schematic flowchart illustrating the main steps of a point cloud mapping method according to an embodiment of the present invention. Figure 1 As shown, the point cloud mapping method in this embodiment of the invention mainly includes the following steps S101 to S104.

[0129] Step S101: Obtain key frames of the lidar point cloud collected after the authorized vehicle enters the designated area.

[0130] Step S102: Create a new map of the specified area based on the key frames of the lidar point cloud.

[0131] In this embodiment of the invention, the interval for dynamically creating new maps can be preset, and a new map is created every certain interval. Due to factors such as seasons, weather, and construction, regional elements such as roads and buildings within a specified area may change. By dynamically creating new maps of the specified area, the latest status of regional elements within the specified area (including but not limited to the latest location and shape) can be obtained as much as possible. This improves the accuracy of the map after updating the specified area based on these new maps.

[0132] In addition to dynamically building new maps according to pre-set intervals, some implementations can also respond to received mapping instructions by acquiring key frames of lidar point clouds collected by authorized vehicles within a specified time period to build new maps, thereby improving the flexibility of mapping.

[0133] It should be noted that those skilled in the art can flexibly set the specific value of the interval according to actual needs. The embodiments of the present invention do not impose specific limitations on this. For example, the interval can be 1 month.

[0134] Step S103: Obtain the base map of the specified area.

[0135] Specifically, when the new map established in step S102 is the first new map established, the regional base map is the initial map established based on the historical lidar point cloud key frames collected after the authorized vehicle enters the designated area. When the new map established in step S102 is not the first new map established, the regional base map is the updated base map based on the previously established new map.

[0136] In this embodiment of the invention, an initial map can be first established for a specified area, and then updated continuously using each newly established map as the basis for updating. Therefore, for the first newly established map, this initial map is the regional base map, and the base map formed after updating this initial map is the regional base map when updating the map based on the next newly established map. That is, for new maps that are not newly established for the first time, the base map updated based on the previous newly established map is the regional base map.

[0137] Historical lidar point cloud keyframes refer to lidar point cloud keyframes collected when authorized vehicles drive into a designated area before a new map is first created.

[0138] Step S104: Based on each first map tile on the new map, update the map for each second map tile on the regional base map to obtain the updated regional base map.

[0139] This invention first establishes an initial map for a specified area, and then uses this initial map as the basis for continuous updates by creating new maps each time. The coordinate system of this initial map can be a latitude-longitude-height coordinate system, i.e., a three-dimensional coordinate system formed by longitude, latitude, and height. This can be achieved by dividing the initial map into tiles after its creation, forming multiple map tiles. Similarly, the coordinate system of the new map can also be a latitude-longitude-height coordinate system. Before updating the regional base map, the new map can also be divided into tiles using the same method as the initial map, forming multiple map tiles. For simplicity, the map tiles formed by dividing the new map into tiles are referred to as first map tiles, and the map tiles formed by dividing the region into tiles are referred to as second map tiles. Each first map tile corresponds one-to-one with each second map tile.

[0140] Since the first and second map tiles are divided in the same way, it is possible to obtain the first and second map tiles with the same tile location (same coordinates), and then use the map elements in the first map tile to update the map elements in the second map tile. The map elements include at least traffic signs and / or other objects that can serve as markers on the road. Traffic signs include at least lane lines, stop lines, road signs (such as left-turn arrows), traffic lights, and traffic signs. Other objects that can serve as markers include at least pole-shaped objects.

[0141] LiDAR point clouds are data obtained from LiDAR in the LiDAR coordinate system. Since the LiDAR coordinate system is a local coordinate system, a new map built from keyframes of the LiDAR point cloud is actually built with the local coordinate system as the reference coordinate system, thus possessing local consistency. However, due to the random noise in the LiDAR point cloud output, the new map and the regional base map will differ in the global coordinate system, resulting in a lack of global consistency. If this difference is not removed, directly updating the regional base map based on the new map will reduce the map's accuracy. Therefore, in some embodiments of this invention, the new map and the regional base map can be aligned first to achieve global consistency, and then the regional base map can be updated based on the aligned new map.

[0142] Based on the methods described in steps S101 to S104 above, no specialized map data collection vehicles are required. Even ordinary vehicles can acquire the necessary LiDAR point clouds for a designated area by initiating a mapping authorization, thereby completing map creation. In other words, a large number of ordinary vehicles can efficiently, accurately, and cost-effectively acquire LiDAR point clouds for various designated areas to complete map creation. Furthermore, since the new map is dynamically acquired, map updates are performed based on each newly created map, enabling dynamic map updates for designated areas, allowing for periodic map updates. During the base map update process, map tiles are used as the update dimension, and the regional base map is updated modularly, improving the flexibility of map updates.

[0143] The following provides further explanation of steps S102 and S104.

[0144] I. Further explanation of step S102.

[0145] In this embodiment of the invention, after obtaining the keyframes of the LiDAR point cloud, in addition to using conventional point cloud map building methods in the field of map technology to build a map of a specified area based on these LiDAR point cloud keyframes, it is also possible to obtain LiDAR point cloud keyframes in a specified area through multi-vehicle acquisition and / or multiple acquisitions by the same vehicle. Then, using the conventional point cloud map building methods described above, a single map is built based on each LiDAR point cloud keyframe acquired. Finally, these single maps are overlaid, and the overlaid map is used as a new map.

[0146] Specifically, see Appendix Figure 2 In some embodiments of step S102 above, multiple single maps can be created and superimposed to form a new map through the following steps S1021 to S1025.

[0147] Step S1021: Obtain key frames of the lidar point cloud collected after the authorized vehicle has driven into the designated area multiple times.

[0148] Step S1022: Based on the key frames of the lidar point cloud collected each time the authorized vehicle enters the designated area, create a single map for each instance.

[0149] Step S1023: Create a factor graph based on a single map. The factor graph may include factor nodes and relative pose constraints.

[0150] Each factor node corresponds one-to-one with each LiDAR point cloud keyframe. The relative pose constraint is the constraint between two LiDAR point cloud keyframes that can form a loop on different single maps.

[0151] Step S1024: Optimize the pose of each LiDAR point cloud keyframe on each single map based on the factor graph to obtain the optimized pose of each LiDAR point cloud keyframe.

[0152] Step S1025: Based on the optimized pose of each LiDAR point cloud keyframe, overlay each LiDAR point cloud keyframe to form a new map of the specified area.

[0153] The following sections will explain steps S1023 to S1025.

[0154] (a) Explanation of step S1023.

[0155] For simplicity, two LiDAR point cloud keyframes that form a loop on different single-map maps are referred to as loop-loop LiDAR point cloud keyframes. The "constraint term for the relative pose between loop-loop LiDAR point cloud keyframes" can constrain the actual relative pose and the optimal relative pose between these keyframes on different single-map maps, ensuring that the actual relative pose continuously approaches this optimal relative pose. The methods for determining loop-loop LiDAR point cloud keyframes on different single-map maps and for creating the aforementioned constraint term are explained below.

[0156] 1. Method for determining keyframes of loop-closing LiDAR point clouds on different single-map maps

[0157] In this embodiment of the invention, two LiDAR point cloud keyframes capable of forming a loop can be determined on different single maps based on the absolute pose of each keyframe on each single map. Specifically, the difference in absolute pose between two LiDAR point cloud keyframes can be calculated, and it can be determined whether this difference is less than a set threshold. If so, the two LiDAR point cloud keyframes are determined to be capable of forming a loop; otherwise, they are determined not to be capable of forming a loop. Here, absolute pose includes absolute position and absolute attitude, and the difference in absolute pose includes the difference in absolute position and the difference in absolute attitude. Determining whether the difference is less than the set threshold means determining whether the difference in absolute position and absolute attitude is less than their respective set thresholds. The set thresholds for absolute position and absolute attitude can be different. Those skilled in the art can flexibly set the specific values ​​of the above-mentioned set thresholds according to actual needs; this embodiment of the invention does not specifically limit this.

[0158] Since the absolute pose is obtained from the keyframes of the LiDAR point cloud using a GPS device on an authorized vehicle, the positioning signal of the GPS device may be abnormal in scenarios such as underground parking lots. In such cases, the pose of the keyframes of the LiDAR point cloud cannot be accurately obtained, and therefore the loopback LiDAR point cloud keyframes cannot be accurately determined. A GPS device refers to a device that uses satellite navigation and positioning technology for positioning, such as a device based on the Global Navigation Satellite System (GNSS).

[0159] In this case, the keyframes of the loop-closing LiDAR point cloud can be determined using the external images of the vehicle acquired by the image acquisition device on the authorized vehicle. Specifically, in this embodiment of the invention, the following steps 11 to 13 can be used to determine two keyframes of the LiDAR point cloud that can form a loop on different single maps.

[0160] Step 11: Obtain the vehicle exterior image acquired by the same image acquisition device on the authorized vehicle at the same time as the authorized vehicle acquires each lidar point cloud keyframe.

[0161] The same image acquisition device refers to an image acquisition device with the same acquisition range or field of view. Generally speaking, image acquisition devices installed at the same location and at the same installation angle on authorized vehicles have the same acquisition range.

[0162] Step 12: Perform similarity analysis on the vehicle exterior images corresponding to each keyframe of the LiDAR point cloud on different single maps to obtain at least one similar image group, wherein the similar image group includes two similar vehicle exterior images that correspond to the keyframes of the LiDAR point cloud on different single maps respectively.

[0163] In this embodiment of the invention, image features of vehicle exterior images can be extracted. The similarity between two vehicle exterior images is obtained based on their image features, and then compared with a preset similarity threshold. If the similarity is greater than the preset threshold, it indicates that the two vehicle exterior images are highly similar and are matched or similar. If the similarity is less than or equal to the preset threshold, it indicates that the two vehicle exterior images are relatively similar and are not matched or dissimilar. After similarity analysis using the above method, two similar vehicle exterior images are grouped into a similar image group. It should be noted that those skilled in the art can use conventional image similarity acquisition methods in the field of image processing technology to obtain the similarity between two vehicle exterior images, and this embodiment of the invention does not specifically limit this. For example, in some embodiments, the Euclidean distance between the image features of two vehicle exterior images can be calculated, and then the reciprocal of the Euclidean distance can be calculated and used as the similarity.

[0164] Step 13: For each similar image group, take the key frames of the LiDAR point cloud corresponding to the two vehicle exterior images in the similar image group as the two key frames of the LiDAR point cloud that can form a loop on different single maps.

[0165] When two vehicle exterior images are highly similar, it indicates that the two vehicle exterior images were captured by the image acquisition device when the vehicles were driving to the same position. Therefore, the key frames of the lidar point cloud acquired at the same time as the two vehicle exterior images can form a loop.

[0166] Based on the method described in steps 11 to 13 above, key frames of loop-loop lidar point clouds on different single maps can be accurately determined using images collected by the vehicle when the positioning signal of the global satellite positioning device is abnormal.

[0167] 2. Method for establishing constraints on relative pose between keyframes of loop-loop lidar point clouds on different single maps

[0168] As described above, the main function of the "constraint term for relative pose between loop-loop LiDAR point cloud keyframes" is to constrain the actual relative pose and optimal relative pose between two LiDAR point cloud keyframes that can form a loop on different single maps, so that the actual relative pose continuously approaches this optimal relative pose. Therefore, the accuracy of the optimal relative pose will greatly affect the constraint capability of the constraint term. In order to maximize the constraint capability of the above constraint term, in this embodiment of the invention, the optimal relative pose can be obtained through the following steps 21 to 25, and then the constraint term for relative pose between loop-loop LiDAR point cloud keyframes on different single maps can be established based on the optimal relative pose.

[0169] Step 21: Obtain two keyframes of LiDAR point clouds that can form a loop on different single maps.

[0170] Step 22: From the single map containing the first LiDAR point cloud keyframe in the two LiDAR point cloud keyframes, obtain the first local map including the first LiDAR point cloud keyframe.

[0171] In some preferred embodiments, multiple lidar point cloud keyframes located before and after the first lidar point cloud keyframe can be selected from a single map, and then the multiple lidar point cloud keyframes before and after the first lidar point cloud keyframe are superimposed to form a first local map.

[0172] Step 23: From the single map containing the second LiDAR point cloud keyframe, obtain the second local map.

[0173] In some preferred embodiments, multiple lidar point cloud keyframes located before and after the second lidar point cloud keyframe can be selected from a single map, and then the multiple lidar point cloud keyframes before and after the second lidar point cloud keyframe are superimposed to form a second local map.

[0174] Step 24: Perform point cloud matching on the first local map and the second local map to determine the matching relative pose between the first local map and the second local map, and use the matching relative pose as the optimal relative pose between the two lidar point cloud keyframes.

[0175] Point cloud matching can be used to obtain the relative pose between the first local map and the second local map. Based on the relative pose between the two maps, the relative pose between any two LiDAR point cloud keyframes in the two maps can be obtained.

[0176] In this embodiment of the invention, conventional point cloud matching methods in the field of autonomous driving technology can be used to perform point cloud matching between the first local map and the second local map. For example, a point cloud matching method based on the ICP (Iterative Closest Point) algorithm can be used to perform point cloud matching on the local map. The point cloud matching method based on the ICP algorithm can be a point cloud matching method based on the plane-to-plane method in the ICP algorithm. This embodiment of the invention does not specifically limit the method for performing point cloud matching on the local map.

[0177] Step 25: Based on the optimal relative pose, establish constraints for the relative pose between two LiDAR point cloud keyframes. Specifically, the matched relative pose is taken as the optimal relative pose between two LiDAR point cloud keyframes that can form a loop, and constraints are established based on this optimal relative pose.

[0178] Based on the methods described in steps 21 to 25 above, a more accurate optimal relative pose can be obtained, thereby improving the constraint capability of the constraint terms established based on this optimal relative pose.

[0179] Furthermore, in some preferred embodiments of step S1023 above, before executing step 24, a step (step 26) may be included to obtain the initial relative pose between two lidar point cloud keyframes that can form a loop. After obtaining the initial relative pose through this step, the initial relative pose is used as the initial value of the relative pose between the first local map and the second local map (i.e., the relative pose between the first local map and the second local map is initialized). Then, when executing step 25, point cloud matching is performed on the first local map and the second local map according to this initial value to determine the matching relative pose between the first local map and the second local map.

[0180] Specifically, the step of obtaining the initial relative pose (step 26) may include the following steps 261 to 263.

[0181] Step 261: Obtain two keyframes of LiDAR point clouds that can form a loop on different single maps.

[0182] Step 262: Obtain two external images of the vehicle acquired by the same image acquisition device on the authorized vehicle at the same time as the acquisition of the first and second lidar point cloud keyframes.

[0183] Step 263: Perform image matching on the two vehicle exterior images, and determine the initial relative pose between the first and second lidar point cloud keyframes based on the image matching results.

[0184] Specifically, by matching two vehicle exterior images, the relative pose between the two vehicle exterior images can be obtained. This relative pose is then transformed from the image coordinate system to the LiDAR coordinate system, and finally, the relative pose transformed to the LiDAR coordinate system is used as the initial relative pose. It should be noted that in this embodiment of the invention, conventional image matching methods in the field of image processing technology can be used to match the two vehicle exterior images to obtain the relative pose between them. This embodiment of the invention does not specifically limit the above image matching method.

[0185] Based on the method described in steps 261 to 263 above, the initial relative pose between key frames of the loop-loop lidar point cloud can be obtained more accurately using the vehicle's external image. This initial relative pose can be used as the initial value of the relative pose between the first local map and the second local map, which can overcome the problem that the relative pose between the first local map and the second local map cannot be accurately obtained through point cloud matching due to the inaccuracy of the initial pose.

[0186] (ii) Explanation of step S1024.

[0187] In the process of optimizing the pose of each LiDAR point cloud keyframe on each single map using factor graphs, the "constraint term of relative pose between loop-loop LiDAR point cloud keyframes on different single maps" can accurately and efficiently eliminate the pose deviation of LiDAR point cloud keyframes at the same location on different single maps, reduce the difference at the same location on different single maps, realize the coordinate system alignment of each LiDAR point cloud keyframe on different single maps, obtain a more accurate pose, and thus quickly and accurately build a point cloud map.

[0188] In this embodiment of the invention, the conventional method of optimizing the pose of LiDAR point clouds based on factor graphs in the field of autonomous driving technology can be adopted. The pose of each key frame of LiDAR point cloud on each single map is optimized by factor graphs. This embodiment of the invention will not elaborate on the specific method of optimization by factor graphs.

[0189] (III) Explanation of step S1025.

[0190] In this embodiment of the invention, each keyframe of the LiDAR point cloud can be projected onto the same coordinate system and superimposed based on its optimized pose to form a new map corresponding to a specified area. For example, if the optimized pose is an absolute pose, then each keyframe of the LiDAR point cloud can be projected onto an absolute coordinate system and superimposed to form a new map corresponding to the specified area.

[0191] The above is an explanation of steps S1023 to S1025.

[0192] Based on the method described in steps S1021 to S1025 above, a new map of a specified area can be established by using key frames of LiDAR point clouds in a specified area obtained through multi-vehicle acquisition and / or multiple acquisitions by the same vehicle. Compared with establishing a new map of a specified area using key frames of LiDAR point clouds acquired by a single vehicle or a single acquisition by the same vehicle, the accuracy of the new map can be significantly improved.

[0193] Furthermore, it should be noted that the method described in steps S1021 to S1025 above can also be used to establish an initial map of a specified area in the embodiments of the present invention. The only difference is that the key frame of the lidar point cloud obtained in step S1021 is not a new lidar point cloud key frame, but a historical lidar point cloud key frame.

[0194] The above is a further explanation of step S102.

[0195] II. Further explanation of step S104.

[0196] As can be seen from the aforementioned method embodiments, in step S104, the new map and the regional base map can be aligned first, and then the second map tiles of the regional base map are updated based on the first map tiles of the aligned new map to obtain the updated base map. It is evident that the accuracy of map alignment greatly affects the accuracy of map updating. For further details, please refer to the appendix. Figure 3 In some embodiments of step S104 above, each keyframe of the lidar point cloud of the new map can be aligned with the regional base map through the following steps S1041 to S1044, so as to achieve alignment of the entire new map with the regional base map.

[0197] Step S1041: Based on the pose of the current lidar point cloud keyframe on the new map, search for the nearest neighbor lidar point cloud keyframe from the lidar point cloud keyframes on the regional base map.

[0198] Specifically, based on the pose of the lidar point cloud keyframes, the distance between the current lidar point cloud keyframe and each lidar point cloud keyframe in the region base map can be calculated. Lidar point cloud keyframes with a distance less than a preset distance threshold are selected from the region base map as the nearest neighbor lidar point cloud keyframes. Those skilled in the art can flexibly set the value of the preset distance threshold according to actual needs; for example, the preset distance threshold could be 5 meters. This embodiment of the invention does not specifically limit this value.

[0199] Step S1042: Obtain a local map from the regional base map, including the keyframes of the nearest neighbor lidar point cloud.

[0200] In some preferred embodiments, multiple lidar point cloud keyframes located before and after the nearest neighbor lidar point cloud keyframe can be selected from the regional base map, and then the multiple lidar point cloud keyframes before and after the nearest neighbor lidar point cloud keyframe are superimposed to form a local map.

[0201] Step S1043: Perform point cloud matching between the current LiDAR point cloud keyframe and the local map to determine the relative pose between the current LiDAR point cloud keyframe and the local map.

[0202] In this embodiment of the invention, conventional point cloud matching methods in the field of autonomous driving technology can be used to perform point cloud matching between the current LiDAR point cloud keyframe and the local map. For example, a point cloud matching method based on the ICP algorithm can be used to perform point cloud matching on the local map. The point cloud matching method based on the ICP algorithm can be a point cloud matching method based on the point-to-plane method in the ICP algorithm. This embodiment of the invention does not specifically limit the point cloud matching method described above.

[0203] Step S1044: Perform pose transformation on the current lidar point cloud keyframe according to the relative pose, so that the current lidar point cloud keyframe is aligned with the region base map.

[0204] Since the new map and the regional base map do not have global consistency, the relative pose of each LiDAR point cloud keyframe on the new map may be different from that of its corresponding local map. After obtaining the accurate relative pose, the pose transformation of each LiDAR point cloud keyframe on the new map can be performed using this relative pose to eliminate the difference between the new map and the regional base map in the global coordinate system, so that the new map and the regional base map have global consistency.

[0205] In addition, to further improve the accuracy of map updates, in some other embodiments of step S104 above, map updates can be performed using a map difference-based update method or a manual update method.

[0206] 1. Map difference-based update method

[0207] The map difference-based update method mainly includes: obtaining a first map tile and a second map tile with the same tile location (i.e., a first map tile and a second map tile with the same coordinates), then performing map difference on the first map tile and the second map tile with the same tile location, and updating the second map tile based on the map difference result. Map difference mainly involves matching map elements on the first map tile with map elements on the second map tile, and determining the differences between the first map tile and the second map tile based on the matching result, i.e., obtaining the map difference result. After determining the differences, the second map tile can be updated based on the differences, such as adding new map elements or deleting existing map elements on the second map tile. It should be noted that in the embodiments of the present invention, conventional map difference methods in the field of map technology can be used to perform map difference on the first map tile and the second map tile, and then update the second map tile based on the map difference result. The embodiments of the present invention do not specifically limit the above map difference method.

[0208] Map updates can be performed automatically and efficiently without manual intervention using the map difference-based update method described above.

[0209] 2. Manual update method

[0210] The manual update method mainly includes: outputting the first map tile of the aligned new map and the second map tile of the regional base map; responding to the received manual update instruction for the second map tile; and updating the second map tile according to the manual update instruction. The manual update instruction may contain content to update the second map tile (such as adding or deleting map elements), and the second map tile is updated according to the above update content.

[0211] Updating maps manually allows for the updating of the base map of a specified area at any time based on update needs, thus improving the flexibility of map updates.

[0212] It should be noted that although the steps in the above embodiments are described in a specific order, those skilled in the art will understand that in order to achieve the effects of the present invention, different steps do not necessarily have to be executed in such an order. They can be executed simultaneously (in parallel) or in other orders, and these variations are all within the scope of protection of the present invention.

[0213] Those skilled in the art will understand that all or part of the processes in the method of the above embodiment of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable storage medium can include any entity or device capable of carrying the computer program code, a medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content included in the computer-readable storage medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0214] Furthermore, the present invention also provides a computer device.

[0215] See appendix Figure 4 , Figure 4 This is a schematic diagram of the main structure of a computer device according to an embodiment of the present invention. Figure 4 As shown, the computer device in this embodiment of the invention mainly includes a storage device and a processor. The storage device can be configured to store a program for executing the point cloud mapping method of the above-described method embodiments, and the processor can be configured to execute the program in the storage device. This program includes, but is not limited to, a program for executing the point cloud mapping method of the above-described method embodiments. For ease of explanation, only the parts related to the embodiments of the present invention are shown. For specific technical details not disclosed, please refer to the method section of the embodiments of the present invention.

[0216] In embodiments of the present invention, the computer device may be a control device comprising various electronic devices. In some possible implementations, the computer device may include multiple storage devices and multiple processors. The program executing the point cloud mapping method of the above method embodiments can be divided into multiple subroutines, each subroutine can be loaded and run by a processor to execute different steps of the point cloud mapping method of the above method embodiments. Specifically, each subroutine can be stored in different storage devices, and each processor can be configured to execute programs in one or more storage devices to jointly implement the point cloud mapping method of the above method embodiments, that is, each processor executes different steps of the point cloud mapping method of the above method embodiments to jointly implement the point cloud mapping method of the above method embodiments.

[0217] The aforementioned multiple processors can be processors deployed on the same device. For example, the aforementioned computer device can be a high-performance device composed of multiple processors, and the aforementioned multiple processors can be processors configured on that high-performance device. Alternatively, the aforementioned multiple processors can also be processors deployed on different devices. For example, the aforementioned computer device can be a server cluster, and the aforementioned multiple processors can be processors on different servers within the server cluster.

[0218] Furthermore, the present invention also provides a computer-readable storage medium.

[0219] In one embodiment of the computer-readable storage medium according to the present invention, the computer-readable storage medium may be configured to store a program for performing the point cloud mapping method of the above-described method embodiments. This program may be loaded and run by a processor to implement the point cloud mapping method. For ease of explanation, only the parts related to the embodiments of the present invention are shown; for specific technical details not disclosed, please refer to the method section of the embodiments of the present invention. The computer-readable storage medium may be a storage device comprising various electronic devices. Optionally, in the embodiments of the present invention, the computer-readable storage medium is a non-transitory computer-readable storage medium.

[0220] The technical solution of the present invention has been described above with reference to one embodiment shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions resulting from such changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A point cloud mapping method, characterized in that, The method includes: The key frame of the lidar point cloud is acquired after the authorized vehicle enters the designated area, and the pose change between the key frame and the previous lidar point cloud frame is greater than a preset threshold. A new map of the specified area is created based on the keyframes of the lidar point cloud. Obtain the base map of the specified area; Based on each first map tile on the new map, update the map for each second map tile on the regional base map to obtain the updated regional base map; The authorized vehicle is the vehicle that initiates the authorization for mapping the designated area, and each first map tile corresponds one-to-one with each second map tile; Creating a new map of the specified area includes: Based on the keyframes of the lidar point cloud collected each time the authorized vehicle enters the designated area, a single map is created. A factor graph is created based on the single map. The factor graph includes factor nodes and relative pose constraints. Each factor node corresponds one-to-one with each key frame of the LiDAR point cloud. The relative pose constraints are the constraints on the relative pose between two key frames of the LiDAR point cloud that can form a loop on different single maps. Based on the factor map, the pose of each key frame of the LiDAR point cloud on each single map is optimized to obtain the optimized pose of each key frame of the LiDAR point cloud. Based on the optimized pose of each LiDAR point cloud keyframe, the keyframes of each LiDAR point cloud are superimposed to form a new map of the specified area.

2. The point cloud mapping method according to claim 1, characterized in that, Before the step of "updating the map for each second map tile on the regional base map based on each first map tile on the new map to obtain the updated regional base map", the method further includes: Align the new map with the base map of the region.

3. The point cloud mapping method according to claim 2, characterized in that, The step of "aligning the new map with the regional base map" specifically includes aligning each keyframe of the LiDAR point cloud on the new map with the regional base map using the following methods: Based on the pose of the current lidar point cloud keyframe on the new map, search for the nearest neighbor lidar point cloud keyframe from the lidar point cloud keyframes on the regional base map. Obtain a local map from the base map of the region, including the keyframes of the nearest neighbor lidar point cloud; Point cloud matching is performed between the current lidar point cloud keyframe and the local map to determine the relative pose between the current lidar point cloud keyframe and the local map. The pose transformation of the current lidar point cloud keyframe is performed based on the relative pose so that the current lidar point cloud keyframe is aligned with the region base map.

4. The point cloud mapping method according to claim 3, characterized in that, The step of "obtaining a local map from the base map of the region, including the keyframes of the nearest neighbor lidar point cloud" specifically includes: Select multiple lidar point cloud keyframes from the base map of the region that are located before and after the nearest neighbor lidar point cloud keyframe. The local map is formed by overlaying multiple key frames of the lidar point cloud before and after the aforementioned key frames, as well as the key frames of the nearest-neighbor lidar point cloud.

5. The point cloud mapping method according to claim 1, characterized in that, The step of "updating the map for each second map tile on the regional base map based on each first map tile on the new map, to obtain the updated regional base map" specifically includes: Obtain the first map tile and the second map tile with the same tile position; Map difference is performed on the first map tile and the second map tile at the same tile position, and the second map tile is updated based on the result of the map difference.

6. The point cloud mapping method according to claim 1, characterized in that, The step of "updating the map for each second map tile on the regional base map based on each first map tile on the new map, to obtain the updated regional base map" further includes: Output the first map tile of the new map and the second map tile of the region base map; In response to a received manual update instruction for the second map tile, the second map tile is updated according to the manual update instruction.

7. The point cloud mapping method according to claim 1, characterized in that, The method further includes creating the relative pose constraint term in the following manner: Acquire two keyframes of LiDAR point clouds that can form a loop on different single maps; From the single map containing the first LiDAR point cloud keyframe in the two LiDAR point cloud keyframes, obtain a first local map including the first LiDAR point cloud keyframe. From the single map containing the second LiDAR point cloud keyframe in the two LiDAR point cloud keyframes, obtain a second local map including the second LiDAR point cloud keyframe. Point cloud matching is performed on the first local map and the second local map to determine the matching relative pose between the first local map and the second local map, and the matching relative pose is used as the optimal relative pose between the two lidar point cloud keyframes. Based on the optimal relative pose, constraints are established for the relative pose between the two lidar point cloud keyframes.

8. The point cloud mapping method according to claim 7, characterized in that, Before the step of "performing point cloud matching between the first local map and the second local map to determine the relative pose of the matching between the first local map and the second local map", the method further includes: The authorized vehicle acquires two external images captured by the same image acquisition device on the authorized vehicle at the same time as the first and second lidar point cloud keyframes are acquired. Image matching is performed on the two vehicle exterior images, and the initial relative pose between the first lidar point cloud keyframe and the second lidar point cloud keyframe is determined based on the image matching result. The step of "performing point cloud matching between the first local map and the second local map to determine the relative pose of the matching between the first local map and the second local map" specifically includes: The initial pose is used as the initial value of the relative pose between the first local map and the second local map. Point cloud matching is performed on the first local map and the second local map based on the initial value to determine the matching relative pose between the first local map and the second local map.

9. The point cloud mapping method according to claim 7, characterized in that, The step of "obtaining a first local map including the first LiDAR point cloud keyframe from the single map containing the first LiDAR point cloud keyframe" specifically includes: Select multiple lidar point cloud keyframes from the single map, located before and after the first lidar point cloud keyframe. The first local map is formed by overlaying multiple keyframes of the lidar point cloud before and after the first lidar point cloud keyframe.

10. The point cloud mapping method according to claim 7, characterized in that, The step of "obtaining a second local map including the second LiDAR point cloud keyframe from the single map containing the second LiDAR point cloud keyframe" specifically includes: Select multiple lidar point cloud keyframes from the single map, located before and after the second lidar point cloud keyframe. The second local map is formed by overlaying multiple key frames of the LiDAR point cloud before and after the first LiDAR point cloud and the second LiDAR point cloud key frame.

11. The point cloud mapping method according to claim 7, characterized in that, The steps for "obtaining two keyframes of LiDAR point clouds that can form a loop on different single maps" specifically include: Based on the absolute pose of each LiDAR point cloud keyframe on each single map, determine two LiDAR point cloud keyframes on different single maps that can form a loop.

12. The point cloud mapping method according to claim 7, characterized in that, The steps of "obtaining two keyframes of LiDAR point clouds that can form a loop on different single maps" also include: The external image of the vehicle is acquired by the same image acquisition device on the authorized vehicle while the authorized vehicle is acquiring each key frame of the lidar point cloud; Similarity analysis is performed on the vehicle exterior images corresponding to each keyframe of the LiDAR point cloud on different single maps to obtain at least one similar image group, wherein the similar image group includes two similar vehicle exterior images that correspond to the keyframes of the LiDAR point cloud on different single maps respectively. For each group of similar images, the key frames of the LiDAR point cloud corresponding to the two vehicle exterior images in the group of similar images are taken as the two key frames of the LiDAR point cloud that can form a loop on different single maps.

13. A computer device comprising a processor and a storage device, said storage device being adapted to store a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by the processor to perform the point cloud mapping method according to any one of claims 1 to 12.

14. A computer-readable storage medium storing a plurality of program codes, characterized in that, The program code is adapted to be loaded and run by a processor to perform the point cloud mapping method according to any one of claims 1 to 12.

Images