A map construction method and device, electronic equipment and storage medium

Abstract

A map construction method and device, electronic equipment and storage medium are disclosed. The method comprises: obtaining point cloud data and inertial measurement data, and fusing the point cloud data and the inertial measurement data into mapping data; determining a current reference map block based on a sub-mapping strategy of a target road section, and extracting, from the mapping data, to-be-matched mapping data corresponding to a to-be-registered map block according to the current reference map block; obtaining a coordinate mapping relationship of the to-be-matched mapping data to the current reference map block; mapping the to-be-matched mapping data to standard mapping data according to the coordinate mapping relationship, and splicing the to-be-registered map block to a map of the target road section based on the standard mapping data. The technical scheme of the embodiment of the present application can solve the problem of the decline in the precision and quality of the point cloud map constructed by the SLAM technology in a large-scale scene and the resulting map ghosting phenomenon by splicing the to-be-registered map block to the map of the target road section based on the standard mapping data.

Description

Technical Field

[0001] This invention relates to the field of image processing, and more particularly to a map construction method, apparatus, electronic device, and storage medium. Background Technology

[0002] In the field of point cloud map construction based on 3D LiDAR Simultaneous Localization and Mapping (SLAM), to ensure the accuracy and consistency of maps in large-scale scenes, the industry typically uses LiDAR inertial odometry for continuous pose estimation and map construction. However, due to factors such as sensor extrinsic parameter calibration errors and measurement noise, the front-end odometry inevitably has accumulated errors. This will cause the point cloud map construction trajectory based on SLAM technology to continuously shift in large-scale scenes, and will also cause additional distortion in the distortion correction process of a single frame of LiDAR point cloud, thus affecting the final map construction effect and accuracy. In addition, when the number of single-frame LiDAR point clouds input to the SLAM system is too large, it will lead to an increase in point cloud matching time.

[0003] To reduce data processing volume, existing technologies typically employ feature extraction or storage sequence interval downsampling. However, these methods are prone to losing crucial environmental information in feature-sparse scenarios, and may even lead to SLAM system failure, thus affecting the stability of the SLAM system. Therefore, how to effectively suppress accumulated errors and improve the accuracy and efficiency of large-scale point cloud map construction without sacrificing information integrity and system stability has become a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0004] This invention provides a map building method, apparatus, electronic device, and storage medium to solve the problem that the accuracy and quality of point cloud maps built by SLAM technology in large-scale scenes are reduced, and map ghosting phenomenon is caused.

[0005] According to one aspect of the present invention, a map construction method is provided, the method comprising: Acquire point cloud data and inertial measurement data, and fuse the point cloud data and inertial measurement data into mapping data; The current reference map block is determined based on the map segmentation strategy of the target road segment, and the map data to be matched for the corresponding map block to be registered is extracted from the map data based on the current reference map block. The current reference map block and the map block to be registered have some overlap. Obtain the coordinate mapping relationship between the mapping data to be matched and the current base map block; The mapping data to be matched is mapped to standard mapping data according to the coordinate mapping relationship, and the map to be registered is pieced together into the map of the target road segment based on the standard mapping data.

[0006] According to another aspect of the present invention, a map building apparatus is provided, the apparatus comprising: The data fusion module is used to acquire point cloud data and inertial measurement data, and fuse the point cloud data and the inertial measurement data into mapping data; The candidate determination module is used to determine the current reference map block based on the map segmentation strategy of the target road segment, and extract the map data to be matched for the corresponding map block to be registered in the map data according to the current reference map block, wherein the current reference map block and the map block to be registered have partial overlap. The synchronization relationship module is used to obtain the coordinate mapping relationship between the mapping data to be matched and the current base map block; The map stitching module is used to map the mapping data to be matched into standard mapping data according to the coordinate mapping relationship, and stitch the map to be registered into the target road segment map based on the standard mapping data.

[0007] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the map construction method according to any embodiment of the present invention.

[0008] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute a map construction method according to any embodiment of the present invention.

[0009] The technical solution of this invention involves acquiring point cloud data and inertial measurement data, fusing them into mapping data, determining the current reference map block based on the mapping strategy of the target road segment, extracting the mapping data to be matched for the corresponding map block to be registered within the mapping data based on the current reference map block, obtaining the coordinate mapping relationship between the mapping data to be matched and the current reference map block, mapping the mapping data to be matched into standard mapping data based on the coordinate mapping relationship, and stitching the map block to be registered onto the map of the target road segment based on the standard mapping data. The above technical solution determines the current reference map block by using a segmentation strategy based on the target road segment, and extracts the matching mapping data of the corresponding map block to be registered from the mapping data based on the current reference map block. This can segment large-scale scenes, achieve data dimensionality reduction and localized registration, and map the matching mapping data to be registered into standard mapping data through coordinate mapping relationship. Based on the standard mapping data, the map block to be registered is stitched to the map of the target road segment, which can synthesize a high-precision point cloud map and improve the quality of the point cloud map.

[0010] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart of a map construction method provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of a map construction method provided in Embodiment 2 of the present invention; Figure 3 This is a flowchart illustrating the overall principle of a high-precision point cloud map construction system according to Embodiment 3 of the present invention. Figure 4 This is a schematic diagram of a road segmentation principle provided in Embodiment 3 of the present invention; Figure 5 This is a schematic diagram of the principle process of block map registration according to Embodiment 3 of the present invention; Figure 6 This is a schematic diagram of the candidate frame selection principle for loop closure detection provided in Embodiment 3 of the present invention; Figure 7This is a schematic diagram of a portable and detachable data acquisition device for a point cloud map construction system according to Embodiment 3 of the present invention; Figure 8 This is a schematic diagram of a map building device according to Embodiment 4 of the present invention; Figure 9 This is a schematic diagram of the structure of an electronic device that implements the map construction method of Embodiment 5 of the present invention. Detailed Implementation

[0013] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0014] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0015] Example 1 Figure 1 This invention provides a flowchart of a map building method according to Embodiment 1. This embodiment is applicable to high-precision point cloud map construction. The method can be executed by a map building device, which can be implemented in hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes: S110. Acquire point cloud data and inertial measurement data, and fuse the point cloud data and inertial measurement data into mapping data.

[0016] The point cloud data can include raw sensor data containing three-dimensional spatial information of the environment, collected by LiDAR, which can include mechanical LiDAR, semi-solid-state LiDAR, etc. The inertial measurement data can include raw sensor data describing the motion state, collected by an inertial measurement unit. The mapping data can include data units that fuse point cloud data and inertial measurement data through timestamp alignment and spatial coordinate alignment.

[0017] Specifically, the front-end odometer unit can acquire point cloud data collected by the lidar and inertial measurement data collected by the inertial measurement unit through the data preprocessing control center, and then fuse the point cloud data and inertial measurement data into mapping data by aligning the timestamps and spatial coordinates.

[0018] S120. Determine the current reference map block based on the target road segment sub-map strategy, and extract the matching mapping data of the corresponding map block to be registered in the mapping data according to the current reference map block.

[0019] The target road segment can include the complete road segment for which a point cloud map needs to be built, and the target road segment can be composed of multiple continuous road segments stitched together. The map segmentation strategy can be a strategy that divides the target road segment for which a point cloud map needs to be built into several units according to the controllable range of cumulative error of the front-end odometer, based on a time series and a single road segment mileage threshold. The map segmentation strategy can include sequentially dividing according to the time series, setting fixed or variable overlapping areas between adjacent blocks for map segmentation constraints; setting a single road segment mileage threshold to divide the map into blocks, ensuring that the cumulative error within each block does not exceed a preset error upper limit, etc. The reference map block can include map blocks that serve as a registration reference. The map block to be registered can include map blocks that need to be transformed into the coordinate system of the reference map block, and the reference map block and the map block to be registered have partial overlap. The mapping data to be matched can include data units extracted from the mapping data used for registration operations with the current reference map block.

[0020] Specifically, a map segmentation strategy is obtained for each target road segment. This strategy divides the target road segment into several units, which may or may not overlap. Based on the controllable range of the cumulative error of the front-end odometer, the target road segment for which a point cloud map needs to be constructed can be divided into several units according to the map segmentation strategy. Then, the map segment with the completed map segmentation strategy is determined as the current reference map block, and the next adjacent map segment is determined as the map block to be registered. The map data to be matched for the map block to be registered is extracted from the mapping data.

[0021] S130. Obtain the coordinate mapping relationship of the mapping data to be matched and synchronized to the current base map block.

[0022] The coordinate mapping relationship uniquely represents a position vector in one coordinate system as a position vector in another coordinate system. The coordinate mapping relationship can include spatial transformation parameters that transform the mapping data to be matched to the coordinate system of the base map tiles. These spatial transformation parameters can include rotation matrices, translation vectors, etc.

[0023] Specifically, the partial overlap between the base map block and the map block to be registered can be solved to obtain spatial transformation parameters for registering the mapping data to be matched to the coordinate system of the current base map block, so as to establish the coordinate mapping relationship between the mapping data to be matched and the current base map block. The solution method can include iterative nearest point algorithm and point cloud registration algorithm.

[0024] S140. Map the mapping data to be matched to standard mapping data according to the coordinate mapping relationship, and stitch the map to be registered into the target road segment map based on the standard mapping data.

[0025] Standard mapping data refers to a standardized set of input data used to construct environmental maps. Standard mapping data can include fused data that, after being transformed by coordinate mapping relationships, is in the same coordinate system as the current reference map tiles. The map can include a complete point cloud map corresponding to the target road segment, and the map can be synthesized from multiple registered and stitched map tiles.

[0026] Specifically, based on the coordinate mapping relationship, the mapping data to be matched can be transformed to obtain standard mapping data aligned with the coordinate system of the current reference map block. The standard mapping data is then fused with the current reference map block, and the fused map block to be registered is stitched to the corresponding position in the target road segment map.

[0027] The technical solution of this invention involves acquiring point cloud data and inertial measurement data, fusing them into mapping data, determining the current reference map block based on the mapping strategy of the target road segment, extracting the mapping data to be matched for the corresponding map block to be registered within the mapping data based on the current reference map block, obtaining the coordinate mapping relationship between the mapping data to be matched and the current reference map block, mapping the mapping data to be matched into standard mapping data based on the coordinate mapping relationship, and stitching the map block to be registered onto the map of the target road segment based on the standard mapping data. The above technical solution determines the current reference map block by using a segmentation strategy based on the target road segment, and extracts the matching mapping data of the corresponding map block to be registered from the mapping data based on the current reference map block. This can segment large-scale scenes, achieve data dimensionality reduction and localized registration, and map the matching mapping data to be registered into standard mapping data through coordinate mapping relationship. Based on the standard mapping data, the map block to be registered is stitched to the map of the target road segment, which can synthesize a high-precision point cloud map and improve the quality of the point cloud map.

[0028] Example 2 Figure 2 This is a flowchart of a map construction method provided in Embodiment 2 of the present invention. This embodiment further refines the above embodiment: like Figure 2 As shown, the method includes: S210. Acquire point cloud data and inertial measurement data, and fuse the point cloud data and inertial measurement data into mapping data.

[0029] The point cloud data can include raw sensor data containing three-dimensional spatial information of the environment, collected by LiDAR (LiDAR System). LiDAR can include mechanical LiDAR, semi-solid-state LiDAR, etc. The inertial measurement data can include raw sensor data describing the motion state, collected by an inertial measurement unit (IMU). The mapping data can be the basic data units used to construct an environmental map. Mapping data can include data units that fuse point cloud data and inertial measurement data through timestamp alignment, spatial coordinate alignment, etc.

[0030] Specifically, point cloud data collected by lidar and inertial measurement data collected by inertial measurement unit can be obtained through the data preprocessing control center, and the point cloud data and inertial measurement data can be fused into mapping data by aligning timestamps and spatial coordinates.

[0031] S220. Divide the target road segment into at least one map segment according to the preset single road segment mileage threshold and the repeated road segment mileage threshold.

[0032] The map segmentation can include several continuous sub-segments obtained by dividing the target road segment according to time series and mileage thresholds. The mileage threshold for a single road segment can include the maximum mileage length of a single map segment set according to the controllable range of cumulative error of the front-end odometer unit. The mileage threshold for repeated road segments can include the maximum mileage length of the overlapping portion between two adjacent map segments.

[0033] Specifically, the overlapping area between two adjacent segments can be determined based on the mileage threshold of repeated road segments. The target road segment can be divided into several segments sequentially based on the mileage threshold of a single road segment. Each pair of adjacent segments can contain overlapping areas. Each segment of the target road segment generated by the division can be used as a separate map segment.

[0034] S230. Determine each map segment as the current base map block in the order of the target road segment.

[0035] The current reference map block can include the map block currently used as the registration reference.

[0036] Specifically, the first map segment of the target road segment can be set as the initial current reference map block, and the next map segment can be determined as the new current reference map block according to the direction of travel of the target road segment, until all map segments are determined as current reference map blocks. The determination method can include determining each map segment as the current reference map block at the same time, or determining each map segment as the current reference map block in sequence according to the processing order, etc.

[0037] S240. Extract the endpoint location timestamp of the mapping data of the current base map block within the map of the target road segment.

[0038] The timestamp of the endpoint location of the mapping data can include the timestamp corresponding to the last frame of standard mapping data in the current base map tile.

[0039] Specifically, based on the identifier information of the current base map block, the mapping data sequence corresponding to the current base map block can be extracted from the map of the target road segment. The timestamp corresponding to the last frame of standard mapping data in the mapping data sequence can be read and determined as the timestamp of the end point of the mapping data. The identifier information may include the block number and block name of the current base map block.

[0040] S250. Starting from the timestamp of the endpoint location, extract the mapping data of the mileage length of the corresponding map block to be registered as the mapping data to be matched.

[0041] The mileage length can include the path length covered by the map tile to be registered. The mileage length can be pre-configured for extracting different map tiles to be registered.

[0042] Specifically, starting from the timestamp of the endpoint of the mapping data, a portion of the mapping data can be extracted from the mapping data according to the length of the mileage to be matched as the mapping data to be matched. The extraction methods can include time interval truncation and mileage-triggered extraction.

[0043] For example, the tile map registration unit uses the timestamp of the endpoint of the mapping data as the starting point and the corresponding LiDAR data as the starting data for the map tile to be registered. The front-end odometer starts performing SLAM calculations from this starting data and accumulates the travel mileage in real time. When the accumulated mileage reaches the preset threshold for a single road segment, the current round of data collection is terminated, and the mapping data collected within that mileage range is used as the mapping data to be matched.

[0044] S260. Obtain the coordinate mapping relationship of the mapping data to be matched and synchronized to the current base map block.

[0045] The coordinate mapping relationship can include spatial transformation parameters that transform the mapping data to be matched to the coordinate system of the base map block. The spatial transformation parameters can include rotation matrix and translation vector.

[0046] Specifically, the partial overlap between the base map block and the map block to be registered can be solved to obtain spatial transformation parameters for registering the mapping data to be matched to the coordinate system of the current base map block. These spatial transformation parameters can be used to establish the coordinate mapping relationship between the mapping data to be matched and the current base map block. The solution method for the partial overlap between the base map block and the map block to be registered can include iterative nearest point algorithm, point cloud registration algorithm, etc.

[0047] S270. Convert the mapping data to be matched into standard mapping data according to the coordinate mapping relationship.

[0048] Specifically, based on the spatial transformation parameters in the coordinate mapping relationship, each data point in the mapping data to be matched can be spatially transformed one by one, converting the mapping data to be matched from the original acquisition coordinate system to the current base map block coordinate system, thus obtaining standard mapping data.

[0049] S280. Remove overlapping data in the standard mapping data based on the overlapping parts of the current reference map block and the map block to be registered.

[0050] Overlapping data can include data from standard mapping data located within the overlapping area of ​​the current reference map tile and the map tile to be registered.

[0051] Specifically, the overlapping area between the current base map block and the map block to be registered can be determined based on a preset threshold for repeated road segment mileage. The data segment corresponding to this overlapping area is then located in the standard mapping data. The parts of the standard mapping data belonging to the overlapping area are then removed, retaining only the data from the non-overlapping areas.

[0052] S290. The cropped standard mapping data is stitched together with the baseline mapping data of the current baseline map block within the map.

[0053] The baseline mapping data may include mapping data that corresponds to the current baseline map tiles and whose coordinate system is fixed.

[0054] Specifically, the standard mapping data obtained after removing overlapping data can be stitched together with the baseline mapping data of the current baseline map block. Stitching methods include, but are not limited to, direct stitching, fine-grained stitching, and graph-optimized global stitching. During stitching, the starting point of the standard mapping data is spatially connected to the ending point of the baseline mapping data of the current baseline map block within the map to form a complete map.

[0055] The technical solution of this invention involves acquiring point cloud data and inertial measurement data, fusing them into mapping data, dividing the target road segment into at least one map segment according to preset single-segment mileage thresholds and repeated segment mileage thresholds, and then sequentially determining each map segment as the current reference map block according to the order of the target road segment. The endpoint timestamp of the mapping data of the current reference map block is extracted from the map of the target road segment. Using the endpoint timestamp as the starting point, mapping data of the mileage length of the corresponding map block to be registered is extracted as the mapping data to be matched. The coordinate mapping relationship between the mapping data to be matched and the current reference map block is obtained. The mapping data to be matched is then converted into standard mapping data according to the coordinate mapping relationship. Overlapping data in the standard mapping data is trimmed according to the overlapping portion between the current reference map block and the map block to be registered. Finally, the trimmed standard mapping data is stitched into the reference mapping data of the current reference map block within the map. The above technical solution, by sequentially determining each map segment as the current reference map block according to the target road segment order, extracts the endpoint timestamp of the mapping data of the current reference map block within the map of the target road segment, and then extracts the mapping data of the corresponding mileage length of the map block to be registered as the matching mapping data, using the endpoint timestamp as the starting point. This allows for segmented processing of large-scale scenes, achieving data dimensionality reduction and localized registration. By obtaining the coordinate mapping relationship of the matching mapping data synchronized to the current reference map block, the matching mapping data is converted into standard mapping data according to the coordinate mapping relationship. The overlapping data in the standard mapping data is then trimmed according to the overlapping part of the current reference map block and the map block to be registered. Finally, the trimmed standard mapping data is stitched to the reference mapping data of the current reference map block within the map, enabling the synthesis of high-precision point cloud maps and improving the quality of point cloud maps.

[0056] Furthermore, based on the above embodiments of the invention, obtaining the coordinate mapping relationship of the mapping data to be matched and synchronized to the current base map tile includes: Extract the baseline mapping data corresponding to the current baseline map tile within the map; Obtain the baseline overlap data and the overlap data to be matched in the overlapping part of the baseline mapping data and the mapping data to be matched, respectively; Singular value decomposition is performed on the baseline overlapping data and the overlapping data to be matched to obtain a coarse transformation matrix that transforms the coordinate system of the map block to be registered to the coordinate system of the current baseline map block. The mapping data to be matched and mapped is converted into mapping data to be synchronized based on the coarse transformation matrix of the map blocks to be registered; The fine transformation matrix between the mapping data to be synchronized and the baseline mapping data is determined according to the iterative nearest point rule, which serves as the coordinate mapping relationship.

[0057] The mapping data can include mapping data corresponding to the current reference map block and with a fixed coordinate system. The reference overlap data can include data from the reference mapping data located within the overlap area between the current reference map block and the map block to be registered. The overlapping data to be matched can include data from the mapping data to be matched located within the overlap area between the current reference map block and the map block to be registered. The coarse transformation matrix can include transformation parameters calculated using singular value decomposition to transform the data in the coordinate system of the map block to be registered to the coordinate system of the current reference map block. The mapping data to be synchronized can include data obtained by transforming the mapping data to be matched of the map block to be registered according to the coarse transformation matrix. The fine transformation matrix can include coordinate transformation parameters obtained using the iterative nearest point rule algorithm to perform fine-grained registration of the mapping data to be synchronized and the reference mapping data. The coordinate mapping relationship can include spatial transformation parameters that transform the mapping data to be synchronized to the coordinate system of the reference mapping data; these spatial transformation parameters can include rotation matrices, translation vectors, etc.

[0058] Specifically, the baseline mapping data corresponding to the current baseline map tile can be extracted from the map. Then, the baseline overlapping data and the overlapping data to be matched can be extracted from the baseline mapping data and the mapping data to be matched, respectively. Singular value decomposition can be used to calculate a coarse transformation matrix to convert the coordinate system of the map tile to be registered to the coordinate system of the current baseline map tile. The mapping data to be matched from the map tile to be registered is then transformed according to the coarse transformation matrix to obtain the mapping data to be synchronized. Using the baseline mapping data as the target, the iterative nearest point rule algorithm is used for fine registration of the mapping data to be synchronized, calculating the fine transformation matrix between the mapping data to be synchronized and the baseline data. This fine transformation matrix is ​​then used as the coordinate mapping relationship.

[0059] Furthermore, based on the above embodiments, the invention also includes: Extract each original point cloud data and divide each original point cloud data into a layer point cloud dataset with a first threshold number of data. Within each layer of the point cloud dataset, the original point cloud data is extracted as point cloud data according to the second threshold interval.

[0060] The raw point cloud data may include a set of 3D spatial points acquired by LiDAR without any processing. The first threshold number may include the number of layers into which the raw point cloud data is divided according to the LiDAR scan lines. The layer point cloud dataset may include several subsets into which the raw point cloud data is divided according to the first threshold number, and each subset may correspond to point cloud data on one or more consecutive scan lines. The second threshold interval may include sampling parameters for sampling points on each scan line at intervals within each layer point cloud dataset.

[0061] Specifically, raw point cloud data can be acquired, and the raw point cloud data can be divided into several layers of point cloud datasets according to the scan line number of each point cloud dataset and a first threshold number. Within each layer of point cloud datasets, the sequence of points on each scan line is sampled at a second threshold interval. The sampling method can include extracting one point from the sequence of points on each scan line at intervals of the second threshold number of points. The extracted raw point cloud data is then used as the point cloud data. Different layers can have the same or different second threshold intervals.

[0062] Furthermore, based on the above embodiments, the invention also includes: Acquire keyframe data from the map, and extract real-time localization and map building data and real-time dynamic differential localization data from the keyframe data; The first optimization function is constructed based on the binary edges corresponding to the instant positioning and map construction data and the unary edges of the real-time dynamic differential positioning data. The key frame data is then optimized based on the first optimization function. Based on loop closure detection of keyframe data; A second optimization function is constructed based on the binary edges corresponding to the instant positioning and map construction data, the unary edges of the real-time dynamic differential positioning data, and the binary edges of the loop closure detection. The keyframes are then optimized based on the second optimization function.

[0063] The keyframe data may include data used for optimization. Instantaneous localization and mapping (IAM) data refers to the localization and mapping data output by the front-end odometry system. Real-time dynamic differential localization (RTD) data may include localization data output by the RTD device. Binary edges may include constraint terms formed by the relationship between the current keyframe data and the preceding and following keyframe data. Unary edges may include constraint terms formed by the current keyframe data itself. The first optimization function may include an optimization objective function constructed based on the binary edges corresponding to the IAM data and the unary edges corresponding to the RTD data. Loop closure detection may include registration detection between keyframes with spatially close locations. The binary edges for loop closure detection may include constraint terms formed by the relationship between the preceding and following keyframe data in the path formed after loop closure detection. The second optimization function may include an optimization objective function constructed by adding the binary edges from loop closure detection to the first optimization function.

[0064] Specifically, keyframe data of the map can be acquired, and real-time localization and map construction data, as well as real-time dynamic differential localization data, can be extracted from the keyframe data. Constraints formed by the relationship between the current keyframe data and its preceding and following edges are used as binary edges, and current conditional terms formed by the current keyframe are used as unary edges. A first optimization function is constructed based on the binary edges corresponding to the real-time localization and map construction data and the unary edges of the real-time dynamic differential localization data. The keyframe data is then optimized based on this first optimization function. Registration detection can be performed between keyframes with similar spatial locations. A second optimization function is constructed based on the binary edges corresponding to the real-time localization and map construction data, the unary edges of the real-time dynamic differential localization data, and the binary edges of loop closure detection. The keyframes are then optimized based on this second optimization function. By constructing the second optimization function using the binary edges corresponding to the real-time localization and map construction data, the unary edges of the real-time dynamic differential localization data, and the binary edges of loop closure detection, and optimizing the keyframes based on this second optimization function, real-time optimization of the point cloud map can be achieved, reducing map ghosting and improving the quality of the point cloud map.

[0065] Example 3 Figure 3 This is a flowchart illustrating the overall principle of a high-precision point cloud map construction system provided in Embodiment 3 of the present invention.

[0066] For details, see Figure 3 This embodiment describes a flowchart illustrating the overall principle of a high-precision point cloud map construction system. The system comprises three parts: a front-end odometry unit, a segmented map registration unit, and a back-end optimization unit. Specifically: The front-end odometry employs a 3D LiDAR-Inertial Tightly Coupled Odometry (LIO) system. Raw sensor data is packaged and relayed by the data preprocessing control center before entering the front-end odometry system. First, point cloud downsampling is performed. Then, along with synchronously processed Inertial Measurement Unit (IMU) data, the data is fed into the core algorithm of the front-end odometry system to generate Simultaneous Localization and Mapping (SLAM) odometry output data. This output data includes synchronously processed Real-Time Kinematic (RTK) data. After one round of front-end odometry calculation, it is determined whether the point cloud matching residual value has reached a set threshold for the first time. If the threshold is reached, the odometry output data is sent to the segmented map registration unit.

[0067] The tiled map registration unit is used to register two adjacent time-series maps to eliminate the impact of accumulated errors from the front-end odometry. First, the tiled map receives and stores keyframe data from the front-end odometry unit. Once the point cloud map registration conditions are met, trajectory data from the common overlapping areas of the reference trajectory and the trajectory to be registered are extracted. Singular value decomposition is used to calculate the transformation matrix from the trajectory to the reference trajectory, completing coarse registration. Based on the two coarsely registered trajectories, the reference map and the map to be registered are synthesized. A multi-threaded concurrent iterative closest point (ICP) algorithm is used for fine matching, further obtaining the precise transformation matrix from the map to the reference map. Finally, based on the coarse and fine transformation matrices, the original trajectory to be registered is transformed to the coordinate system of the reference trajectory, and new keyframe data is generated according to set conditions and sent to the back-end optimization unit.

[0068] The backend optimization unit receives and stores keyframe data sent by the segmented map registration unit. Once all keyframe data is received and stored, optimization begins. First, a Levenberg-Marquardt optimization problem is constructed based on the simultaneous localization and mapping (SMR) trajectory and the relative real-time dynamic difference trajectory. Optimization vertices, multi-element edges for SMR, and unary edges for real-time dynamic difference are added, and the first layer of optimization is performed to obtain the first-round optimized trajectory. Based on the first-round optimized trajectory, loop closure candidate frames are selected according to set conditions, and loop closure detection is performed using the Normal Distributions Transform (NDT) algorithm. Then, a second-layer Levenberg-Marquardt (LM) optimization problem is constructed, adding optimization vertices, multi-element edges for SMR, unary edges for real-time dynamic difference, and multi-element edges for loop closure. The second layer of optimization is then performed to obtain the final optimized trajectory. Based on the final optimized trajectory, the corresponding point cloud data is read, a high-precision point cloud map is synthesized, and the map is output and stored.

[0069] This embodiment also proposes a lossless hierarchical interval downsampling method, specifically: In the preprocessing of multi-line LiDAR point cloud data, if the number of points in a single frame of LiDAR point cloud is too large (e.g., a 128-line mechanical LiDAR has approximately 600,000 points in one frame), downsampling of the point cloud in a single frame is necessary to ensure the real-time operation of the odometry calculation. However, it is also necessary to retain the original information of each point, and the density requirements for different locations on the point cloud map vary in different application scenarios (e.g., high-precision map rendering requires clearer ground points, while surrounding buildings do not require denser points, so the point cloud density on the ground can be higher than that of the surrounding buildings). Based on these factors, this invention designs a hierarchical interval downsampling algorithm based on Ring information for multi-line LiDAR (mechanical or semi-solid-state). Specific implementation: First, the LiDAR point cloud is divided into... N Layers (e.g., lines 1-30) Lines 31-50 are For each layer, an indirect filtering parameter is set. M (For example: The indirect filtering parameters of the layer are , The indirect filtering parameters of the layer are ),parameter The meaning refers to in Within each layer, points corresponding to each Ring are individually sampled at intervals of M-1 (e.g., M=2 means that one point is sampled every 1 point in the point cloud of the corresponding Ring). This method enables downsampling at different densities between different layers while preserving the original information of each point for point cloud distortion correction applications in LIO front-end odometry.

[0070] Figure 4 This is a schematic diagram of a road segmentation principle provided in Embodiment 3 of the present invention.

[0071] Figure 5 This is a schematic diagram of the principle process of block map registration according to Embodiment 3 of the present invention.

[0072] For details, see Figure 4 and Figure 5 This embodiment illustrates a schematic diagram of a road segmentation principle and a schematic diagram of a block map registration principle. The block map registration principle consists of three cyclical steps: segmentation, registration, and synthesis. This cyclical execution is based on a time series, continuing until the last map block completes registration. Finally, the map enters the optimization center for trajectory optimization, thus enabling the construction of a large-scale map. Specifically: Segmentation: Based on the performance of simultaneous localization and mapping (SMR) and front-end odometry, the target road segments requiring point cloud map construction are segmented according to time series and single-segment mileage thresholds. Divided into several units ( ,i =1, 2, 3, ..., n), the segmentation size principle is determined based on the controllable range of cumulative error of the simultaneous localization and mapping (SMR) front-end odometry (e.g., within 1km, the error of the SMR front-end odometry is ≤10cm; when the target accuracy requirement is ≤10cm, then the trajectory length of each unit is ≤1km). Simultaneous localization and mapping (SMR) calculations are performed on each unit segment using the front-end odometry. The results of each frame of LiDAR SMR and map building calculations are sent to the segmented point cloud map registration center for storage and processing. Every two adjacent units need to have overlapping segments (…). , i =1, 2, 3, ..., n), with repeated road segments used for point cloud map registration. In this method, the point cloud map corresponding to the n road segments is ( , i =1, 2, 3, ..., n) have only two attributes: the base map ( ) and maps to be registered ( According to this definition, the first map The attributes are only for the base map, not the end map. The attribute is only the map to be registered; other maps are both the base map and the map to be registered. The starting point is End point location and threshold length of repeated road segments Decision, record The timestamp of the last frame of LiDAR data is Starting from this position, the cumulative mileage is calculated recursively backwards. When the mileage length is not less than At that time, find the first frame of LiDAR data that just meets the threshold condition and record the corresponding timestamp. According to timestamp The corresponding lidar data is then used as The starting data is used to begin the simultaneous localization and mapping (Localization and Mapping) front-end mileage calculation method. When the accumulated mileage reaches a threshold... Afterwards, the simultaneous localization and mapping (SMR) calculations for this round of data are terminated, and the data enters the registration center for point cloud map registration. Once the tile map registration center completes the registration, it will calculate the starting timestamp of the next round of SMR calculations. The front-end odometer is based on A new round of simultaneous localization and mapping (SMR) calculations begins.

[0073] Registration: This step registers overlapping areas between adjacent maps in order to calculate the trajectory of the map to be registered. To the baseline map trajectory The transformation parameters, thereby transforming Precisely transform to In the coordinate system, cropping and stitching are performed to obtain a new simultaneous localization and map-building trajectory. According to the previous definition, when... When the attribute is the base map, then The attribute is the map to be registered, denoted as... The map of the repeating areas is (The corresponding trajectory is) Similarly, remember The map of the repeating areas is (The corresponding trajectory is) ).because It exists within its own world coordinate system, and... Because the coordinate systems are different, singular value decomposition is used first for calculation. arrive coarse transformation matrix ,use Will coarse transformation to In the coordinate system, the corresponding map to be registered is then generated. Registration was performed using an improved version of the Iterative Closest Point (ICP) algorithm. and get and Exact transformation matrix between The iterative closest point algorithm uses multi-threaded concurrency, effectively accelerating the registration process. Finally, based on the transformation matrix... and Will Precise transformation to In the coordinate system, the accurate transformation trajectory is obtained. , trajectory and The repeated regions are cropped, and finally, new simultaneous localization and map building keyframes can be obtained based on the trajectory of the non-repeating regions.

[0074] Synthesis: The result obtained after completing the registration process between the reference trajectory and the trajectory to be registered. Cut it and After the trajectory of the repeated region, the remaining trajectory is denoted as... To reduce the amount of data, based on a distance threshold... and angle threshold extract The keyframes extracted from the data will be sent to the backend optimization center for storage and processing, thus completing the synthesis of the registered trajectory.

[0075] After performing the above operations, perform backend optimization, specifically: After all the map tiles have been registered and keyframes extracted, the backend optimization center, having obtained all the keyframe data, begins trajectory optimization. (The simultaneous localization and mapping (SMR) trajectory and the real-time dynamic differential trajectory involved in the optimization are both trajectories in a relative coordinate system; that is, the first point of the SMR trajectory and the real-time dynamic differential trajectory is defined as the origin, thus establishing the coordinate system of the SMR trajectory and the real-time dynamic differential trajectory.) In this embodiment of the invention, the backend optimization center uses a two-layer optimization architecture based on the Levenberg-Marquardt nonlinear least squares method to optimize the SMR trajectory.

[0076] Figure 6 This is a schematic diagram of a loop closure detection candidate frame selection principle provided in Embodiment 3 of the present invention.

[0077] For details, see Figure 6 This embodiment illustrates a schematic diagram of a loop closure detection candidate frame selection principle. This loop closure detection candidate frame selection includes a first-layer optimization and a second-layer optimization, specifically: First-level optimization (SLAM+RTK): Optimization targets (original LiDAR SLAM positioning points) and constraint edges (LiDAR SLAM binary edges and RTK unary edges) are added based on keyframe data. The weight coefficients of the RTK unary edges are set based on the RTK variance. SLAM binary edges are fixed with the same weight coefficient. After the optimization problem configuration is completed, the optimization is executed, optimizing the optimization target (original SLAM localization points) to fit the normal RTK trajectory, thus obtaining the first layer of optimized trajectory. , This will serve as the optimization objective for the second layer and will also be used to filter candidate frames for loop closure detection.

[0078] The second layer of optimization (SLAM+RTK+loop closure factor) is based on the optimization results of the first layer. First, candidate frames for loop closure detection are selected (candidate frames for loop closure detection exist in pairs). One round of iteration is then performed. The trajectory is considered to be in a relatively close spatial distance (a distance threshold parameter is set). However, the time intervals are relatively large (using keyframe identifiers instead of time thresholds, setting parameters). A loop closure detection is performed on the keyframe of the test. To reduce the number of detection points, an identifier interval parameter is set. Each interval Each keyframe is checked once. After all loop closure candidate frames have been filtered, registration is performed on each candidate frame pair using a single-frame point cloud to sub-map approach based on the Normal Distributions Transform (NDT) algorithm. Candidate frame pairs are defined. , , Generate a single-frame point cloud. The process involves extracting nearby keyframes to construct a sub-map, using the NDT algorithm to register the single-frame point cloud onto the sub-map, and calculating the transformation matrix parameters from the single-frame point cloud to the sub-map. This operation is performed on all candidate frame pairs. This system employs a task-based parallel programming model (Intel Threading Building Blocks, TBB) and a concurrent NDT algorithm to register loop closure candidate frame pairs, significantly improving detection efficiency. After completing all loop closure detections, an optimization problem is constructed, and optimization objectives are added based on the keyframe data. (Location points), and add constraint edges (LiDAR SLAM binary edges, RTK unary edges, and closure factor binary edges). Set the weight coefficients of the RTK unary edges based on the RTK variance. In order to reduce the constraint strength of the RTK unary edge in the second round of optimization and avoid affecting the optimization effect of the closure factor binary edge, it is necessary to... Multiply by a decay factor Obtain the weight coefficients of the second-layer RTK constraint edges. (Right now = * SLAM binary edges are fixed with the same weight coefficient. Generally, it is related to the first layer of optimization. Same. The same weight coefficient is always set for the binary edge of the laparoscing factor. , The size depends on the accuracy of the loop closure; the higher the accuracy, the better. The larger the value, the smaller the value. After constructing the optimization problem, the second-level optimization work begins, and the completed trajectory is denoted as... ,at this time The coordinate system is in a relative coordinate system, and it needs to be transformed to the RTK world coordinate system to obtain the final point cloud map construction trajectory. , Each point in the image corresponds to a frame of LiDAR point cloud, based on... By stitching together the corresponding LiDAR point clouds, a target point cloud map can be obtained. To prevent the point cloud map from becoming too large, this method divides all maps into blocks, i.e., setting a mileage threshold for each unit of the point cloud map. Starting from the starting point, each accumulated mileage reaches Then the synthesis of one unit map is completed, and the total mileage is set to 1. Finally, we get N = / (Round up) map blocks.

[0079] Figure 7 This is a schematic diagram of a portable and detachable data acquisition device for a point cloud map construction system according to Embodiment 3 of the present invention.

[0080] For details, see Figure 7 This embodiment illustrates a portable and detachable data acquisition device for a point cloud map construction system. The device internally includes an RTK host, an RTK main antenna, an RTK slave antenna, functional modules, a 3D LiDAR, and four fixed suction cups. The functional modules include a processor, read-only memory (ROM), random access memory (RAM), a communication unit, and a storage unit. Specifically: RTK host refers to the central processing unit used to receive and process real-time dynamic differential positioning signals. RTK main antenna refers to the primary antenna used for receiving signals, typically mounted on the top of the device in an open location. RTK slave antenna refers to a secondary antenna used to assist in receiving signals. Functional module refers to a circuit board assembly integrating a processor, memory, communication unit, and storage unit, responsible for data acquisition, processing, storage, and communication. 3D LiDAR refers to a sensor used to acquire three-dimensional point cloud data of the environment; 3D LiDAR can include mechanical LiDAR and semi-solid-state LiDAR. A mounting cup refers to a detachable mounting component used to secure the entire data acquisition device. Processor refers to the arithmetic unit used to execute computer programs and control the operation of various modules. Read-only memory refers to memory used to store firmware and system parameters. Random access memory refers to memory used to temporarily store running data and program code. Communication unit refers to the communication interface used to enable data exchange between the device and an external computer. Storage unit refers to storage devices used to persistently store acquisition results such as point cloud data and trajectory data.

[0081] In this embodiment, the portable and detachable data acquisition device of the point cloud map building system is firmly attached to the top of the vehicle by four fixed suction cups. The RTK main antenna and RTK receive signals from the antenna and transmit them to the RTK host. The RTK host calculates the positioning data. The 3D LiDAR collects point cloud data of the surrounding environment. The processor in the functional module runs the point cloud map building program. Using the read-only memory and temporary data in the random access memory, the system interacts with external devices through the communication unit and saves the final map data to the storage unit.

[0082] The technical solution of this invention can segment large-scale scenes through a front-end odometer unit to achieve data dimensionality reduction and localized registration. Through a block map registration unit and a back-end optimization unit, it can achieve the synthesis of high-precision point cloud maps, solving the problem that the accuracy and quality of point cloud maps are reduced in large-scale scenes in the existing SLAM technology, and that map ghosting occurs.

[0083] Example 4 Figure 8 This is a schematic diagram of a map building device provided in Embodiment 4 of the present invention. Figure 8 As shown, the device includes: a data fusion module 310, a candidate determination module 320, a synchronization relationship module 330, and a map stitching module 340; wherein, The data fusion module 310 is used to acquire point cloud data and inertial measurement data, and to fuse the point cloud data and inertial measurement data into mapping data.

[0084] The candidate determination module 320 is used to determine the current reference map block based on the map segmentation strategy of the target road segment, and extract the map data to be matched for the corresponding map block to be registered in the map data according to the current reference map block. The current reference map block and the map block to be registered have some overlap.

[0085] The synchronization relationship module 330 is used to obtain the coordinate mapping relationship of the mapping data to be matched to the current base map block.

[0086] The map stitching module 340 is used to map the mapping data to be matched to the standard mapping data according to the coordinate mapping relationship, and stitch the map to be registered into the target road segment map based on the standard mapping data.

[0087] The technical solution of this invention involves acquiring point cloud data and inertial measurement data through a data fusion module, and fusing the point cloud data and inertial measurement data into mapping data. A candidate determination module determines the current reference map block based on the mapping strategy of the target road segment, and extracts the mapping data to be matched for the corresponding map block to be registered within the mapping data based on the current reference map block, wherein the current reference map block and the map block to be registered partially overlap. A synchronization relationship module obtains the coordinate mapping relationship between the mapping data to be matched and the current reference map block. A map stitching module maps the mapping data to be matched into standard mapping data according to the coordinate mapping relationship, and stitches the map block to be registered onto the map of the target road segment based on the standard mapping data. The above technical solution determines the current reference map block by using a segmentation strategy based on the target road segment, and extracts the matching mapping data of the corresponding map block to be registered from the mapping data based on the current reference map block. This allows for segmentation processing of large-scale scenes, achieving data dimensionality reduction and localized registration. By mapping the matching mapping data to standard mapping data through coordinate mapping relationships, and stitching the map block to be registered to the map of the target road segment based on the standard mapping data, a high-precision point cloud map can be synthesized. This solves the problem in existing technologies where the accuracy and quality of point cloud maps decrease in large-scale scenes due to SLAM technology, and also causes map ghosting.

[0088] Optional, the data fusion module 310 is specifically used for: Acquire point cloud data and inertial measurement data, and then fuse the point cloud data and inertial measurement data into mapping data.

[0089] Optionally, the candidate determination module 320 is specifically used for: The map segmentation strategy based on the target road segment determines the current reference map block, and extracts the matching map data of the corresponding map block to be registered in the map data based on the current reference map block. There is some overlap between the current reference map block and the map block to be registered.

[0090] Optionally, the current baseline map tiles are determined based on the target road segment's subdivision strategy, including: The target road segment is divided into at least one map segment according to the preset single road segment mileage threshold and the repeated road segment mileage threshold; Each map segment is determined as the current base map block in the order of the target road segment.

[0091] Optionally, based on the current base map tiles, extract the mapping data to be matched for the corresponding map tile to be registered within the mapping data, including: Extract the timestamp of the endpoint location of the mapping data of the current base map block within the map of the target road segment; Starting from the timestamp of the endpoint, extract the mapping data of the mileage length of the corresponding map block to be registered as the mapping data to be matched.

[0092] Optional, the synchronization module 330 is specifically used for: Obtain the coordinate mapping relationship of the mapping data to be matched and synchronize it to the current base map block.

[0093] Optionally, obtain the coordinate mapping relationship between the mapping data to be matched and the current base map tile, including: Extract the baseline mapping data corresponding to the current baseline map tile within the map; Obtain the baseline overlap data and the overlap data to be matched in the overlapping part of the baseline mapping data and the mapping data to be matched, respectively; Singular value decomposition is performed on the baseline overlapping data and the overlapping data to be matched to obtain a coarse transformation matrix that transforms the coordinate system of the map block to be registered to the coordinate system of the current baseline map block. The mapping data to be matched and mapped is converted into mapping data to be synchronized based on the coarse transformation matrix of the map blocks to be registered; The fine transformation matrix between the mapping data to be synchronized and the baseline mapping data is determined according to the iterative nearest point rule, which serves as the coordinate mapping relationship.

[0094] Optional, map stitching module 340, specifically used for: Based on the coordinate mapping relationship, the mapping data to be matched is mapped to standard mapping data, and the map to be registered is pieced together into the map of the target road segment based on the standard mapping data.

[0095] Optionally, the mapping data to be matched is mapped to standard mapping data according to the coordinate mapping relationship, and the map to be registered is pieced together from the standard mapping data and stitched onto the map of the target road segment, including: Convert the mapping data to be matched into standard mapping data according to the coordinate mapping relationship; Remove overlapping data from the standard mapping data based on the overlapping parts of the current reference map blocks and the map blocks to be registered; The cropped standard mapping data is stitched together with the baseline mapping data of the current baseline map block within the map.

[0096] Optional, also includes: Extract each original point cloud data and divide each original point cloud data into a layer point cloud dataset with a first threshold number of data. Within each layer of the point cloud dataset, the original point cloud data is extracted as point cloud data according to the second threshold interval.

[0097] Optional, also includes: Acquire keyframe data from the map, and extract real-time localization and map building data and real-time dynamic differential localization data from the keyframe data; The first optimization function is constructed based on the binary edges corresponding to the instant positioning and map construction data and the unary edges of the real-time dynamic differential positioning data. The key frame data is then optimized based on the first optimization function. Based on loop closure detection of keyframe data; A second optimization function is constructed based on the binary edges corresponding to the instant positioning and map construction data, the unary edges of the real-time dynamic differential positioning data, and the binary edges of the loop closure detection. The keyframes are then optimized based on the second optimization function.

[0098] The map building apparatus provided in the embodiments of the present invention can execute the map building method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0099] It is worth noting that the various modules included in the above-mentioned map building device are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of each functional module are only for easy differentiation and are not used to limit the protection scope of the embodiments of the present invention.

[0100] Example 5 Figure 9 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0101] like Figure 9As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0102] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0103] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as map building methods.

[0104] In some embodiments, the map building method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the map building method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the map building method by any other suitable means (e.g., by means of firmware).

[0105] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0106] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0107] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0108] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0109] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0110] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0111] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the map construction method provided in any embodiment of this invention.

[0112] In implementing a computer program product, computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include, but are not limited to, object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0113] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0114] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A map construction method, characterized in that, The method includes: Acquire point cloud data and inertial measurement data, and fuse the point cloud data and inertial measurement data into mapping data; The current reference map block is determined based on the map segmentation strategy of the target road segment, and the map data to be matched for the corresponding map block to be registered is extracted from the map data based on the current reference map block. The current reference map block and the map block to be registered have some overlap. Obtain the coordinate mapping relationship between the mapping data to be matched and the current base map block; The mapping data to be matched is mapped to standard mapping data according to the coordinate mapping relationship, and the map to be registered is pieced together into the map of the target road segment based on the standard mapping data.

2. The method according to claim 1, characterized in that, Also includes: Extract each original point cloud data, and divide each original point cloud data into a layer point cloud dataset with a first threshold number of data. Within each of the aforementioned layer point cloud datasets, the original point cloud data is extracted as the point cloud data according to a second threshold interval.

3. The method according to claim 1, characterized in that, The target road segment-based map segmentation strategy determines the current baseline map tiles, including: The target road segment is divided into at least one map segment according to the preset single road segment mileage threshold and the repeated road segment mileage threshold; The map segments are sequentially determined as the current base map blocks according to the order of the target road segments.

4. The method according to claim 1 or 3, characterized in that, The step of extracting the matching mapping data of the corresponding map block to be registered within the mapping data based on the current reference map block includes: Extract the endpoint location timestamp of the mapping data of the current reference map block within the map of the target road segment; Starting from the timestamp of the endpoint location, the mapping data corresponding to the mileage length of the map block to be registered is extracted as the mapping data to be matched.

5. The method according to claim 1, characterized in that, The step of obtaining the coordinate mapping relationship of the mapping data to be matched and synchronized to the current base map tile includes: Extract the baseline mapping data corresponding to the current baseline map block within the map; Obtain the baseline overlap data and the overlap data to be matched in the overlapping part of the baseline mapping data and the mapping data to be matched, respectively; Singular value decomposition is performed on the reference overlapping data and the overlapping data to be matched to obtain a coarse transformation matrix that transforms the coordinate system of the map block to be registered to the coordinate system of the current reference map block. The mapping data to be matched in the map block to be registered is converted into mapping data to be synchronized according to the coarse transformation matrix. The fine transformation matrix between the mapping data to be synchronized and the reference mapping data is determined according to the iterative nearest point rule and used as the coordinate mapping relationship.

6. The method according to claim 1, characterized in that, The step of mapping the mapping data to be matched to standard mapping data according to the coordinate mapping relationship, and then stitching the map to be registered into blocks based on the standard mapping data to the map of the target road segment, includes: The mapping data to be matched is converted into the standard mapping data according to the coordinate mapping relationship; The overlapping data in the standard mapping data is removed according to the overlapping part of the current reference map block and the map block to be registered; The cropped standard mapping data is stitched together with the baseline mapping data of the current baseline map block within the map.

7. The method according to claim 1, characterized in that, Also includes: Acquire keyframe data of the map, and extract real-time positioning and map construction data and real-time dynamic differential positioning data from the keyframe data; A first optimization function is constructed based on the binary edges corresponding to the instant positioning and map construction data and the unary edges of the real-time dynamic differential positioning data, and the keyframe data is optimized based on the first optimization function. Based on loop closure detection of the keyframe data; A second optimization function is constructed based on the binary edges corresponding to the instant positioning and map construction data, the unary edges of the real-time dynamic differential positioning data, and the binary edges of the loop closure detection. The keyframe is then optimized based on the second optimization function.

8. A map building device, characterized in that, The device includes: The data fusion module is used to acquire point cloud data and inertial measurement data, and fuse the point cloud data and the inertial measurement data into mapping data; The candidate determination module is used to determine the current reference map block based on the map segmentation strategy of the target road segment, and extract the map data to be matched for the corresponding map block to be registered in the map data according to the current reference map block, wherein the current reference map block and the map block to be registered have partial overlap. The synchronization relationship module is used to obtain the coordinate mapping relationship between the mapping data to be matched and the current base map block; The map stitching module is used to map the mapping data to be matched into standard mapping data according to the coordinate mapping relationship, and stitch the map to be registered into the target road segment map based on the standard mapping data.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the map construction method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a processor to perform the map construction method according to any one of claims 1-7.

Images