Point cloud map processing method and device

Abstract

The invention discloses a point cloud map processing method and device. The method relates to the technical field of automatic driving, and comprises the following steps: acquiring a current point cloud map of a target range; each point in the current point cloud map is subjected to quality inspection, a target point set in the current point cloud map is obtained, the distance between each target point in the target point set and the fitting plane is within a preset distance range, and the precision of the target point meets the preset precision; the fitting plane is used for representing a plane obtained by fitting each point in the current point cloud map; clustering the abnormal points in the target point set based on the semantic information of each target point in the target point set to obtain a clustering result; at least one abnormal area in the target range is determined based on the clustering result, and the semantic information is used for representing the type of the target object corresponding to the target point. According to the invention, the technical problem of low detection accuracy of the point cloud map in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a method and device for processing a point cloud map. Background Art

[0002] In the field of autonomous driving technology, the construction and maintenance of high-precision maps are crucial for vehicle navigation and positioning. Traditional calibration site maps rely on manual collection and construction to ensure the high accuracy and reliability of maps. However, during the long-term use of these calibration sites, due to natural environmental factors and human factors such as heavy trucks running over them, the terrain of the ground and retaining walls may change, resulting in a decrease in the detection accuracy of point cloud maps.

[0003] In the related art, calibrated collection vehicles or drones are used to collect point clouds to build high-precision maps. However, since it is impossible to conduct regular inspections of the calibration site, the environmental changes of the calibration site caused by the rolling of vehicles make it difficult to detect rough areas of the calibration site, resulting in random errors in the detected point cloud map, which in turn leads to low accuracy of point cloud map detection in the related art.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a method and device for processing a point cloud map, so as to at least solve the technical problem of low point cloud map detection accuracy in the related art.

[0006] According to one aspect of an embodiment of the present invention, a method for processing a point cloud map is provided, comprising: obtaining a current point cloud map of a target range; performing quality inspection on each point in the current point cloud map to obtain a target point set in the current point cloud map, wherein the distance between each target point in the target point set and a fitting plane is within a preset distance range, and the accuracy of the target point satisfies the preset accuracy, and the fitting plane is used to characterize the plane obtained by fitting each point in the current point cloud map; clustering abnormal points in the target point set based on semantic information of each target point in the target point set to obtain a clustering result; and determining at least one abnormal area within the target range based on the clustering result, wherein the semantic information is used to characterize the type of target object corresponding to the target point.

[0007] In an embodiment of the present invention, the current point cloud map of the target range is first obtained, and each point in the current point cloud map is quality inspected to obtain a target point set in the current point cloud map. Then, based on the semantic information of each target point in the target point set, the abnormal points in the target point set are clustered to obtain a clustering result. Finally, at least one abnormal area in the target range is determined based on the clustering result, so as to achieve the purpose of improving the detection accuracy of the point cloud map. It is easy to notice that the distance between each target point in the target point set and the fitting plane is within a preset distance range, and the accuracy of the target point meets the preset accuracy. After obtaining the target point set, the abnormal points in the target set points are clustered based on the semantic information, and finally the abnormal area is determined based on the clustering result, and then the abnormal area is adjusted to improve the detection accuracy of the point cloud map. In the above process, a target point set that meets the requirements is screened out by a preset distance range and a preset accuracy, and then the abnormal point clustering result is determined from the target point set based on the semantic information, and the abnormal area is identified, so that the abnormal area can be adjusted, and the abnormal area with potential problems is processed, thereby solving the technical problem of low detection accuracy of the point cloud map in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0009] Figure 1 is a flow chart of a method for processing a point cloud map according to an embodiment of the present invention;

[0010] Figure 2 is a graph of abnormal point detection results of a point cloud map according to an embodiment of the present invention;

[0011] Figure 3 is another abnormal point detection result diagram of a point cloud map according to an embodiment of the present invention;

[0012] Figure 4 is a rough area detection result diagram of a point cloud map according to an embodiment of the present invention;

[0013] Figure 5 is a calibration result diagram of a point cloud map according to an embodiment of the present invention;

[0014] Figure 6 is a schematic diagram of a point cloud map processing device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0016] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] According to an embodiment of the present invention, an embodiment of a method for processing a point cloud map is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0018] Figure 1 is a flow chart of a method for processing a point cloud map according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0019] Step S102, obtaining the current point cloud map of the target range.

[0020] The target range in the above steps is used to describe the specific area that is set to be processed in the perception system of the autonomous vehicle. The target range may include but is not limited to a static target range and a dynamic target range. The static target range may be a calibration site. The actual target range depends on the task requirements and environmental conditions. The role of the target range may include but is not limited to clarifying the calibration site to achieve more accurate environmental understanding and decision-making.

[0021] The current point cloud map in the above steps is used to describe the current point cloud representation of the target range, that is, the coordinates of each point on the target range in three-dimensional space. The functions of the current point cloud map may include but are not limited to determining the location coordinates of the calibration site, environmental perception, obstacle detection, positioning and navigation, etc.

[0022] The acquisition method in the above steps may include but is not limited to lidar scanning, structured light scanning, etc. By using a lidar device to scan the calibration site, a three-dimensional point cloud map can be generated to obtain the position coordinates of each point on the calibration site, which is convenient for subsequent quality inspection of the acquired points.

[0023] In an optional embodiment, a current point cloud map of the target range is obtained through a laser radar or other equipment to obtain the position coordinate information of multiple points in the calibration site, so as to facilitate subsequent quality inspection of the multiple points and screen out the accurate point coordinate positions. At the same time, with limited computing resources, by focusing on the target range, the data processing process can be optimized and efficiency can be improved.

[0024] Step S104, performing quality inspection on each point in the current point cloud map to obtain a target point set in the current point cloud map, wherein the distance between each target point in the target point set and the fitting plane is within a preset distance range, and the accuracy of the target point meets the preset accuracy, and the fitting plane is used to characterize the plane obtained by fitting each point in the current point cloud map.

[0025] The quality inspection in the above steps is used to describe the detection and screening of each point in the acquired point cloud map, and select points that meet specific quality standards to improve the accuracy and reliability of the map. The types of quality inspection may include but are not limited to target range roughness inspection, data integrity inspection, outlier detection, etc. The specific quality inspection needs to be determined according to actual needs and is not limited here.

[0026] The target point set in the above steps is used to describe the set of points in the point cloud that meet specific conditions after quality inspection. The role of the target point set can include but is not limited to further analysis and processing, such as target range roughness detection, coordinate position detection and identification, etc.

[0027] The fitting plane in the above steps is used to describe the plane calculated from the point cloud data by mathematical methods, and make the plane as close to these points as possible, wherein the mathematical method may include but is not limited to the least squares method, and the fitting plane may include but is not limited to a two-dimensional plane such as a ground plane, and a three-dimensional plane such as a room ceiling. The specific fitting plane depends on the actual application scenario and is not limited here. The role of the fitting plane may include but is not limited to characterizing the geometric structure in the point cloud data, and can be used to identify the ground, walls or other plane structures.

[0028] The preset distance range in the above steps is used to describe the range of distance settings between points in the point cloud and the fitting plane. Only points within this range will be considered as target points. The types of preset distance ranges may include but are not limited to fixed preset distance ranges, dynamic preset distance ranges, etc. The specific preset distance ranges depend on the actual application requirements and environmental conditions, which are not limited here. The role of the preset distance range may include but is not limited to filtering out points that are too far or too close to the fitting plane to ensure that the relationship between the points in the target point set and the fitting plane is meaningful.

[0029] The preset accuracy in the above steps is used to describe the accuracy requirements for the points in the point cloud data. Only points that meet this accuracy requirement will be considered as target points. The types of preset accuracy can include but are not limited to absolute accuracy and relative accuracy. The specific preset accuracy needs to be determined according to the actual situation and is not limited here. The role of the preset accuracy can include but is not limited to ensuring that the points in the target point set have sufficient accuracy to meet the needs of subsequent processing.

[0030] In an optional embodiment, all points in the current point cloud map are first fitted by the least squares method to determine a fitting plane, and then each point is quality checked to determine whether they meet a specific preset distance range and preset accuracy. After quality inspection, all points that meet the conditions are formed into a target point set, which will be used for further analysis and application. In the above process, all points are quality checked by a specific preset distance range and preset accuracy, and points that meet the requirements are screened out, which improves the quality of the data set and reduces the impact of errors. At the same time, the high-quality target point set obtained can be used for more accurate measurement and analysis, improving the accuracy and reliability of subsequent processing.

[0031] Step S106: clustering the abnormal points in the target point set based on the semantic information of each target point in the target point set to obtain a clustering result.

[0032] The semantic information in the above steps is used to describe the type or characteristics of the target object corresponding to the target point, and the role of the voice information may include but is not limited to helping understand and identify the type or characteristics of the target point.

[0033] The outliers in the above steps are used to describe points whose semantic information is significantly different from that of most target points. The types of outliers may include but are not limited to data with errors in coordinate positions, wrongly labeled data, and noise data. The specific outliers need to be determined based on actual conditions and are not limited here. The functions of outliers may include but are not limited to rough area determination, anomaly detection, safety monitoring, etc.

[0034] The clustering in the above steps is used to describe dividing the points in the target point set into multiple groups so that the target points with similar semantic information are clustered in the same group. Here, the abnormal points are clustered based on the semantic information. The types of clustering methods may include but are not limited to hierarchical clustering, spectral clustering, density clustering, partition clustering, etc. The specific clustering method needs to be determined according to the actual situation and is not limited here.

[0035] The clustering result in the above step is used to describe the result of clustering the outliers in the target point set according to the semantic information, and each cluster contains semantically similar target points. The role of the clustering result may include but is not limited to determining different outlier areas according to multiple outliers.

[0036] In an optional embodiment, by analyzing and utilizing the semantic information of the target point, it is possible to identify and cluster those abnormal points that are inconsistent with the normal data pattern. The clustering results can help us better understand the abnormal patterns in the data set, thereby achieving more accurate analysis and decision-making in various application scenarios, such as anomaly detection, anomaly classification and identification. The above process can extract the required information from a large amount of complex data, cluster the required information, improve the efficiency and accuracy of subsequent processing, and also provide a basis for further data analysis.

[0037] Step S108: determining at least one abnormal area within the target range based on the clustering result, wherein the semantic information is used to characterize the type of the target object corresponding to the target point.

[0038] The abnormal area in the above steps is used to describe the area where the abnormal points are located in the target range. The types of abnormal areas may include but are not limited to position abnormal areas, density abnormal areas, cluster abnormal areas, etc. The specific abnormal areas need to be determined based on the distribution of abnormal points and quality inspection conditions, which are not limited here. The role of abnormal areas may include but are not limited to error identification, timely processing of abnormal areas, etc.

[0039] In an optional embodiment, based on the clustering results obtained by clustering the abnormal points in the target point set, the abnormal area within the target range is determined. This process can process each cluster by solving the convex hull to obtain multiple abnormal areas. The above method uses anomaly detection technology to identify patterns and structures in all points, and determines multiple abnormal areas based on semantic information, thereby improving the efficiency of anomaly detection, preventing potential risks and improving safety.

[0040] In an embodiment of the present invention, the current point cloud map of the target range is first obtained, and each point in the current point cloud map is quality inspected to obtain a target point set in the current point cloud map. Then, based on the semantic information of each target point in the target point set, the abnormal points in the target point set are clustered to obtain a clustering result. Finally, at least one abnormal area in the target range is determined based on the clustering result, so as to achieve the purpose of improving the detection accuracy of the point cloud map. It is easy to notice that the distance between each target point in the target point set and the fitting plane is within a preset distance range, and the accuracy of the target point meets the preset accuracy. After obtaining the target point set, the abnormal points in the target set points are clustered based on the semantic information, and finally the abnormal area is determined based on the clustering result, and then the abnormal area is adjusted to improve the detection accuracy of the point cloud map. In the above process, a target point set that meets the requirements is screened out by a preset distance range and a preset accuracy, and then the abnormal point clustering result is determined from the target point set based on the semantic information, and the abnormal area is identified, so that the abnormal area can be adjusted, and the abnormal area with potential problems is processed, thereby solving the technical problem of low detection accuracy of the point cloud map in the related art.

[0041] Optionally, quality inspection is performed on each point in the current point cloud map to obtain a target point set in the current point cloud map, including: quality inspection is performed on each point in the current point cloud map to obtain a first candidate point set in the current point cloud map, wherein the distance between each first candidate point in the first candidate point set and the fitting plane is within a preset distance range; based on the semantic information of each first candidate point in the first candidate point set, quality inspection is performed on multiple first candidate points with the same semantic information to obtain a second candidate point set in the first candidate point set, wherein the accuracy of each second candidate point in the second candidate point set meets the preset accuracy; based on the position coordinates of multiple marking devices deployed within the target range, quality inspection is performed on device points in the second candidate point set corresponding to multiple marking devices to obtain a target point set, wherein the accuracy of the device points with the same semantic information in the target point set meets the preset accuracy.

[0042] The first candidate point set mentioned above is used to describe a set of all points in the point cloud map whose distances from the fitting plane are within a preset range.

[0043] The first candidate point mentioned above is used to describe a single point in the first candidate point set. The first candidate point represents a point that meets a preset distance range and is the basis for further quality inspection.

[0044] The second candidate point set is used to describe a set of points further selected from the first candidate point set based on the same semantic information and preset accuracy.

[0045] The second candidate point mentioned above is used to describe a single point in the second candidate point set. The second candidate point represents a point that has passed the semantic information quality inspection and the accuracy quality inspection, and is a candidate for the final target point set.

[0046] The above-mentioned marking device is used to describe the device deployed within the target range, which is used to provide location coordinates and help determine the accuracy of points related to the device in the point cloud. The types of marking devices may include but are not limited to dotting devices and other devices that can provide location information. The specific marking device needs to be determined according to actual needs and is not limited here.

[0047] The above-mentioned device point is used to describe the point in the second candidate point set corresponding to the marked device.

[0048] The above target point set is used to describe the points that meet the preset distance range and preset accuracy map selected by the quality inspection step. The role of the target point set can include but is not limited to providing high-quality point cloud data for subsequent analysis, modeling or other applications.

[0049] In an optional embodiment, the accuracy and reliability of the point cloud map are improved by a quality inspection method for the target point set in the point cloud map. The method first detects the quality of each point and selects a first candidate point set whose distance from the fitting plane is within a preset distance range; then, based on the semantic information of the first candidate point, further selects a second candidate point set that meets the preset accuracy; finally, combined with the position coordinates of multiple marking devices within the target range, the device points in the second candidate point set are quality inspected to finally obtain the target point set. The above process effectively improves the accuracy of the point cloud map, reduces errors, and provides more reliable data support for subsequent applications.

[0050] Optionally, the quality of each point in the current point cloud map is detected to obtain a first candidate point set in the current point cloud map, including: rasterizing the current point cloud map to obtain multiple grids; determining the roughness of the grid based on the point located at the center of each grid, wherein the roughness is used to characterize the distance between the point located at the center of the grid and the fitting plane; determining at least one target grid among the multiple grids based on the roughness of the multiple grids, wherein the roughness of each target grid is less than a first preset roughness; and adding the point corresponding to the at least one target grid to the first candidate point set.

[0051] The above-mentioned rasterization process is used to describe a method of discretizing continuous point cloud data, that is, dividing the point cloud space into a number of small grids, each grid containing a certain number of points. The types of rasterization processes may include but are not limited to regular grid-based rasterization, octree-based rasterization, etc. The role of rasterization processes may include but are not limited to simplifying point cloud data to facilitate subsequent calculations and analysis, such as roughness calculation, target grid determination, etc.

[0052] The above-mentioned target grid is used to describe a grid whose roughness is less than the first preset roughness after rasterization. The types of target grids may include but are not limited to target grids based on roughness, target grids based on shape features, etc. The specific target grid needs to be determined according to actual needs and is not limited here. The role of the target grid may include but is not limited to screening out grids with specific features from a large number of grids for further processing, such as feature extraction, point cloud classification, etc.

[0053] The first preset roughness mentioned above is used to describe a preset threshold value, which is used to measure the distance between the center point of the grid and the fitting plane, so as to evaluate the roughness of the grid. The first preset roughness may include but is not limited to a fixed first preset roughness, a dynamic first preset roughness, etc. The specific first preset roughness needs to be determined according to actual needs and is not limited here. The size of the first preset roughness may include but is not limited to 0.2 meters, 0.4 meters, 0.6 meters, 0.8 meters, etc. The specific first preset roughness needs to be determined according to the application scenario and accuracy requirements and is not limited here. The role of the first preset roughness may include but is not limited to providing a standard for distinguishing between grids with lower roughness and grids with higher roughness, so as to determine the target grid.

[0054] In an optional embodiment, the point cloud map is first rasterized to divide the map into multiple grid units; then, the roughness of each grid is evaluated by calculating the distance between the center point of each grid and the fitting plane; then, the system selects those grids whose roughness is less than the first preset roughness as target grids based on the first preset roughness; finally, the points corresponding to these target grids are included in the first candidate point set to determine the points that meet the requirements and use them for subsequent map construction or analysis. The above process helps to screen out representative and high-precision points from a large amount of point cloud data, thereby improving the quality and usability of the map.

[0055] Optionally, based on the point located at the center of each grid, the roughness of the grid is determined, including: determining the maximum distance and the minimum distance between the point located at the center of each grid and the fitting plane; obtaining the difference between the maximum distance and the minimum distance to obtain the roughness of the grid.

[0056] In an optional embodiment, the maximum and minimum distances between each grid center point and the fitting plane are calculated, and then the difference between the two distances is obtained to determine the roughness of the grid. The above method can quantify the undulation of the grid surface, providing important reference data for fields such as terrain analysis, geological exploration and land planning. At the same time, by measuring and comparing these distances, the complexity and variability of the terrain can be effectively evaluated.

[0057] Optionally, based on the roughness of multiple grids, at least one target grid among the multiple grids is determined, including: based on the roughness of the multiple grids, at least one candidate grid among the multiple grids is determined, wherein the roughness of each candidate grid is less than a first preset roughness; obtaining an average value of the roughness of at least one candidate grid to obtain a first roughness mean; when the first roughness mean is less than a preset mean and the proportion of at least one candidate grid in the multiple grids is greater than the first preset proportion, determining at least one candidate grid as at least one target grid.

[0058] The candidate grids are used to describe grids whose roughness is less than the first preset roughness among multiple grids. The candidate grids represent grids whose roughness meets a specific roughness condition. Determining the candidate grids is helpful to narrow the search range and is the basis for determining the target grid.

[0059] The first roughness mean value is used to describe the average value of the roughness of all candidate grids. The first roughness mean value can be used to compare with a preset mean value to determine whether to determine the candidate grid as the target grid.

[0060] The above-mentioned preset mean is used to describe a preset value for comparison with the first roughness mean. The preset mean can be used as a threshold to help determine whether the candidate grid is smooth enough to be determined as the target grid. The preset mean can be any value such as 0.1 meter, 0.2 meter, 0.3 meter, 1 meter, etc. The specific preset mean needs to be determined according to the application scenario and the required roughness level, and is not limited here.

[0061] The above-mentioned first preset proportion is used to describe the proportion of the preset candidate grids in all grids. The first preset proportion is a percentage value, which may include but is not limited to 10%, 20%, 30%, etc. The specific first preset proportion needs to be determined according to actual needs and is not limited here. The first preset proportion can be used to ensure that the candidate grids occupy a sufficient proportion of all grids.

[0062] In an optional embodiment, by evaluating the roughness of multiple grids, candidate grids with roughness lower than a preset threshold are screened out, and the first roughness mean of these candidate grids is calculated. When the first roughness mean is lower than the preset mean, and the proportion of the candidate grids in all grids exceeds the preset proportion, these candidate grids will be determined as target grids. The above process effectively improves the accuracy of target grid selection, helps to more accurately identify and process target objects, thereby improving the performance and efficiency of the system.

[0063] Optionally, the method also includes: performing voxel filtering on the points in the current point cloud map to obtain a first filter point set; performing discrete point filtering on the first filter point set to obtain a second filter point set; performing rasterization processing on the current point cloud map to obtain multiple grids, including: performing rasterization processing on the second filter point set to obtain multiple grids.

[0064] The above-mentioned voxel filtering is used to describe a method for downsampling point cloud data. Voxel filtering divides the point cloud data into three-dimensional voxels and then processes the points in each voxel to reduce the amount of data and reduce the computational complexity. The effects of voxel filtering can include but are not limited to eliminating duplicate points or point clouds with close distances, removing redundancy, simplifying data, retaining features, etc.

[0065] The first filter point set mentioned above is used to describe a point cloud data set after voxelization filtering processing.

[0066] The above-mentioned discrete point filtering is used to describe a method for further processing point cloud data after voxel filtering. The function of discrete point filtering may include but is not limited to removing points whose position coordinates are too discrete.

[0067] The second filtering point set mentioned above is used to describe the point cloud data set after discrete point filtering.

[0068] In an optional embodiment, the points in the current point cloud map are first voxelized and filtered to convert the point cloud data into a first filter point set that is easier to process; then the first filter point set is discretely filtered to obtain a more accurate second filter point set; finally, the point cloud data is converted into multiple grids by rasterizing these finely processed point sets. The above process helps to simplify the representation of the map, facilitates subsequent map analysis and application, and at the same time improves the accuracy and practicality of the map data, providing strong data support for various geographic information systems and autonomous driving applications.

[0069] Optionally, based on the semantic information of each first candidate point in the first candidate point set, quality inspection is performed on multiple first candidate points with the same semantic information to obtain a second candidate point set in the first candidate point set, including: based on the semantic information of each first candidate point in the first candidate point set, determining at least one semantic set, wherein different semantic sets are used to characterize sets consisting of first candidate points with different semantic information; based on each first candidate point in each semantic set, determining the roughness of the semantic set; based on the roughness of at least one semantic set, determining a target semantic set in at least one semantic set, wherein the roughness of the target semantic set is less than a second preset roughness; and adding the first candidate points in the target semantic set to the second candidate point set.

[0070] The above-mentioned semantic set is used to describe a set that groups first candidate points with the same or similar semantic information based on the semantic information of each first candidate point in the first candidate point set. The function of the semantic set may include but is not limited to classifying the candidate points according to their semantic information for more refined processing and analysis.

[0071] The above-mentioned target semantic set is used to describe the semantic set selected according to specific criteria among all semantic sets, where the specific criteria may include but are not limited to roughness, features, etc. The role of the target semantic set may include but is not limited to further narrowing the scope of candidate points and improving the efficiency and accuracy of the method.

[0072] The above-mentioned second preset roughness is used to describe a preset numerical value, which is used to measure and evaluate the roughness of the points in the first candidate point set. The types of the second preset roughness may include but are not limited to fixed preset roughness, dynamic preset roughness, etc. The specific second preset roughness needs to be determined according to the application scenario and actual needs, and is not limited here. The size of the second preset roughness may include but is not limited to 0.2 meters, 0.4 meters, 0.6 meters, 0.8 meters, etc. The specific second preset roughness needs to be determined according to the application scenario and accuracy requirements, and is not limited here. The role of the second preset roughness may include but is not limited to screening candidate points, performing quality control on candidate points, etc.

[0073] In an optional embodiment, first, multiple semantic sets are determined based on the semantic information of each first candidate point, each set represents a different semantic information; then, the candidate points in each semantic set are subjected to roughness evaluation to determine the roughness of the set. Then, by comparing the roughness of each semantic set with the second preset roughness, target semantic sets with a roughness lower than the second preset roughness are identified; finally, the first candidate points in these target semantic sets are included in the second candidate point set. The above process effectively selects candidate points with higher quality and more accurate semantic information from a large number of candidate points through semantic information and roughness evaluation, thereby improving the accuracy and efficiency of subsequent processing.

[0074] Figure 2 is a graph of abnormal point detection results of a point cloud map according to an embodiment of the present invention, such as Figure 2 As shown, Figure 2 The medium gray area represents the range of the current point cloud map, and the white points represent the detected anomalies.

[0075] Figure 3 is another abnormal point detection result diagram of a point cloud map according to an embodiment of the present invention, such as Figure 3 As shown, Figure 3 The middle gray area represents the range of the current point cloud map, and the white points represent the detected anomalies.

[0076] Optionally, the method further includes: segmenting each first candidate point in the first candidate point set to determine semantic information of the first candidate point.

[0077] The above-mentioned segmentation method may include but is not limited to the region growing method, the random sampling continuous allocation method, etc. By segmenting the first candidate point by the above-mentioned segmentation method, multiple types of point clouds such as ground point cloud, static retaining wall point cloud, and dynamic target point cloud can be obtained.

[0078] In an optional embodiment, the semantic information of these points is determined by segmenting each first candidate point in the first candidate point set. The above process can more accurately identify and understand the semantic content of the candidate points through segmentation and analysis, thereby improving the accuracy and efficiency of data processing. The above method is suitable for scenarios where semantic analysis of large amounts of data is required, such as natural language processing, image recognition and other fields. Segmenting the data helps to improve the system's ability to understand and process the data.

[0079] Optionally, based on the position coordinates of multiple marking devices deployed within the target range, quality inspection is performed on device points in the second candidate point set corresponding to the multiple marking devices to obtain a target point set, including: based on the position coordinates of each marking device, searching in the second candidate point set to determine the device point in the second candidate point set corresponding to the marking device, wherein the distance between the device point and the position coordinates is less than the distance between other second candidate points and the position coordinates, and the other second candidate points are used to represent any second candidate points in the second candidate point set except the device point; obtaining the difference between the elevation of the marking device and the elevation of the device point to obtain the elevation difference; based on the elevation difference, determining the target device point among the device points corresponding to the multiple marking devices, wherein the elevation difference corresponding to the target device point is greater than an elevation difference threshold; when the proportion of the target device point in the device points corresponding to the multiple marking devices is less than a second preset proportion, determining the second candidate point set as the target point set; preferably, the elevation difference threshold is a threshold determined based on the semantic information of the target device point.

[0080] The above-mentioned elevation is used to describe the height of a point relative to a reference datum. The types of elevation may include but are not limited to absolute elevation, relative elevation, etc. The specific elevation should be determined according to actual needs and is not limited here. The role of elevation may include but is not limited to map making, navigation, geological exploration, etc. The elevation can be obtained through ground measurement, satellite remote sensing technology, aerial photogrammetry, etc.

[0081] The above elevation difference is used to describe the vertical distance difference between two points, that is, the vertical distance difference between the elevation of the marking device and the elevation of the corresponding device point.

[0082] The above-mentioned elevation difference threshold is used to describe the pre-set elevation difference value. The elevation difference threshold can be determined according to the semantic information of the target device point. The elevation difference threshold can be used to screen out device points whose elevation difference meets the requirements and use them as target device points. The elevation difference threshold may include but is not limited to 0.1 meters, 0.3 meters, 0.6 meters, 0.9 meters, etc. The specific elevation difference threshold needs to be determined according to the application scenario and accuracy requirements, and is not limited here.

[0083] The above-mentioned target device point is used to describe the device point whose elevation difference is greater than the elevation difference threshold. The types of target device points may include but are not limited to marking device points, sensor device points, etc. The specific target device point needs to be determined based on the device type and the elevation difference threshold, which is not limited here.

[0084] The above-mentioned second preset proportion is used to describe the proportion of the target device points in the device points corresponding to multiple marking devices. The second preset proportion is a percentage value, which may include but is not limited to 10%, 20%, 30%, etc. The specific first preset proportion needs to be determined according to actual needs and is not limited here. The first preset proportion can be used to ensure that the target device points occupy a sufficient proportion of all device points, thereby ensuring the quality of the device points in the target point set.

[0085] In an optional embodiment, the device point closest to the position coordinates of each marked device is first searched in the second candidate point set, and then the difference between these device points and the elevation of the marked device, i.e., the elevation difference, is calculated; then the target device point is screened out by comparing the elevation difference with a preset elevation difference threshold. If the proportion of the target device point in the device points corresponding to all marked devices is lower than the second preset proportion, the second candidate point set will be regarded as the target point set. In addition, the elevation difference threshold can be adjusted according to the semantic information of the target device point to improve the accuracy of the screening. The above method helps to improve the quality of the target point set and ensure that the device points it contains match the position and elevation information of the marked device.

[0086] Optionally, based on the semantic information of each target point in the target point set, the outliers in the target point set are clustered to obtain a clustering result, including: based on the semantic information of each target point, determining a third preset roughness corresponding to the target point; extracting target points whose roughness is greater than the third preset roughness from the target point set to obtain outliers; clustering the outliers to obtain at least one first cluster set, wherein different first cluster sets are used to characterize different areas; and determining at least one first cluster set as a clustering result.

[0087] The third preset roughness mentioned above is used to describe a threshold value pre-set based on the semantic information of the target point. The function of the third preset roughness may include but is not limited to measuring the roughness of the target point and screening out target points greater than the third preset roughness. The third preset roughness may include but is not limited to 0.2 meters, 0.4 meters, 0.6 meters, 0.8 meters, etc. The specific third preset roughness needs to be determined according to the application scenario and accuracy requirements, and is not limited here.

[0088] The first cluster set mentioned above is used to describe a set of target points obtained after cluster analysis of outliers. The role of the first cluster set may include but is not limited to characterizing different regions or features to help identify and understand patterns and structures in data.

[0089] In an optional embodiment, first determine the third preset roughness corresponding to each target point, then select the points whose roughness exceeds the preset value from the target point set, and regard these points as abnormal points; then, perform cluster analysis on these abnormal points and form at least one first cluster set, each set representing a different area; finally, determine at least one such cluster set as a clustering result. The above process can identify and classify abnormal points, provide structured abnormal data for further analysis and processing, and help improve data availability and analysis accuracy.

[0090] Figure 4 is a rough area detection result diagram of a point cloud map according to an embodiment of the present invention, such as Figure 4 As shown, Figure 4 The light gray area in the middle rectangle represents the ground point cloud of the calibration site, the white polygon on the light gray represents the rough ground area, the dark gray area outside the light gray rectangle represents the retaining wall point cloud of the calibration site, and the gray polygon on the dark gray area represents the rough area of ​​the wall. The white polygon and the gray polygon are obtained by solving the convex hull of the cluster set.

[0091] Optionally, the method also includes: when the number of at least one abnormal area is greater than a preset number, or the total area of ​​at least one abnormal area accounts for a proportion of the area of ​​the target range that is greater than a third preset proportion, outputting a prompt message, wherein the prompt message is used to prompt trimming of the target range.

[0092] The above-mentioned preset number is used to describe the maximum number of abnormal areas within the target range that is preset. The preset number may include but is not limited to 5, 10, 15, 20, etc. The specific preset number needs to be set based on historical data, experience or specific business logic, and is not limited here. The role of the preset number may include but is not limited to identifying whether the number of abnormal areas exceeds the normal range. If the number of abnormal areas exceeds this preset value, the system will consider that there may be a problem and further inspection or adjustment is required.

[0093] The third preset ratio mentioned above is used to describe the percentage threshold of the total area of ​​the preset abnormal area to the total area of ​​the target range. The third preset ratio may include but is not limited to 10%, 20%, 30%, etc. The specific third preset ratio needs to be set based on historical data, experience or business needs, and is not limited here. The role of the third preset ratio may include but is not limited to evaluating whether the area of ​​the abnormal area occupies an unreasonable proportion of the target range, and determining whether the system needs to trim the target range based on the evaluation results.

[0094] In an optional embodiment, when the number of detected abnormal areas exceeds a preset number, or the total area of ​​these abnormal areas accounts for a proportion of the target range area that exceeds a third preset proportion, the system will automatically output a prompt message. These prompt messages are intended to remind users or operators that the target range may require further trimming or adjustment to ensure the accuracy and effectiveness of the analysis or processing. The above process helps to improve the quality and efficiency of data processing, especially in scenarios where precise control of abnormal data is required. Through automated anomaly detection and prompt mechanisms, users can identify and handle potential problem areas more quickly, thereby making timely adjustments.

[0095] Optionally, the method also includes: obtaining a historical point cloud map of the target range; matching historical points in the historical point cloud map with current points in the current point cloud map to determine at least one displacement point in the historical point cloud map, wherein the at least one displacement point is used to characterize a historical point that is displaced from the current point in the current point cloud map; clustering the at least one displacement point to obtain at least one second cluster set, wherein different second cluster sets are used to characterize different areas.

[0096] The above-mentioned historical point cloud map is used to describe a collection of three-dimensional spatial data acquired before a specific point in time. The types of historical point cloud maps may include but are not limited to static point cloud maps, dynamic point cloud maps, etc. The specific historical point cloud map needs to be determined according to the application scenario and is not limited here. The role of the historical point cloud map may include but is not limited to being used as a reference for the current point cloud map to identify environmental changes.

[0097] The above historical points are used to describe a single three-dimensional spatial coordinate point recorded in the historical point cloud map.

[0098] The above-mentioned current point is used to describe a single three-dimensional spatial coordinate point recorded in the current point cloud map.

[0099] The above-mentioned displacement points are used to describe points whose positions in the historical point cloud map do not match the positions in the current point cloud map. The role of the displacement points may include but is not limited to exploring the movement or changes of the displacement points between two time points, identifying and tracking changes in the environment.

[0100] The above-mentioned second clustering set is used to describe the set obtained by grouping the displacement points through the clustering algorithm. The types of the second clustering set may include but are not limited to the second clustering set based on the displacement speed and the second clustering set based on the displacement distance. The specific second clustering set needs to be determined according to actual needs and is not limited here. The role of the second clustering set may include but is not limited to further analyzing and understanding changes in the environment.

[0101] In an optional embodiment, a historical point cloud map of the target area is first obtained; then the historical points are compared with the current points in the current point cloud map through matching technology to identify the historical points that have been displaced and mark them as displaced points; then these displaced points are clustered to form different second cluster sets, each of which represents a specific change area. The above process can not only reveal the dynamic changes of the environment, but also provide valuable spatial information for further analysis and decision-making.

[0102] Optionally, historical points in the historical point cloud map are matched with current points in the current point cloud map to determine at least one displacement point in the historical point cloud map, including: determining the degree of overlap between a boundary area of ​​the historical point cloud map and a boundary area of ​​the current point cloud map; when the degree of overlap is greater than a preset degree of overlap, matching the historical points in the historical point cloud map with the current point in the current point cloud map to determine at least one displacement point.

[0103] The above-mentioned overlap is used to describe the degree of matching between the historical point cloud map and the corresponding points in the current point cloud map in terms of spatial position. The overlap can be calculated by comparing the positional relationship of the same or similar feature points in the two maps. The role of the overlap can include but is not limited to evaluating the similarity between the two point cloud maps, so as to determine whether further matching processing is required.

[0104] The above-mentioned preset overlap is used to describe a pre-set threshold value, which is used to determine whether the matching degree between the two maps is high enough to ensure the accuracy and reliability of the matching. The preset overlap can be any parameter such as 70%, 75%, 80%, 85%, 90%, etc. The specific preset overlap needs to be determined according to the application scenario and accuracy requirements, and is not limited here.

[0105] In an optional embodiment, the matching process between historical points and current points is optimized by evaluating the overlap between the boundary area of ​​the historical point cloud map and the current point cloud map. When it is detected that the overlap between the two exceeds the preset overlap, the system will perform an exact match to identify at least one point in the historical point cloud map that has been displaced. The above process improves the accuracy and efficiency of point cloud map matching, and has important practical application value for map updates and navigation systems in dynamic environments.

[0106] Optionally, historical points in the historical point cloud map are matched with current points in the current point cloud map to determine at least one displacement point, including: matching each historical point in the historical point cloud map with each current point in the current point cloud map to determine the matching point corresponding to each historical point, wherein the distance between the matching point and the historical point is less than the distance between other current points in the current point cloud map and the historical point, and the other current points are used to represent any current point in the current point cloud map except the matching point; when the distance between the historical point and the matching point is greater than a preset distance, the historical point is determined to be a displacement point.

[0107] The above-mentioned matching point is used to describe the point in the current point cloud map that has the shortest distance from a certain historical point in the historical point cloud map. The role of the matching point may include but is not limited to helping determine the correspondence between the historical point cloud map and the current point cloud map.

[0108] The above-mentioned preset distance is used to describe a pre-set threshold. The size of the preset distance may include but is not limited to 1 meter, 2 meters, 3 meters, etc. The specific preset distance needs to be determined according to the application scenario and accuracy requirements, and is not limited here. The role of the preset distance may include but is not limited to determining whether the distance between the historical point and the matching point is large enough, and then determining whether the historical point has shifted.

[0109] In an optional embodiment, all historical points in the historical point cloud map are compared with all current points in the current point cloud map to find the closest matching point to the historical point. If the distance between the matching point and the historical point exceeds a preset distance, the historical point is marked as a displacement point, indicating that its position has changed. The above process helps to monitor and analyze changes in the environment or objects over time, and has important application value in the fields of autonomous driving, robot navigation, geographic information systems, etc.

[0110] Figure 5 is a calibration result diagram of a point cloud map according to an embodiment of the present invention, such as Figure 5 As shown, Figure 5 The rectangular area in the middle represents the point cloud map, the light gray above represents the current point cloud map, the dark gray below represents the historical point cloud map, and the white area represents the detected displacement points.

[0111] According to another aspect of an embodiment of the present invention, an embodiment of a point cloud map processing device is also provided. The device can execute the point cloud map processing method provided in the above embodiment. The specific implementation method and preferred application scenario are the same as the above embodiment and will not be repeated here.

[0112] Figure 6 is a schematic diagram of a point cloud map processing device according to an embodiment of the present invention, such as Figure 6As shown, the device includes: a first acquisition module 602, used to acquire the current point cloud map of the target range; a quality inspection module 604, used to perform quality inspection on each point in the current point cloud map to obtain a target point set in the current point cloud map, wherein the distance between each target point in the target point set and the fitting plane is within a preset distance range, and the accuracy of the target point meets the preset accuracy, and the fitting plane is used to characterize the plane obtained by fitting each point in the current point cloud map; a first clustering module 606, used to cluster abnormal points in the target point set based on the semantic information of each target point in the target point set to obtain a clustering result; a first determination module 608, used to determine at least one abnormal area within the target range based on the clustering result, wherein the semantic information is used to characterize the type of target object corresponding to the target point.

[0113] Optionally, the quality inspection module includes: a detection unit, which is used to detect the quality of each point in the current point cloud map to obtain a first candidate point set in the current point cloud map, wherein the distance between each first candidate point in the first candidate point set and the fitting plane is within a preset distance range; a first quality inspection unit, which is used to perform quality inspection on multiple first candidate points with the same semantic information based on the semantic information of each first candidate point in the first candidate point set to obtain a second candidate point set in the first candidate point set, wherein the accuracy of each second candidate point in the second candidate point set meets the preset accuracy; a second quality inspection unit, which is used to perform quality inspection on device points corresponding to multiple marking devices in the second candidate point set based on the position coordinates of multiple marking devices deployed within the target range to obtain a target point set, wherein the accuracy of the device points with the same semantic information in the target point set meets the preset accuracy.

[0114] Optionally, the detection unit includes: a processing subunit, used to rasterize the current point cloud map to obtain multiple grids; a first determination subunit, used to determine the roughness of the grid based on the point located at the center of each grid, wherein the roughness is used to characterize the distance between the point located at the center of the grid and the fitting plane; a second determination subunit, used to determine at least one target grid among the multiple grids based on the roughness of the multiple grids, wherein the roughness of each target grid is less than a first preset roughness; and a first adding subunit, used to add points corresponding to at least one target grid to the first candidate point set.

[0115] Optionally, the first determination subunit includes: determining the maximum distance and the minimum distance between a point located at the center of each grid and the fitting plane; obtaining the difference between the maximum distance and the minimum distance to obtain the roughness of the grid.

[0116] Optionally, the second determination subunit includes: determining at least one candidate grid among the multiple grids based on the roughness of the multiple grids, wherein the roughness of each candidate grid is less than a first preset roughness; obtaining an average value of the roughness of at least one candidate grid to obtain a first roughness mean; when the first roughness mean is less than a preset mean and the proportion of at least one candidate grid in the multiple grids is greater than the first preset proportion, determining at least one candidate grid as at least one target grid.

[0117] Optionally, the device also includes: a first filtering module, used to perform voxel filtering on the points in the current point cloud map to obtain a first filter point set; a second filtering module, used to perform discrete point filtering on the first filter point set to obtain a second filter point set; a processing module, used to perform rasterization processing on the current point cloud map to obtain multiple grids, including: performing rasterization processing on the second filter point set to obtain multiple grids.

[0118] Optionally, the first quality inspection unit includes: a third determination subunit, used to determine at least one semantic set based on the semantic information of each first candidate point in the first candidate point set, wherein different semantic sets are used to represent sets consisting of first candidate points with different semantic information; a fourth determination subunit, used to determine the roughness of the semantic set based on each first candidate point in each semantic set; a fourth determination subunit, used to determine a target semantic set in at least one semantic set based on the roughness of at least one semantic set, wherein the roughness of the target semantic set is less than a second preset roughness; and a second adding subunit, used to add the first candidate point in the target semantic set to the second candidate point set.

[0119] Optionally, the device further includes: segmenting each first candidate point in the first candidate point set to determine semantic information of the first candidate point.

[0120] Optionally, the second quality inspection unit includes: a fifth determination subunit, used to search in the second candidate point set based on the position coordinates of each marking device, and determine the device point corresponding to the marking device in the second candidate point set, wherein the distance between the device point and the position coordinates is less than the distance between other second candidate points and the position coordinates, and the other second candidate points are used to represent any second candidate points in the second candidate point set except the device point; an acquisition subunit, used to acquire the difference between the elevation of the marking device and the elevation of the device point, and obtain the elevation difference; a sixth determination subunit, used to determine the target device point among the device points corresponding to multiple marking devices based on the elevation difference, wherein the elevation difference corresponding to the target device point is greater than an elevation difference threshold; a seventh determination subunit, used to determine the second candidate point set as the target point set when the proportion of the target device point in the device points corresponding to multiple marking devices is less than a second preset proportion; preferably, the elevation difference threshold is a threshold determined based on the semantic information of the target device point.

[0121] Optionally, the first clustering module includes: a first determination unit, used to determine the third preset roughness corresponding to each target point based on the semantic information of the target point; an extraction unit, used to extract target points whose roughness is greater than the third preset roughness from the target point set to obtain abnormal points; a clustering unit, used to cluster the abnormal points to obtain at least one first cluster set, wherein different first cluster sets are used to characterize different areas; and a second determination unit, used to determine at least one first cluster set as a clustering result.

[0122] Optionally, the device also includes: when the number of at least one abnormal area is greater than a preset number, or the total area of ​​at least one abnormal area accounts for a proportion of the area of ​​the target range that is greater than a third preset proportion, outputting a prompt message, wherein the prompt message is used to prompt the target range to be trimmed.

[0123] Optionally, the device also includes: a second acquisition module, which acquires a historical point cloud map of the target range; a second determination module, which matches historical points in the historical point cloud map with current points in the current point cloud map, and determines at least one displacement point in the historical point cloud map, wherein the at least one displacement point is used to characterize a historical point that is displaced from the current point in the current point cloud map; a second clustering module, which clusters the at least one displacement point to obtain at least one second cluster set, wherein different second cluster sets are used to characterize different areas.

[0124] Optionally, the second acquisition module includes: a third determination unit, used to determine the degree of overlap between the boundary area of ​​the historical point cloud map and the boundary area of ​​the current point cloud map; a fourth determination unit, used to match the historical points in the historical point cloud map with the current points in the current point cloud map when the degree of overlap is greater than a preset degree of overlap, and determine at least one displacement point.

[0125] Optionally, the fourth determination unit includes: matching each historical point in the historical point cloud map with each current point in the current point cloud map, and determining the matching point corresponding to each historical point, wherein the distance between the matching point and the historical point is smaller than the distance between other current points in the current point cloud map and the historical point, and the other current points are used to represent any current point in the current point cloud map except the matching point; when the distance between the historical point and the matching point is greater than a preset distance, the historical point is determined to be a displacement point.

[0126] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, comprising: a memory storing an executable program; and a processor for running the program, wherein the method in each embodiment of the present invention is executed when the program is running.

[0127] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the methods in various embodiments of the present invention.

[0128] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0129] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium is used to store a computer program, and when the computer program is executed by a processor, the method in each embodiment of the present invention is implemented.

[0130] According to another aspect of an embodiment of the present invention, a computer program is further provided. When the computer program is executed by a processor, the methods in the above-mentioned embodiments of the present invention are implemented.

[0131] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0132] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0133] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0134] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0135] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0136] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for processing a point cloud map, characterized in that: include: Get the current point cloud map of the target range; Performing quality inspection on each point in the current point cloud map to obtain a target point set in the current point cloud map, wherein the distance between each target point in the target point set and the fitting plane is within a preset distance range, and the accuracy of the target point meets the preset accuracy, and the fitting plane is used to represent the plane obtained by fitting each point in the current point cloud map; Based on the semantic information of each target point in the target point set, clustering the abnormal points in the target point set to obtain a clustering result; At least one abnormal area within the target range is determined based on the clustering result, wherein the semantic information is used to characterize the type of the target object corresponding to the target point.

2. The method according to claim 1, characterized in that The step of performing quality inspection on each point in the current point cloud map to obtain a set of target points in the current point cloud map includes: Detecting the quality of each point in the current point cloud map to obtain a first candidate point set in the current point cloud map, wherein the distance between each first candidate point in the first candidate point set and the fitting plane is within the preset distance range; Based on the semantic information of each first candidate point in the first candidate point set, quality inspection is performed on multiple first candidate points with the same semantic information to obtain a second candidate point set in the first candidate point set, wherein the accuracy of each second candidate point in the second candidate point set meets the preset accuracy; Based on the position coordinates of multiple marking devices deployed within the target range, quality inspection is performed on the device points in the second candidate point set corresponding to the multiple marking devices to obtain the target point set, wherein the accuracy of the device points with the same semantic information in the target point set meets the preset accuracy.

3. The method according to claim 2, characterized in that The detecting the quality of each point in the current point cloud map to obtain a first candidate point set in the current point cloud map includes: Performing rasterization processing on the current point cloud map to obtain a plurality of grids; Determine the roughness of the grid based on the point located at the center of each grid, wherein the roughness is used to characterize the distance between the point located at the center of the grid and the fitting plane; Based on the roughness of the plurality of grids, determining at least one target grid among the plurality of grids, wherein the roughness of each target grid is less than a first preset roughness; The points corresponding to the at least one target grid are added to the first candidate point set.

4. The method according to claim 3, characterized in that The method of determining the roughness of each grid based on a point located at the center of each grid comprises: Determine the maximum distance and the minimum distance between a point located at the center of each grid and the fitting plane; The difference between the maximum distance and the minimum distance is obtained to obtain the roughness of the grid.

5. The method according to claim 3, characterized in that: The determining, based on the roughness of the plurality of grids, at least one target grid among the plurality of grids comprises: Determine at least one candidate grid among the plurality of grids based on the roughness of the plurality of grids, wherein the roughness of each candidate grid is less than the first preset roughness; Obtaining an average value of the roughness of the at least one candidate grid to obtain a first roughness average value; When the first roughness mean is smaller than a preset mean and the proportion of the at least one candidate grid in the plurality of grids is greater than a first preset proportion, the at least one candidate grid is determined to be the at least one target grid.

6. The method according to claim 3, characterized in that The method further comprises: Performing voxel filtering on the points in the current point cloud map to obtain a first filtered point set; Performing discrete point filtering on the first filter point set to obtain a second filter point set; The rasterizing the current point cloud map to obtain a plurality of raster blocks includes: rasterizing the second filter point set to obtain the plurality of raster blocks.

7. The method according to any one of claims 2 to 6, characterized in that The step of performing quality inspection on a plurality of first candidate points with the same semantic information based on the semantic information of each first candidate point in the first candidate point set to obtain a second candidate point set in the first candidate point set includes: Based on the semantic information of each first candidate point in the first candidate point set, determining at least one semantic set, wherein different semantic sets are used to represent sets consisting of first candidate points with different semantic information; Based on each first candidate point in each semantic set, determining the roughness of the semantic set; Determining a target semantic set in the at least one semantic set based on the roughness of the at least one semantic set, wherein the roughness of the target semantic set is less than a second preset roughness; The first candidate point in the target semantic set is added to the second candidate point set.

8. The method according to claim 7, characterized in that The method further comprises: Each first candidate point in the first candidate point set is segmented to determine semantic information of the first candidate point.

9. The method according to any one of claims 2 to 8, characterized in that The step of performing quality inspection on device points corresponding to the multiple marking devices in the second candidate point set based on the position coordinates of the multiple marking devices deployed within the target range to obtain the target point set includes: Based on the position coordinates of each marking device, searching in the second candidate point set to determine a device point in the second candidate point set corresponding to the marking device, wherein a distance between the device point and the position coordinates is less than a distance between other second candidate points and the position coordinates, and the other second candidate points are used to represent any second candidate point in the second candidate point set except the device point; Obtain the difference between the elevation of the marking device and the elevation of the device point to obtain the elevation difference; Based on the elevation difference, determining a target device point among the device points corresponding to the multiple marking devices, wherein the elevation difference corresponding to the target device point is greater than an elevation difference threshold; When the proportion of the number of target device points in the device points corresponding to the multiple marking devices is less than a second preset proportion, determining the second candidate point set as the target point set; Preferably, the elevation difference threshold is a threshold determined based on semantic information of the target device point.

10. The method according to any one of claims 1 to 9, characterized in that The clustering of abnormal points in the target point set based on the semantic information of each target point in the target point set to obtain a clustering result includes: Based on the semantic information of each target point, determining a third preset roughness corresponding to the target point; Extracting a target point whose roughness is greater than the third preset roughness from the target point set to obtain the abnormal point; Clustering the outliers to obtain at least one first cluster set, wherein different first cluster sets are used to represent different areas; The at least one first clustering set is determined as the clustering result.

11. The method according to any one of claims 1 to 10, characterized in that The method further comprises: When the number of the at least one abnormal area is greater than a preset number, or the total area of ​​the at least one abnormal area accounts for a proportion of the area of ​​the target range that is greater than a third preset proportion, a prompt message is output, wherein the prompt message is used to prompt the target range to be trimmed.

12. The method according to any one of claims 1 to 10, characterized in that The method further comprises: Obtaining a historical point cloud map of the target range; Matching the historical points in the historical point cloud map with the current point in the current point cloud map to determine at least one displacement point in the historical point cloud map, wherein the at least one displacement point is used to represent the historical point that is displaced from the current point in the current point cloud map; The at least one displacement point is clustered to obtain at least one second cluster set, wherein different second cluster sets are used to characterize different regions.

13. The method according to claim 12, characterized in that The matching of the historical points in the historical point cloud map with the current points in the current point cloud map to determine at least one displacement point in the historical point cloud map includes: Determining the degree of overlap between the boundary area of ​​the historical point cloud map and the boundary area of ​​the current point cloud map; When the degree of overlap is greater than a preset degree of overlap, the historical points in the historical point cloud map are matched with the current points in the current point cloud map to determine the at least one displacement point.

14. The method according to claim 13, characterized in that The matching of the historical point in the historical point cloud map with the current point in the current point cloud map to determine the at least one displacement point includes: Matching each historical point in the historical point cloud map with each current point in the current point cloud map, and determining a matching point corresponding to each historical point, wherein a distance between the matching point and the historical point is less than a distance between other current points in the current point cloud map and the historical point, and the other current points are used to represent any current point in the current point cloud map except the matching point; When the distance between the historical point and the matching point is greater than a preset distance, the historical point is determined to be a displacement point.

15. A point cloud map processing device, characterized in that: include: A first acquisition module is used to acquire a current point cloud map of a target range; A quality inspection module, used for performing quality inspection on each point in the current point cloud map to obtain a target point set in the current point cloud map, wherein the distance between each target point in the target point set and the fitting plane is within a preset distance range, and the accuracy of the target point meets the preset accuracy, and the fitting plane is used to characterize the plane obtained by fitting each point in the current point cloud map; A first clustering module, used for clustering abnormal points in the target point set based on semantic information of each target point in the target point set to obtain a clustering result; The first determination module is used to determine at least one abnormal area within the target range based on the clustering result, wherein the semantic information is used to characterize the type of the target object corresponding to the target point.

Images