Vehicle-mounted laser point cloud data repeatability accuracy evaluation method and device

Abstract

The present application relates to the technical field of point cloud data processing, in particular to a kind of vehicle-mounted laser point cloud data repeatability precision evaluation method and device, by obtaining vehicle-mounted trajectory data and vehicle-mounted laser point cloud data;According to vehicle-mounted trajectory data, vehicle-mounted laser point cloud data is divided into multiple segmented point clouds, according to segmented point cloud, obtain point cloud query feature, then according to point cloud query feature, repeatability precision evaluation is obtained.In this way, from the angle of repeatability precision evaluation, the area with repeated point cloud is evaluated for repeatability precision, and any collected point cloud data can be evaluated without deliberately laying out control points by hand.Avoid manual control point layout work, thereby solving the technical problems of high work intensity, long time consumption and low efficiency in the precision detection process of the prior art, improving the evaluation efficiency of point cloud precision, and providing a reference for point cloud quality evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of point cloud data processing, and particularly relates to a vehicle-mounted laser point cloud data repeatability precision evaluation method and device. BACKGROUND

[0002] The vehicle-mounted mobile measurement system can efficiently and densely obtain high-precision three-dimensional point cloud and image texture information of the road environment, is a new high-resolution earth observation technology, and can significantly improve the efficiency of three-dimensional spatial information acquisition. However, in the actual engineering application of the vehicle-mounted mobile measurement system, the comprehensive influence of the positioning error of the global positioning system, the attitude determination error of the inertial navigation system, the ranging and angle measurement error of the scanner, the sensor time synchronization and calibration error, and the data solving error, the point cloud data precision is difficult to reach the requirement of centimeter level, and even reaches decimeter level or meter level in the signal loss area. There is also a non-rigid deformation between the revisit point clouds of the same area at different times, which may lead to low reliability of the point cloud data. Therefore, the precision of the collected point cloud data needs to be evaluated to screen out reliable point cloud data.

[0003] Some scholars have studied the evaluation method of point cloud precision. In 2019, Xu Shuzhi et al. proposed to use an outdoor three-dimensional detection field, with the help of self-made reflective paper, spherical targets and feature objects, to obtain high-precision three-dimensional coordinates of feature points, and to compare the obtained coordinates with the point cloud solved by the system to evaluate the precision of the point cloud. In 2021, Huang Changdi et al. obtained feature point coordinates by setting targets on control points and measuring terrain feature point coordinates by high-precision total station or three-dimensional scanner, and compared the feature point coordinates with the point cloud results to evaluate the precision of the point cloud. However, the current evaluation of point cloud precision mainly uses manual control point setting, which leads to high work intensity, long time consumption and low efficiency in the precision detection process, and cannot meet the requirements of large-scale application of point cloud data. SUMMARY

[0004] In view of the deficiencies in the prior art, the present application provides a vehicle-mounted laser point cloud data repeatability precision evaluation method and device to solve the technical problems of high work intensity, long time consumption and low efficiency in the precision detection process in the prior art.

[0005] In a first aspect, the present application provides a vehicle-mounted laser point cloud data repeatability precision evaluation method.

[0006] In a first implementation, a vehicle-mounted laser point cloud data repeatability precision evaluation method comprises:

[0007] obtaining vehicle-mounted trajectory data and vehicle-mounted laser point cloud data;

[0008] dividing the vehicle-mounted laser point cloud data into multiple segmented point clouds according to the vehicle-mounted trajectory data;

[0009] Obtain point cloud query features based on each segment of the point cloud;

[0010] Repeatability accuracy evaluation is obtained based on point cloud query features.

[0011] In the second feasible method, in conjunction with the first feasible method, the acquisition of vehicle trajectory data and vehicle laser point cloud data includes:

[0012] During vehicle inspection, an onboard laser scanning system is used to collect point cloud data. The collected point cloud data is identified as onboard laser point cloud data, and the vehicle inspection route is identified as onboard trajectory data.

[0013] In the third feasible method, in conjunction with the second feasible method, the vehicle-mounted laser point cloud data is divided into multiple segmented point clouds based on the vehicle trajectory data, including:

[0014] Obtain the angular velocity of all trajectory points in the vehicle trajectory data;

[0015] Trajectory points with angular velocities greater than a first preset threshold are defined as segmentation points;

[0016] The vehicle-mounted laser point cloud data is divided into multiple segmented point clouds according to each segment point.

[0017] In the fourth possible implementation method, combined with the second implementation method, point cloud query features are obtained based on each segment of the point cloud, including:

[0018] A query cuboid is constructed with the query point as the center and the direction perpendicular to the vehicle trajectory on the horizontal plane as the main direction; the length of the cuboid is determined by the road width and the lane where the trajectory is located.

[0019] The point cloud query features are obtained by sequentially querying each segment of the point cloud using a query cuboid.

[0020] In the fifth possible implementation method, in conjunction with the first feasible method, repeatability accuracy evaluation is obtained based on point cloud query features, including:

[0021] Based on the point cloud query features, a plane fitting is performed to obtain the fitting plane;

[0022] Obtain the elevation accuracy value based on the fitted plane;

[0023] Roadside lines are extracted from the point cloud query features to obtain the roadside lines.

[0024] Obtain the planar accuracy value based on the roadside line;

[0025] Repeatability accuracy is evaluated based on elevation and planar accuracy values.

[0026] Combining the fifth feasible method, the sixth feasible method involves performing plane fitting based on point cloud query features to obtain the fitted plane, including:

[0027] Perform an initial fitting on the point cloud query features to obtain an initial fitting plane;

[0028] Obtain the residuals and mean square error of the initial fitted plane;

[0029] Noise is removed based on each residual and standard error;

[0030] The initial fitted plane after noise removal is fitted a second time to obtain the final fitted plane.

[0031] In conjunction with the fifth feasible method, the seventh feasible method involves obtaining elevation accuracy values ​​based on the fitted plane, including:

[0032] Obtain the first distance value from the query point to each fitted plane;

[0033] Obtain the first distance difference between each first distance value;

[0034] The largest first distance difference is determined as the elevation accuracy value.

[0035] Combining the fifth feasible method, in the eighth feasible method, roadside lines are extracted from the point cloud query features to obtain the roadside lines, including:

[0036] Voxel division is performed on the point cloud query features to obtain multiple voxels;

[0037] Feature extraction is performed on each voxel to obtain voxel features;

[0038] The vehicle-mounted point cloud data is clustered according to voxel features to obtain the ground point cloud;

[0039] Perform a neighborhood query on the ground point cloud and identify voxels in the neighborhood that have the characteristic of elevation as candidate point clouds for the roadside;

[0040] Select roadside points from the candidate point cloud along the road;

[0041] Connect the curb points according to the adjacency relationship between voxels to obtain the curb line.

[0042] In conjunction with the fifth feasible method, the ninth feasible method involves obtaining the planar accuracy value based on the roadside line, including:

[0043] Obtain the second distance value between the query point and each road along the route;

[0044] Sort the second distance values ​​and obtain the second distance difference between each adjacent second distance value;

[0045] The distance between at least two roadside lines that meet the first and second screening conditions is determined as the planar accuracy value; the first screening condition is that the second distance values ​​corresponding to the roadside lines are sequentially adjacent and the difference between the second distances is greater than the second preset threshold; the second screening condition is that the roadside lines originate from different segmented point clouds.

[0046] Secondly, the present invention provides a device for evaluating the repeatability accuracy of vehicle-mounted laser point cloud data.

[0047] In the tenth possible implementation, a vehicle-mounted laser point cloud data repeatability accuracy evaluation device includes:

[0048] The point cloud data acquisition module is configured to acquire vehicle trajectory data and vehicle laser point cloud data;

[0049] The point cloud segmentation module is configured to divide the vehicle-mounted laser point cloud data into multiple segmented point clouds based on the vehicle trajectory data;

[0050] The point cloud query module is configured to obtain point cloud query features based on each segment of the point cloud;

[0051] The repeatability accuracy evaluation module is configured to obtain repeatability accuracy evaluation based on point cloud query features.

[0052] As can be seen from the above technical solution, the beneficial technical effects of the present invention are as follows:

[0053] Multiple segmented point clouds are acquired based on vehicle-mounted laser point cloud data. Point cloud query features are obtained from these segmented point clouds, and then repeatability accuracy is evaluated based on these query features. This approach, focusing on repeatability accuracy evaluation, allows for the assessment of repeatability accuracy in areas with recurring point clouds, regardless of the collected point cloud data, eliminating the need for manually setting up control points. This avoids the need for manual control point setup, thus solving the technical problems of high workload, long processing time, and low efficiency in existing accuracy testing processes. It improves the efficiency of point cloud accuracy evaluation and provides a reference for evaluating point cloud quality. Attached Figure Description

[0054] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0055] Figure 1 This is a schematic diagram illustrating a method for evaluating the repeatability accuracy of vehicle-mounted laser point cloud data provided in this embodiment.

[0056] Figure 2 This is a flowchart for obtaining elevation accuracy values ​​provided in this embodiment;

[0057] Figure 3 This is a flowchart for obtaining plane accuracy values ​​provided in this embodiment;

[0058] Figure 4 This is a schematic diagram of the structure of a vehicle-mounted laser point cloud data repeatability accuracy evaluation device provided in this embodiment. Detailed Implementation

[0059] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0060] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning understood by those skilled in the art. The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure 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 for implementation of the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. Unless otherwise stated, the term "a plurality of" means two or more. In this disclosure, the character " / " indicates an "or" relationship between the preceding and following objects. For example, A / B means: A or B. The term "and / or" describes an association relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or, A and B. The term "corresponding" can refer to an association or binding relationship; A corresponding to B means that there is an association or binding relationship between A and B.

[0061] Combination Figure 1 As shown, this embodiment provides a method for evaluating the repeatability accuracy of vehicle-mounted laser point cloud data, including:

[0062] Step S01: Acquire vehicle trajectory data and vehicle laser point cloud data;

[0063] Step S02: Divide the vehicle-mounted laser point cloud data into multiple segmented point clouds based on the vehicle trajectory data;

[0064] Step S03: Obtain point cloud query features based on each segment of the point cloud;

[0065] Step S04: Obtain repeatability accuracy evaluation based on point cloud query features.

[0066] In some embodiments, this solution evaluates the repeatability accuracy of areas with repetitive point clouds from the perspective of repeatability accuracy evaluation, avoiding the need for manual control point deployment, improving the efficiency of point cloud quality assessment, and providing a reference for point cloud quality evaluation.

[0067] In some embodiments, when satellite signals are strong, the point clouds in repeatedly acquired areas match well in spatial location. However, in areas with weak satellite signals, the spatial location of repeatedly acquired point clouds often deviates. Low repeatability accuracy of point cloud data directly indicates low absolute accuracy. To assess point cloud data accuracy, it is typically necessary to manually measure control points in the survey area, a labor-intensive and inefficient process. Therefore, evaluating the repeatability accuracy of point clouds in repeated areas can indirectly reflect the accuracy of point cloud data, providing a reference for accuracy evaluation. This avoids the need for manual control point deployment and improves the efficiency of accuracy assessment.

[0068] Optionally, acquiring vehicle trajectory data and vehicle laser point cloud data includes: using a vehicle laser scanning system to collect point cloud data during vehicle inspection, identifying the collected point cloud data as vehicle laser point cloud data, and identifying the vehicle inspection route as vehicle trajectory data.

[0069] Optionally, the vehicle-mounted laser point cloud data is divided into multiple segmented point clouds based on the vehicle-mounted trajectory data, including: obtaining the angular velocity of all trajectory points in the vehicle-mounted trajectory data; determining the trajectory points with angular velocities greater than a first preset threshold as segment points; and dividing the vehicle-mounted laser point cloud data into multiple segmented point clouds according to each segment point.

[0070] Optionally, trajectory points with angular velocities greater than a first preset threshold are determined as segmentation points, i.e., angular velocities Q0 = {q j |(φ j -φ j-1 ) / (t j -t j-1 )≥q th}, where φ j Let q be the trajectory point j The heading angle, φ j-1 Let q be the trajectory point j-1 The heading angle, t j and t j-1 They are the trajectory points q j and trajectory point q j-1 Time, q th This is the first preset threshold.

[0071] Optionally, before dividing the vehicle-mounted laser point cloud data into multiple segmented point clouds based on the vehicle-mounted trajectory data, the method further includes: determining whether the sampling frequency of the vehicle-mounted trajectory data is high; if so, performing a thinning operation on it before segmentation; otherwise, segmentation is performed directly.

[0072] Optionally, when the sampling frequency of the vehicle trajectory data is greater than the preset acquisition frequency, it is high-frequency acquisition; otherwise, it is low-frequency acquisition.

[0073] In some embodiments, when an onboard laser scanning system collects point cloud data along a trajectory line, if segmentation is not performed in time at U-turns or turns, the same ground feature may be scanned multiple times within the same segment of point cloud data. This makes it difficult to evaluate the repeatability accuracy of the point cloud data. This solution obtains the angular velocity of all trajectory points in the onboard trajectory data, then selects segmentation points based on the angular velocity, and further divides the point cloud data according to these segmentation points. Finally, the onboard laser point cloud data is divided into different segmented point clouds, ensuring that each segment of point cloud data contains only laser points scanned once for the same ground feature, which is beneficial for evaluating the repeatability accuracy of the point cloud data.

[0074] Optionally, point cloud query features are obtained based on each segment of the point cloud, including: constructing a query cuboid with the query point as the center and the direction perpendicular to the vehicle trajectory on the horizontal plane as the main direction; the length of the cuboid is determined by the road width and the lane where the trajectory is located; and using the query cuboid to sequentially perform point cloud queries on each segment of the point cloud to obtain point cloud query features.

[0075] In some embodiments, the length of the query cuboid is the direction perpendicular to the vehicle trajectory on the horizontal plane. The length value is determined according to the road width and the lane where the trajectory is located. This length value is greater than the distance from the roadside to the vehicle trajectory. The width of the query cuboid is along the trajectory direction, and the height of the query cuboid is in the Z-axis direction. The width and height of the query cuboid can be determined based on empirical values ​​obtained from engineering experiments.

[0076] Optionally, point cloud query features are obtained based on each segment of the point cloud, including: subtracting the height of the acquisition platform from the ground from the elevation of the vehicle-mounted laser point cloud data to make the vehicle trajectory fit the ground, and then constructing a query geometry with the query point as the center; using the query geometry to perform point cloud queries on each segment of the point cloud in sequence to obtain point cloud query features.

[0077] In some embodiments, the vehicle-mounted laser point cloud data records the position and orientation of the acquisition platform center. The elevation of the vehicle-mounted laser point cloud data is subtracted from the height H of the acquisition platform above the ground to make the vehicle trajectory fit the ground. Then, a query geometry is constructed with the query point as the center. The query geometry can be a sphere or a cube. The query geometry is used to sequentially perform point cloud queries on each segment of the point cloud to obtain the point cloud query features.

[0078] Optionally, before obtaining the repeatability accuracy evaluation based on the point cloud query features, the method further includes: if the number of point cloud query features is less than 2, then perform the point cloud query again until the number of point cloud query features is greater than or equal to 2; if the number of point cloud query features is greater than or equal to 2, then perform the repeatability accuracy evaluation.

[0079] In some embodiments, point cloud queries are performed at regular intervals L along the vehicle trajectory data, and the query features of the retrieved point cloud are represented as {S1, S2, ..., S...} n}, where S n This represents the point cloud query feature of the nth segment. If n = 0, it means that no point cloud feature was found at this position; if n = 1, it means that there is no duplicate point cloud at this position. In both cases, the query continues to the next interval distance L; when n ≥ 2, it means that there is a duplicate point cloud at this position.

[0080] Optionally, the repeatability accuracy evaluation is obtained based on the point cloud query features, including: performing plane fitting based on the point cloud query features to obtain a fitted plane; obtaining elevation accuracy values ​​based on the fitted plane; extracting roadside lines from the point cloud query features to obtain roadside lines; obtaining plane accuracy values ​​based on the roadside lines; and obtaining repeatability accuracy evaluation based on the elevation accuracy values ​​and plane accuracy values.

[0081] Optionally, a plane fitting is performed based on the point cloud query features to obtain a fitting plane, including: performing an initial fitting on the point cloud query features to obtain an initial fitting plane; obtaining the residuals and mean square errors of the initial fitting plane; removing noise based on the residuals and mean square errors; and performing a second fitting on the initial fitting plane after noise removal to obtain the final fitting plane.

[0082] In some embodiments, a least squares or random sampling consensus algorithm is used to perform plane fitting, obtain a fitted plane, and statistically analyze the residuals and mean square errors of the initial fitted plane.

[0083] Optionally, noise removal is performed based on each residual and the mean square error, including: removing noise points in the initial fitting plane where the residual is greater than twice the mean square error.

[0084] Optionally, obtaining the elevation accuracy value based on the fitted plane includes: obtaining the first distance value from the query point to each fitted plane; obtaining the first distance difference between each first distance value; and determining the largest first distance difference as the elevation accuracy value.

[0085] In some embodiments, the first distance value {D1,D2,…,D} from the query point to the fitted plane is obtained by calculating the point-to-plane distance. n}, where D nThis represents the first distance value from the query point to the fitted plane corresponding to the nth segment of the point cloud. Subtracting each first distance value yields the first distance difference, and the largest first distance difference is the elevation accuracy value.

[0086] In some embodiments, the steps for obtaining elevation accuracy values ​​are as follows: Figure 2 As shown:

[0087] Step S11: Acquire vehicle-mounted laser point cloud data;

[0088] Step S12: Subtract the height of the acquisition platform from the trajectory elevation of the vehicle-mounted laser point cloud data to make the vehicle trajectory fit the ground.

[0089] Step S13: Calculate the query point of the spacing L;

[0090] Step S14: Construct the query geometry to perform point cloud query and obtain point cloud query features;

[0091] Step S15: Obtain the fitting plane based on the point cloud query features, and calculate the first distance value from the query point to the fitting plane;

[0092] Step S16: Subtract the first distance values ​​to obtain the first distance difference;

[0093] Step S17: Determine the maximum value of the first distance difference as the elevation accuracy value.

[0094] Optionally, roadside line extraction is performed on the point cloud query features to obtain the roadside lines, which includes: dividing the point cloud query features into voxels to obtain multiple voxels; extracting features from each voxel to obtain voxel features; clustering the vehicle point cloud data according to the voxel features to obtain ground point clouds; performing a neighborhood query on the ground point clouds to identify voxels in the neighborhood with the feature of elevation as candidate roadside line point clouds; selecting roadside points from the candidate roadside line point clouds; and connecting the roadside points according to the adjacency relationship between voxels to obtain the roadside lines.

[0095] Optionally, voxel partitioning of point cloud query features includes: partitioning the point cloud into voxels of size i*j*k, where i, j, and k represent the length, width, and height of the voxel, respectively.

[0096] Optionally, feature extraction is performed on each voxel to obtain voxel features. Voxel features include three categories: planar features, elevation features, and other features. The elevation standard deviation and normal vector of each voxel are calculated separately. An elevation difference threshold is set. Voxels with elevation differences less than the threshold and normal vectors approximately parallel to (0, 0, 1) are designated as planar features. Voxels with normal vectors approximately perpendicular to (0, 0, 1) are designated as elevation features. Other voxel features are designated as other features.

[0097] Optionally, the voxel features of the ground point cloud are all planar. The vehicle point cloud data is clustered according to the voxel features of the ground point cloud to filter out the ground point cloud data.

[0098] Optionally, roadside points are selected from candidate point clouds along the roadside, including: selecting the first point cloud on the elevation where the roadside is located based on a preset elevation difference; determining the mean of the first point cloud; determining the coordinates of the plane where the mean is located as the target plane coordinates; determining the average elevation of the voxels of the candidate point cloud as the target elevation coordinates; and the point cloud corresponding to the target plane coordinates and the target elevation coordinates is the roadside point.

[0099] In some embodiments, candidate voxels v0 containing the curb are divided into two-dimensional grids. The number and elevation difference of point clouds in the two-dimensional grids are counted. Point count thresholds and elevation difference thresholds are set. Point clouds in the grids that meet the point count thresholds and elevation difference thresholds are taken as the first point clouds on the elevation of the curb. Their average coordinates are calculated, and the plane coordinates of their average values ​​are taken as the target plane coordinates of the curb points. The average elevation of the candidate voxels v0 is taken as the target elevation coordinates of the curb points. The target plane coordinates and target elevation coordinates are the plane coordinates and elevation coordinates of the curb points.

[0100] In some embodiments, curb points are connected according to the adjacency relationship between voxels, that is, each curb point is connected to the nearest curb point to form a curb line.

[0101] Optionally, obtaining the planar accuracy value based on the roadside includes: obtaining a second distance value between the query point and each roadside; sorting the second distance values ​​and obtaining a second distance difference between adjacent second distance values; determining the distance between at least two roadsides that satisfy the first and second filtering conditions as the planar accuracy value; the first filtering condition is that the second distance values ​​corresponding to the roadsides are sequentially adjacent and the second distance difference is greater than a second preset threshold; the second filtering condition is that the roadsides originate from different segmented point clouds.

[0102] In some embodiments, each segmented point cloud may or may not contain roadside lines. For existing roadside lines, a second distance value from the query point to each roadside line is calculated. If the roadside line is to the right of the acquisition direction, its second distance value is marked as positive; otherwise, it is marked as negative. Roadside lines in the query results corresponding to all segmented point clouds are sorted according to their second distance values. A distance threshold is set, and at least two roadside lines from different segmented point clouds whose difference between adjacent second distance values ​​is less than the distance threshold are selected. If no roadside line meets the condition, the process moves to the next position to continue with planar accuracy evaluation; otherwise, the maximum distance difference between the roadside lines that meet the condition is used as the planar accuracy value.

[0103] In some embodiments, the steps for obtaining the plane accuracy value are as follows: Figure 3 As shown:

[0104] Step S21: Acquire vehicle-mounted laser point cloud data;

[0105] Step S22: Subtract the height of the acquisition platform from the trajectory elevation of the vehicle-mounted laser point cloud data to make the vehicle trajectory fit the ground.

[0106] Step S23: Calculate the query point of the spacing L;

[0107] Step S24: Construct the query geometry to perform point cloud query and obtain point cloud query features;

[0108] Step S25: Fit the roadside line based on the point cloud query features, and calculate the second distance value from the query point to the roadside line;

[0109] Step S26: Sort and filter a group of roadside lines from different segment point clouds according to the second distance value;

[0110] Step S27: Calculate the distance difference between each group of roadside sections;

[0111] Step S28: Determine the maximum distance difference as the plane accuracy value.

[0112] Optionally, repeatability accuracy evaluation can be obtained based on elevation accuracy values ​​and planar accuracy values, including: setting elevation accuracy levels and planar accuracy levels, setting different colors for different accuracy levels, determining the accuracy level to which the elevation accuracy values ​​and planar accuracy values ​​belong as repeatability accuracy evaluation, rendering the vehicle-mounted laser point cloud data into the color corresponding to the accuracy level, and providing an intuitive display of the repeatability accuracy evaluation.

[0113] In some embodiments, different color values ​​are set for different accuracy levels. Based on the elevation and planar accuracy evaluation results, the vehicle-mounted laser point cloud data is rendered according to the color values ​​corresponding to the elevation accuracy and planar accuracy of the corresponding location, so as to realize the intuitive display of the evaluation results, which helps to improve the evaluation efficiency of the repeatability accuracy of point cloud data and provides a reference for the quality evaluation of point cloud data.

[0114] Combination Figure 4 As shown, a repeatability accuracy evaluation device for vehicle-mounted laser point cloud data includes: a point cloud data acquisition module 101, configured to acquire vehicle trajectory data and vehicle-mounted laser point cloud data; a point cloud segmentation module 102, configured to divide the vehicle-mounted laser point cloud data into multiple segmented point clouds based on the vehicle trajectory data; a point cloud query module 103, configured to obtain point cloud query features based on each segmented point cloud; and a repeatability accuracy evaluation module 104, configured to obtain repeatability accuracy evaluation based on the point cloud query features.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for evaluating the repeatability accuracy of vehicle-mounted laser point cloud data, characterized in that, include: Acquire vehicle trajectory data and vehicle-mounted laser point cloud data; Based on the vehicle trajectory data, the vehicle-mounted laser point cloud data is divided into multiple segmented point clouds; Obtain point cloud query features based on each segment of the point cloud; Repeatability accuracy evaluation is obtained based on point cloud query features, including: Based on the point cloud query features, a plane fitting is performed to obtain the fitting plane; The process of obtaining elevation accuracy values ​​based on fitted planes includes: obtaining the first distance value from the query point to each fitted plane; obtaining the first distance difference between each first distance value; and determining the largest first distance difference as the elevation accuracy value. Roadside lines are extracted from the point cloud query features to obtain roadside lines; planar precision values ​​are obtained based on the roadside lines, including: obtaining the second distance values ​​between the query point and each roadside line; sorting the second distance values ​​and obtaining the second distance difference between each adjacent second distance value; determining the distance between at least two roadside lines that meet the first and second filtering conditions as the planar precision values; the first filtering condition is that the second distance values ​​corresponding to the roadside lines are sequentially adjacent and the second distance difference is greater than a second preset threshold; the second filtering condition is that the roadside lines originate from different segmented point clouds; Repeatability accuracy is evaluated based on elevation and planar accuracy values.

2. The method according to claim 1, characterized in that, Acquire vehicle trajectory data and vehicle-mounted laser point cloud data, including: During vehicle inspection, an onboard laser scanning system is used to collect point cloud data. The collected point cloud data is identified as onboard laser point cloud data, and the vehicle inspection route is identified as onboard trajectory data.

3. The method according to claim 2, characterized in that, Based on the vehicle trajectory data, the vehicle-mounted laser point cloud data is divided into multiple segmented point clouds, including: Obtain the angular velocity of all trajectory points in the vehicle trajectory data; Trajectory points with angular velocities greater than a first preset threshold are defined as segmentation points; The vehicle-mounted laser point cloud data is divided into multiple segmented point clouds according to each segment point.

4. The method according to claim 2, characterized in that, Point cloud query features are obtained based on each segment of the point cloud, including: A query cuboid is constructed with the query point as the center and the direction perpendicular to the vehicle trajectory on the horizontal plane as the main direction; the length of the cuboid is determined by the road width and the lane where the trajectory is located. The point cloud query features are obtained by sequentially querying each segment of the point cloud using a query cuboid.

5. The method according to claim 1, characterized in that, Based on the point cloud query features, a plane fitting is performed to obtain the fitted plane, including: Perform an initial fitting on the point cloud query features to obtain an initial fitting plane; Obtain the residuals and mean square error of the initial fitted plane; Noise is removed based on each residual and standard error; The initial fitted plane after noise removal is fitted a second time to obtain the final fitted plane.

6. The method according to claim 1, characterized in that, Roadside lines are extracted from the point cloud query features, resulting in the following roadside lines: Voxel division is performed on the point cloud query features to obtain multiple voxels; Feature extraction is performed on each voxel to obtain voxel features; The vehicle-mounted point cloud data is clustered according to voxel features to obtain the ground point cloud; Perform a neighborhood query on the ground point cloud and identify voxels in the neighborhood that have the characteristic of elevation as candidate point clouds for the roadside; Select roadside points from the candidate point cloud along the road; Connect the curb points according to the adjacency relationship between voxels to obtain the curb line.

7. A vehicle-mounted laser point cloud data repeatability accuracy evaluation device, characterized in that, The method for evaluating the repeatability accuracy of vehicle-mounted laser point cloud data according to any one of claims 1-6 includes: The point cloud data acquisition module is configured to acquire vehicle trajectory data and vehicle laser point cloud data; The point cloud segmentation module is configured to divide the vehicle-mounted laser point cloud data into multiple segmented point clouds based on the vehicle trajectory data; The point cloud query module is configured to obtain point cloud query features based on each segment of the point cloud; The repeatability accuracy evaluation module is configured to obtain repeatability accuracy evaluation based on point cloud query features.

Images