A mapping method, device and intelligent vehicle based on normal distribution transformation

Abstract

An embodiment of the present invention provides a mapping method, device, and intelligent vehicle based on normal distribution transformation. A lidar is provided on the intelligent vehicle. By acquiring a first point cloud image and a second point cloud image collected by the lidar, first point cloud data and second point cloud data are respectively extracted from the first point cloud image and the second point cloud image. Transformation parameters of the first point cloud data and the second point cloud data are calculated based on the normal distribution. Then, coordinate transformation is performed on the second point cloud data according to the transformation parameters so that the second point cloud data is loaded into the first point cloud image. The credibility weights of the point clouds in the loaded first point cloud image are calculated. The first point cloud image is updated according to the credibility weights, and the updated first point cloud image is output. By calculating the credibility weights of the point clouds during the mapping process and updating the output point cloud image according to the weights, the accuracy of the mapping is improved, and the mapping becomes more automated.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to a mapping method, device and intelligent vehicle based on normal distribution transformation.

Background Art

[0002] With the progress of social science, the laser point cloud mapping algorithm based on the normal distribution transform NDT (Normal Distributions Transform Mapping) has become increasingly mature. However, when using the laser point cloud mapping with normal distribution transformation, there may be dynamic obstacles in the map, such as people or vehicles. For such dynamic obstacles, the existing technologies do not involve the step of removing the dynamic obstacles encountered during the mapping process, or after the mapping is completed, manually removing the obstacles. However, removing the dynamic obstacles in the mapping process in the above manner makes the laser point cloud mapping algorithm based on normal distribution transformation unable to obtain the optimal solution when performing prior global planning in a small area.

Summary of the Invention

[0003] In view of the above problems, embodiments of the present invention provide a mapping method, device and intelligent vehicle based on normal distribution transformation, which overcome the above problems or at least partially solve the above problems.

[0004] To solve the above technical problems, a technical solution adopted in an embodiment of the present invention is: to provide a mapping method based on normal distribution transformation, which is applied to an intelligent vehicle. A lidar is provided on the intelligent vehicle. The method includes:

[0005] Obtain a first point cloud image and a second point cloud image collected by the lidar, extract first point cloud data from the first point cloud image, and extract second point cloud data from the second point cloud image;

[0006] Calculate the transformation parameters of the first point cloud data and the second point cloud data based on the normal distribution;

[0007] Perform coordinate transformation on the second point cloud data according to the transformation parameters, so that the second point cloud data is loaded into the first point cloud image;

[0008] Calculate the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data, and update the first point cloud image after loading the second point cloud data according to the credibility weights, so as to output the updated first point cloud image.

[0009] Optionally, the step of calculating the transformation parameters of the first point cloud data and the second point cloud data based on the normal distribution transformation specifically includes:

[0010] Obtain initial transformation parameters;

[0011] According to the initial transformation parameters, convert and load the second point cloud data into the first point cloud image;

[0012] Calculate the first probability density of each point cloud in the first point cloud image loaded with the second point cloud data based on the normal distribution;

[0013] According to the first probability density, calculate the registration score of each point cloud in the first point cloud image loaded with the second point cloud data, and perform a summation operation on the registration scores;

[0014] Optimize the initial transformation parameters according to the preset convergence conditions to obtain transformation parameters.

[0015] Optionally, the step of calculating the first probability density of each point cloud in the first point cloud image loaded with the second point cloud data based on the normal distribution transformation specifically includes:

[0016] Divide the space occupied by each point cloud in the first point cloud image loaded with the second point cloud data into voxel grids of a preset size;

[0017] Calculate the centroid and covariance of each voxel grid;

[0018] Calculate the first probability density according to the centroid and covariance of each voxel grid and the coordinates of each point cloud.

[0019] Optionally, the step of optimizing the initial transformation parameters according to the preset convergence conditions to obtain transformation parameters specifically includes:

[0020] Calculate the normal distribution probability density of the second point cloud data, and sum the normal distribution probability densities;

[0021] Obtain the normal distribution registration score according to the sum of the normal distribution probability densities;

[0022] Use the Gauss-Newton method to optimize the normal distribution transformation registration score;

[0023] Repeat the above steps until the initial transformation parameters meet the preset convergence conditions, where the convergence conditions include: reaching a preset number of iterative calculations; the tolerance of the initial transformation matrix obtained from two consecutive iterative calculations is less than a preset tolerance value; the distance between corresponding point pairs of point clouds obtained from two consecutive iterative calculations is less than a preset distance.

[0024] Optionally, the calculation loads the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data, and updates the first point cloud image after loading the second point cloud data according to the credibility weights to output an updated first point cloud image, including:

[0025] Obtain a first point cloud in the first point cloud image after loading the second point cloud data;

[0026] Obtain the first voxel center within the preset range of the first point cloud;

[0027] Calculate the distance between the first point cloud and the first voxel center according to the first voxel center;

[0028] Calculate the credibility weight of the first voxel based on the distance according to the credibility calculation formula;

[0029] Update the credibility of each point in the first point cloud image after loading the point cloud data according to the credibility weight, and output an updated first point cloud image.

[0030] Optionally, the method further includes:

[0031] Calculate the moving distance of the second point cloud data according to the transformation parameter;

[0032] If the moving distance is greater than or equal to a preset distance threshold, obtain the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data;

[0033] Add the point clouds with the credibility weights of the point clouds greater than the preset weight threshold to the first point cloud image to update the first point cloud image;

[0034] If the moving distance is less than the preset distance threshold, there is no need to update the first point cloud image.

[0035] To solve the above technical problems, another technical solution adopted in the embodiments of the present invention is: to provide a mapping device based on normal distribution transformation, which is applied to an intelligent vehicle. Among them, a lidar is provided on the intelligent vehicle, and the device includes:

[0036] A first acquisition module, configured to acquire a first point cloud image and a second point cloud image collected by the lidar, extract first point cloud data from the first point cloud image, and extract second point cloud data from the second point cloud image;

[0037] A calculation module, configured to calculate the transformation parameter of the first point cloud data and the second point cloud data based on the normal distribution;

[0038] A transformation module, configured to perform coordinate transformation on the second point cloud data according to the transformation parameters, so as to load the second point cloud data into the first point cloud image;

[0039] An output module, configured to calculate the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data, and update the first point cloud image after loading the second point cloud data according to the credibility weights, so as to output the updated first point cloud image.

[0040] To solve the above technical problems, another technical solution adopted in the embodiments of the present invention is: to provide an intelligent vehicle, the intelligent vehicle includes:

[0041] At least one processor; and,

[0042] A memory communicatively connected to the at least one processor; wherein,

[0043] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor, so that the at least one processor can execute the method described in any one of the above.

[0044] To solve the above technical problems, another technical solution adopted in the embodiments of the present invention is: to provide a non-volatile computer-readable storage medium, the non-volatile computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by an intelligent vehicle, the intelligent vehicle executes the method described in any one of the above.

[0045] To solve the above technical problems, another technical solution adopted in the embodiments of the present invention is: to provide a computer program product, the computer program product includes a computer program stored on a non-volatile computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by an intelligent vehicle, the intelligent vehicle executes the method described in any one of the above.

[0046] Different from the related art, embodiments of the present invention provide a mapping method, device, intelligent vehicle, non-volatile computer-readable storage medium, and computer product program based on normal distribution transformation, which are applied to an intelligent vehicle. A lidar is provided on the intelligent vehicle. The mapping method, device, intelligent vehicle, non-volatile computer-readable storage medium, and computer product program based on normal distribution transformation obtain a first point cloud image and a second point cloud image collected by the lidar, extract first point cloud data from the first point cloud image, and extract second point cloud data from the second point cloud image. Then, the transformation parameters of the first point cloud data and the second point cloud data are calculated based on the normal distribution, and the second point cloud data is subjected to coordinate transformation according to the transformation parameters to load the second point cloud data into the first point cloud image. Finally, the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data are calculated, and the first point cloud image after loading the second point cloud data is updated according to the credibility weights, so as to output the updated first point cloud image. By calculating the credibility weights of the point clouds during the mapping process and updating the output point cloud image according to the weights, the accuracy of the mapping is improved, making the mapping process more automated.

BRIEF DESCRIPTION OF THE DRAWINGS

[0047] One or more embodiments are exemplarily illustrated by corresponding drawings. These exemplary illustrations do not limit the embodiments. Elements with the same reference numerals in the drawings represent similar elements, unless otherwise stated. The drawings in the figures do not constitute a proportional limitation.

[0048] Figure 1 It is a schematic diagram of the hardware structure of an intelligent vehicle that executes the mapping method based on normal distribution transformation provided by an embodiment of the present invention;

[0049] Figure 2 It is a schematic diagram of an intelligent vehicle provided by an embodiment of the invention;

[0050] Figure 3 It is a flowchart of a mapping method based on normal distribution transformation provided by an embodiment of the present invention;

[0051] Figure 4 It is a flowchart of obtaining transformation parameters in a mapping method based on normal distribution transformation provided by an embodiment of the present invention;

[0052] Figures 5a - 5c It is a schematic diagram after mapping based on normal distribution transformation provided by an embodiment of the present invention;

[0053] Figure 6 It is a structural block diagram of a mapping device based on normal distribution transformation provided by an embodiment of the present invention.

DETAILED DESCRIPTION

[0054] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0055] It should be noted that if there is no conflict, the various features in the embodiments of the present invention can be combined with each other and are all within the protection scope of the present invention. In addition, although the functional modules are divided in the device schematic diagram and the logical sequence is shown in the flowchart, in some cases, the steps shown or described can be executed in a different module division from that in the device schematic diagram or a different order from that in the flowchart.

[0056] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not used to limit the present invention. The term "and / or" used in this specification includes any and all combinations of one or more of the related listed items.

[0057] Please refer to Figure 1 , an embodiment of the present invention provides an intelligent vehicle 30, and the intelligent vehicle 30 includes: at least one processor 31, Figure 1 Taking one processor 31 as an example; a memory 32 communicatively connected to the at least one processor 31, Figure 1 Taking the connection through a bus as an example.

[0058] Wherein, the memory 32 stores instructions executable by the at least one processor 31, and the instructions are executed by the at least one processor 31 so that the at least one processor 31 can execute the following mapping method based on normal distribution transformation.

[0059] As a non-volatile computer-readable storage medium, the memory 32 can be used to store non-volatile software programs, non-volatile computer-executable programs and modules, such as the program instructions / modules corresponding to the mapping method based on normal distribution transformation in the embodiments of the present invention. The processor 31 executes various functional applications and data processing of the intelligent vehicle 30 by running the non-volatile software programs, instructions and modules stored in the memory 32, that is, implements the mapping method based on normal distribution transformation in the following method embodiments.

[0060] The memory 32 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function. In addition, the memory 32 may include a high-speed random access memory and may also include a non-volatile memory. For example, it includes at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the memory 32 may optionally include a memory remotely provided with respect to the processor 31.

[0061] The one or more modules are stored in the memory 32 and, when executed by the one or more processors 31, execute the mapping method based on normal distribution transformation in any of the following method embodiments. For example, execute the Figure 3 , Figure 4 method steps therein.

[0062] The intelligent vehicle 30 is also connected to other devices to better execute the method provided by the embodiments of the present invention. For example, it can be electrically connected to a display screen or other displays, and can be remotely communicatively connected to the communication devices of target users, etc., which are not listed one by one here.

[0063] Please refer to Figure 2 , Figure 2 which is a schematic diagram of the intelligent vehicle provided by the embodiments of the invention. As Figure 2 shown, a lidar is provided on the intelligent vehicle, and the lidar is used to obtain a point cloud image.

[0064] The above intelligent vehicle can execute the method provided by the embodiments of the present invention and has functional modules corresponding to the execution of the method. For technical details not described in detail in this embodiment, reference can be made to the method provided by the embodiments of the present invention.

[0065] Please refer to Figure 3 , Figure 3 which is a flowchart of a mapping method based on normal distribution transformation provided by the embodiments of the present invention. The method is applied to the above intelligent vehicle. As Figure 3 shown, the steps of the method include:

[0066] S01. Obtain a first point cloud image and a second point cloud image collected by the lidar, extract first point cloud data from the first point cloud image, and extract second point cloud data from the second point cloud image.

[0067] The lidar refers to a scanning sensor that uses non-contact laser ranging technology. It mainly detects targets by emitting laser beams and forms point clouds and obtains data by collecting the reflected beams. These data can be generated into accurate three-dimensional stereo images after optoelectronic processing.

[0068] Specifically, the first point cloud image refers to a global map coordinate image obtained by a lidar, and the second point cloud image is a point cloud image near the current position of the intelligent vehicle. Then, first point cloud data is extracted from the first point cloud image, and second point cloud data is extracted from the second point cloud image. Among them, the point cloud data of different frames is different.

[0069] S02. Calculate the transformation parameters of the first point cloud data and the second point cloud data based on the normal distribution.

[0070] The normal distribution transformation algorithm refers to a registration algorithm, which is applied to the statistical model of three-dimensional point clouds. It mainly uses standard optimization techniques to determine the optimal matching between two point clouds. Because it does not use the feature calculation and matching of corresponding points during the registration process, the time is faster than other methods.

[0071] Among them, please refer to Figure 4 , Figure 4 is a flowchart for obtaining transformation parameters in a mapping method based on normal distribution transformation provided in an embodiment of the present invention. As Figure 4 shown, optimizing the transformation parameters includes the following steps:

[0072] S021. Obtain the initial transformation parameters.

[0073] The transformation parameter refers to a transformation matrix for converting the second point cloud data from the second point cloud image to the first point cloud image. Among them, in the absence of other sensor inputs, the transformation matrix can be an identity matrix.

[0074] S022. According to the initial transformation parameters, convert and load the second point cloud data into the first point cloud image.

[0075] Specifically, the second point cloud data is converted into the first point cloud image through the initialized transformation parameters. At this time, since the first point cloud image includes the first point cloud data and the second point cloud data, the point cloud data in the first point cloud map is relatively dense. Therefore, the second point cloud data is downsampled according to a set threshold to obtain downsampled point clouds, thereby reducing the memory consumption. And only when the first point cloud data and the second point cloud data both belong to the same point cloud image can the point cloud data be matched and the optimal solution of the transformation parameter can be obtained.

[0076] S023. Calculate the first probability density of each point cloud in the first point cloud image loaded with the second point cloud data based on the normal distribution.

[0077] Specifically, before calculating the first probability density, it is also necessary to divide the space occupied by each point cloud in the first point cloud image loaded with the second point cloud data into voxel grids of a preset size, then calculate the centroid and covariance of each voxel grid, and finally calculate the first probability density according to the centroid and covariance of each voxel grid and the coordinates of each point cloud.

[0078] Among them, the formula for calculating the centroid of the voxel grid is:

[0079]

[0080] The formula for calculating the covariance of the voxel grid is:

[0081]

[0082] The formula for calculating the first probability density is:

[0083]

[0084] Among them, x k is the coordinate of the k-th point in the current grid, and n is the number of points in the current grid.

[0085] S024. Calculate the registration score of each point cloud in the first point cloud image loaded with the second point cloud data according to the first probability density, and perform a summation operation on the registration scores.

[0086] Specifically, first obtain a three-dimensional transformation matrix according to the first probability density:

[0087]

[0088] Among them, p = [t x t y t z T is the transformation for point x, t is the translation matrix, r x , r y , r z are the three rotation axes respectively, and there are:

[0089]

[0090] s = sinφ

[0091] C = cosφ

[0092] e = 1 - cosφ

[0093] Among them, is the rotation angle.

[0094] ​Then, according to the three-dimensional transformation matrix, the second point cloud data is transformed into the voxel grid of the first point cloud image, the normal distribution probability density of the second point cloud data is calculated, and the normal distribution probability densities are summed up; then, a normal distribution registration score is obtained according to the sum of the normal distribution probability densities. The formula for calculating the registration score is as follows:

[0095]

[0096] S025. Optimize the initial transformation parameters according to the preset convergence conditions to obtain the transformation parameters.

[0097] Specifically, the Gaussian-Newton method is used to optimize the normal distribution transformation registration score, that is, to find the transformation parameter T that maximizes the value of s, and repeat until the set convergence conditions are met, then it is considered that the optimized transformation parameters are obtained. Among them, the Gaussian-Newton method refers to approximately replacing the non-linear regression model through the Taylor series expansion, and then through multiple iterations, repeatedly correcting the regression coefficients to make the regression coefficients continuously approach the best regression coefficients of the non-linear regression model, and finally minimizing the sum of the squared residuals of the original model.

[0098] Among them, the convergence conditions for optimizing the transformation parameters include:

[0099] Reaching the preset number of iterative calculations, that is, the maximum number limit of iterative calculations. If this value is too large, it may cause the algorithm to take too long, and if it is too small, it may cause the algorithm not to converge to the optimal solution. Generally, the set value is 100.

[0100] The tolerance of the initial transformation matrix obtained from the previous and subsequent iterative calculations is less than the preset tolerance value. That is, when the tolerance of the transformation matrix obtained from two iterations is less than the set value, it is considered that the optimal solution has been converged. If this value is too large, it may cause the algorithm not to converge completely, and if it is too small, it may cause the algorithm to take too long.

[0101] The distance between the corresponding point pairs of the point clouds obtained from the previous and subsequent iterative calculations is less than the preset distance. That is, when the distance between the corresponding point pairs of the two point clouds obtained by transforming the current frame point cloud through the transformation matrices obtained from two iterations is less than the set value, it is considered that the optimal solution has been converged. If this value is too large, it may cause the algorithm not to converge completely, and if it is too small, it may cause the algorithm to take too long.

[0102] S03. Perform coordinate transformation on the second point cloud data according to the transformation parameters, so that the second point cloud data is loaded into the first point cloud image.

[0103] S04. Calculate the credibility weights of each point cloud in the first point cloud image after loading the second point cloud data, and update the first point cloud image after loading the second point cloud data according to the credibility weights to output the updated first point cloud image.

[0104] Specifically, first obtain the first point cloud in the first point cloud image after loading the second point cloud data, and obtain the first voxel center within the preset range of the first point cloud. Then, according to the first voxel center, calculate the distance between the first point cloud and the first voxel center, and based on the distance, calculate the credibility weight of the first voxel according to the credibility calculation formula. Finally, according to the credibility weight, update the credibility of each point in the first point cloud image after loading the point cloud data, and output the updated first point cloud image.

[0105] The formula for calculating the credibility weight of the first voxel center is:

[0106]

[0107] where x is the Euclidean distance between the point and the voxel center, and μ and σ are the mean and variance of the normal distribution respectively. The Euclidean distance refers to the actual distance between two points in an m-dimensional space, or the natural length of a vector (i.e., the distance from this point to the origin).

[0108] Specifically, update each point cloud in the first point cloud map through the formula of the credibility weight, retain the point cloud with a high calculated credibility weight in the first point cloud map, and remove the point cloud with a low credibility weight. Among them, the point cloud with a low credibility refers to the point cloud generated by dynamic obstacles. Finally, output the updated first point cloud image.

[0109] Specifically, please refer to Figures 5a - 5c , Figures 5a - 5c is a schematic diagram after mapping based on the normal distribution transformation provided by this embodiment. Among them, Figure 5a is the original image when there are obstacles, Figure 5b is the image obtained after removing the point cloud image with a low credibility, Figure 5c is the final image after the mapping is completed.

[0110] In some embodiments, updating the first point cloud map further requires calculating the moving distance of the second point cloud data according to the transformation parameters. If the moving distance is greater than or equal to a preset distance threshold, the confidence weights of the point clouds in the first point cloud image after loading the second point cloud data are obtained, and the point clouds with the confidence weights of the point clouds greater than the preset weight threshold are added to the first point cloud image to update the first point cloud image; if the moving distance is less than the preset distance threshold, the first point cloud image does not need to be updated.

[0111] Different from the related art, the embodiment of the present invention provides a mapping method based on normal distribution transformation, which is applied to an intelligent vehicle. The method obtains a first point cloud image and a second point cloud image collected by the lidar, extracts first point cloud data from the first point cloud image, extracts second point cloud data from the second point cloud image, then calculates the transformation parameters of the first point cloud data and the second point cloud data based on the normal distribution, performs coordinate transformation on the second point cloud data according to the transformation parameters, so that the second point cloud data is loaded into the first point cloud image, and finally calculates the confidence weights of the point clouds in the first point cloud image after loading the second point cloud data, and updates the first point cloud image after loading the second point cloud data according to the confidence weights, thereby outputting an updated first point cloud image. By calculating the confidence weights of the point clouds during the mapping process and updating the output point cloud image according to the weights, the accuracy of the mapping is improved, making the mapping process more automated.

[0112] Please refer to Figure 6 , Figure 6 which is the structural block diagram of a mapping device based on normal distribution change provided by the embodiment of the present invention. As Figure 6 shown, the mapping device 1 based on normal distribution transformation is applied to the above-mentioned intelligent vehicle. The mapping device 1 based on normal distribution transformation includes a first acquisition module 11, a calculation module 12, a transformation module 13, and an output module 14.

[0113] The first acquisition module 11 is used to acquire the first point cloud image and the second point cloud image collected by the lidar, and extract the first point cloud data from the first point cloud image and the second point cloud data from the second point cloud image.

[0114] The calculation module 12 is used to calculate the transformation parameters of the first point cloud data and the second point cloud data based on the normal distribution.

[0115] Among them, the calculation module 11 includes a first acquisition unit 111, a conversion unit 112, a first calculation unit 113, a second calculation unit 114, and an optimization unit 115.

[0116] The first acquisition unit 111 is configured to acquire initial transformation parameters.

[0117] The conversion unit 112 is configured to convert and load the second point cloud data into the first point cloud image according to the initial transformation parameters.

[0118] The first calculation unit 113 is configured to calculate the first probability density of each point cloud in the first point cloud image loaded with the second point cloud data based on a normal distribution.

[0119] The second calculation unit 114 is configured to calculate the registration score of each point cloud in the first point cloud image loaded with the second point cloud data according to the first probability density, and perform a summation operation on the registration scores.

[0120] The optimization unit 115 is configured to optimize the initial transformation parameters according to a preset convergence condition to obtain transformation parameters.

[0121] The transformation module 13 is configured to perform a coordinate transformation on the second point cloud data according to the transformation parameters, so that the second point cloud data is loaded into the first point cloud image.

[0122] The output module 14 is configured to calculate the credibility weight of each point cloud in the first point cloud image after the second point cloud data is loaded, and update the first point cloud image after the second point cloud data is loaded according to the credibility weight, so as to output the updated first point cloud image.

[0123] It should be noted that the above mapping device based on normal distribution transformation can execute the mapping method based on normal distribution transformation provided by the embodiments of the present invention, and has corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in the embodiments of the mapping device based on normal distribution transformation can be found in the mapping method based on normal distribution transformation provided by the embodiments of the present invention.

[0124] Embodiments of the present invention further provide a non-volatile computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are executed by one or more processors. For example, the method steps described above are executed to Figure 3 and Figure 4 realize the functions of each module in Figure 6 .

[0125] An embodiment of the present invention provides a computer program product. The computer program product includes a computer program stored on a non-volatile computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by the intelligent vehicle, the intelligent vehicle can execute the mapping method based on normal distribution transformation in any of the above method embodiments. For example, execute the method steps S01 to S04 described above Figure 3 in the method steps S01 to S04, Figure 4 in the method steps S021 to S025, and implement Figure 6 the functions of modules 11 - 14 in

[0126] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0127] Through the description of the above embodiments, those of ordinary skill in the art can clearly understand that each embodiment can be implemented by means of software plus a general hardware platform, and of course, it can also be implemented by hardware. Those of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, not to limit them; under the idea of the present invention, the technical features in the above embodiments or different embodiments can also be combined, and the steps can be implemented in any order, and there are many other changes in different aspects of the present invention as described above. For the sake of brevity, they are not provided in detail; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A mapping method based on normal distribution transformation, applied to an intelligent vehicle, wherein, The intelligent vehicle is equipped with a lidar. It is characterized in that the method includes: Obtain a first point cloud image and a second point cloud image collected by the lidar, extract first point cloud data from the first point cloud image, and extract second point cloud data from the second point cloud image; Calculate the transformation parameters of the first point cloud data and the second point cloud data based on the normal distribution; Perform a coordinate transformation on the second point cloud data according to the transformation parameters so that the second point cloud data is loaded into the first point cloud image; Calculate the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data, and update the first point cloud image after loading the second point cloud data according to the credibility weights to output an updated first point cloud image; Among them, the step of calculating the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data, and updating the first point cloud image after loading the second point cloud data according to the credibility weights to output an updated first point cloud image includes: Obtain the first point cloud in the first point cloud image after loading the second point cloud data; Obtain the first voxel center within a preset range of the first point cloud; Calculate the distance between the first point cloud and the first voxel center according to the first voxel center; Calculate the credibility weight of the first voxel based on the credibility calculation formula according to the distance; Update the credibility of each point in the first point cloud image after loading the second point cloud data according to the credibility weight, and output an updated first point cloud image; Among them, the method further includes: Calculate the moving distance of the second point cloud data according to the transformation parameters; If the moving distance is greater than or equal to a preset distance threshold, obtain the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data; Add the point clouds with the credibility weights of the point clouds greater than the preset weight threshold to the first point cloud image to update the first point cloud image; If the moving distance is less than the preset distance threshold, there is no need to update the first point cloud image.

2. The method according to claim 1, wherein The step of calculating the transformation parameters of the first point cloud data and the second point cloud data based on the normal distribution specifically includes: Obtain initial transformation parameters; Convert and load the second point cloud data into the first point cloud image according to the initial transformation parameters; Calculate the first probability density of each point cloud in the first point cloud image loaded with the second point cloud data based on the normal distribution; Calculate the registration scores of the point clouds in the first point cloud image loaded with the second point cloud data according to the first probability density, and perform a summation operation on the registration scores; Optimize the initial transformation parameters according to the preset convergence condition to obtain the transformation parameters.

3. The method according to claim 2, characterized in that, The step of calculating the first probability density of each point cloud in the first point cloud image loaded with the second point cloud data based on the normal distribution specifically includes: Divide the space occupied by each point cloud in the first point cloud image loaded with the second point cloud data into voxel grids of a preset size; Calculate the centroid and covariance of each of the voxel grids; Calculate the first probability density based on the centroid and covariance of each of the voxel grids and the coordinates of the respective point clouds.

4. The method according to claim 2, wherein The step of optimizing the initial transformation parameters according to the preset convergence conditions to obtain the transformation parameters specifically includes: Calculate the normal distribution probability density of the second point cloud data and sum the normal distribution probability densities. Obtain the normal distribution registration score based on the sum of the normal distribution probability densities. Optimize the normal distribution transformation registration score using the Gauss-Newton method. Repeat the above steps until the initial transformation parameters satisfy the preset convergence conditions, where the convergence conditions include: reaching a preset number of iterative calculations; the tolerance between the initial transformation matrices obtained from two consecutive iterative calculations being less than a preset tolerance value; the distance between corresponding point pairs of the point clouds obtained from two consecutive iterative calculations being less than a preset distance.

5. A mapping device based on normal distribution transformation, which is applied to an intelligent vehicle, wherein, The intelligent vehicle is equipped with a lidar, characterized in that the device includes: A first acquisition module, configured to acquire a first point cloud image and a second point cloud image collected by the lidar, and extract first point cloud data from the first point cloud image and second point cloud data from the second point cloud image; A calculation module, configured to calculate the transformation parameters of the first point cloud data and the second point cloud data based on the normal distribution; A transformation module, configured to perform a coordinate transformation on the second point cloud data according to the transformation parameters, so that the second point cloud data is loaded into the first point cloud image; An output module, configured to calculate the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data, and update the first point cloud image after loading the second point cloud data according to the credibility weights to output an updated first point cloud image; Wherein, the output module is specifically configured to: Obtain the first point cloud in the first point cloud image after loading the second point cloud data; Obtain the first voxel center within a preset range of the first point cloud; Calculate the distance between the first point cloud and the first voxel center according to the first voxel center; Calculate the credibility weight of the first voxel based on the distance according to the credibility calculation formula; Update the credibility of each point in the first point cloud image after loading the second point cloud data according to the credibility weight, and output an updated first point cloud map; Wherein, the calculation module is further configured to: Calculate the moving distance of the second point cloud data according to the transformation parameters; If the moving distance is greater than or equal to a preset distance threshold, obtain the credibility weights of the point clouds in the first point cloud image after loading the second point cloud data; Add the point clouds with the credibility weights of the point clouds greater than a preset weight threshold to the first point cloud image to update the first point cloud image; If the moving distance is less than the preset distance threshold, there is no need to update the first point cloud image.

6. An intelligent vehicle, characterized in that, The intelligent vehicle includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method according to any one of claims 1-4.

7. A non-volatile computer-readable storage medium, characterized in that, The non-volatile computer-readable storage medium stores computer-executable instructions, which, when executed by the intelligent vehicle, cause the intelligent vehicle to execute the method according to any one of claims 1-4.

8. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-volatile computer-readable storage medium, the computer program including program instructions which, when executed by the intelligent vehicle, cause the intelligent vehicle to execute the method according to any one of claims 1-4.

Images