Multi-domain point cloud registration automatic identification and positioning method for electric vehicle charging port

Through the multi-domain point cloud registration method, combining knowledge filtering, data filtering and iterative closest point registration, the problem of noise interference in the point cloud of electric vehicle charging port is solved, and accurate point cloud registration and charging port recognition is achieved, improving the The stability and reliability of automatic charging are achieved.

CN118587289BActive Publication Date: 2024-12-06国网(山东)电动汽车服务有限公司
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411067418.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-06
Publication Date
2024-12-06
Estimated Expiration
2044-08-06

AI Technical Summary

Technical Problem

The prior art fails to effectively remove noise interference in point cloud registration of electric vehicle charging ports, resulting in poor noise immunity of point clouds, making it difficult to achieve accurate identification and positioning, affecting the stability and reliability of the automatic charging process.

Method used

Multi-domain point cloud registration method is adopted, including knowledge filtering, data filtering, spatial filtering and iterative closest point registration, and interfering noise is removed through historical point cloud information, data feature correlation degree and point cloud space, and based on errors in different domains Scoring with weights, selecting the best point cloud variation method to obtain the best rotation matrix and displacement matrix.

Benefits of technology

Effectively remove interference noise in point clouds, improve point cloud matching accuracy and noise immunity, realize accurate identification and positioning of electric vehicle charging ports, reduce the point cloud computing error rate of automatic charging technology, and ensure reliable Point cloud registration.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118587289B_ABST
    Figure CN118587289B_ABST
Patent Text Reader

Abstract

The present invention proposes a multi-domain point cloud registration automatic identification and positioning method for an electric vehicle charging port, which belongs to the technical field of charging piles and includes the following steps: S1: establishing a point cloud domain of an electric vehicle charging port model, and a sensor collects the original point cloud domain of the charging port; S2: obtaining a visual point cloud domain based on knowledge filtering, data filtering, and spatial filtering; S3: performing wavelet transformation on the visual point cloud domain to obtain a wavelet point cloud domain; S4: performing registration based on iterative nearest points in the original point cloud domain, the visual point cloud domain, and the wavelet point cloud domain, respectively, and scoring the registration results based on the errors and weights of different domains; S5: selecting the best point cloud change method, and the corresponding rotation matrix and displacement matrix as the final result of the registration, and outputting the best rotation matrix and the best displacement matrix. While retaining the features, the present invention removes the interference noise in the original point cloud, and can obtain the best rotation matrix and displacement matrix, thereby achieving accurate and reliable registration of the point cloud.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of charging piles, and in particular relates to a multi-domain point cloud registration automatic identification and positioning method for an electric vehicle charging port. Background Art

[0002] With the vigorous promotion of the new energy vehicle strategy, autonomous driving technology is gradually being applied to electric vehicles. In the past, vehicle charging mainly relied on manual operation, and the charging process was complicated. The emergence of autonomous driving technology has put forward higher requirements for vehicle autonomy and automatic charging technology. Automatic charging technology can improve the charging efficiency of electric vehicles and adapt to autonomous driving systems. The identification and positioning technology of electric vehicle charging ports can identify and locate charging ports, solve problems such as smooth control during the plugging and unplugging of robotic arms, and realize automatic charging services in different scenarios without changing the existing charging socket standards or the appearance of the vehicle.

[0003] However, due to the limitations of the sensor, the inherent noise of the acquisition equipment, the lighting or reflective properties of the surface or artifacts in the scene, the original point cloud of the charging port collected by the sensor is inevitably interfered by noise. Comparative Document 1 (Application No.: 202110954702.3) provides a point cloud registration method based on topological structure and multi-scale features, using fuzzy matching for local matching and using the topological relationship between substructures for global registration, thereby eliminating false matches and obtaining robust matching results, improving the accuracy of point cloud matching. This method does not take into account the noise in the original point cloud, lacks a noise filtering mechanism, and is difficult to support the accurate identification and positioning of the charging port.

[0004] Comparative Document 2 (Application No.: 202211417357.0) provides a point cloud registration method based on point neighborhood condition constraints, using a sampling consistency algorithm to complete the initial registration, and using an iterative closest point algorithm that introduces a kd tree to improve the registration efficiency and achieve accurate registration of the point cloud. This method does not consider the impact of noise that is difficult to filter out on the point cloud. The point cloud of the charging port has poor noise resistance and is easily affected by noise, resulting in large errors in point cloud calculations. It is impossible to accurately identify and locate the charging port in a complex environment, and it is difficult to achieve reliable point cloud registration, which affects the stability and reliability of the entire electric vehicle automatic charging process. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, a multi-domain point cloud registration automatic identification and positioning method for an electric vehicle charging port is provided.

[0006] The technical solution adopted by the present invention to solve its technical problem is:

[0007] This technical solution proposes a multi-domain point cloud registration automatic identification and positioning method for electric vehicle charging ports, including the following steps:

[0008] S1: Establish the point cloud domain of the electric vehicle charging port model, and the sensor collects the original point cloud domain of the charging port;

[0009] S2: Based on the historical point cloud information, the knowledge filtering of the original point cloud domain is realized, based on the data feature correlation, the data filtering of the original point cloud domain is realized, and based on the point cloud space, the spatial filtering of the original point cloud domain is realized to obtain the visual point cloud domain;

[0010] S3: Perform wavelet transformation on the visual point cloud domain to obtain the wavelet point cloud domain;

[0011] S4: Perform registration based on iterative closest point in the original point cloud domain, visual point cloud domain, and wavelet point cloud domain, and score the registration results based on the errors and weights of different domains;

[0012] S5: Select the best point cloud change method, and the corresponding rotation matrix and displacement matrix as the final result of the registration. Output the best rotation matrix and the best displacement matrix to achieve accurate and reliable point cloud registration.

[0013] Preferably, in S2, the knowledge filtering comprises the following steps:

[0014] A historical knowledge base is built based on the interference noise information of historical point clouds. The historical knowledge base contains the sparsity of interference noise in different historical point clouds. q The average Euclidean distance and Euclidean distance variance of each noise point are used to calculate the sparsity of the interference noise points. The formula is as follows:

[0015] (1)

[0016] In the formula, is the sparseness of interference noise points predicted based on the historical knowledge base, Based on l The sparseness of the interference noise points predicted by the historical point cloud, , Respectively l Interference noise points in historical point cloud k and its neighbors q The average Euclidean distance and Euclidean distance variance of the internal noise points, For the l Interference noise points in historical point cloud k Its neighboring areas i The Euclidean distance of points, For the l The prediction adjustment parameters of the sparsity of interference noise points in the historical point cloud;

[0017] Calculate the sparsity of each point in the original point cloud ,like , then the points in the original point clouda The invalid interference noise is removed from the original point cloud domain.

[0018] Preferably, in S2, data filtering includes the following steps:

[0019] The original point cloud and the model point cloud are meshed to obtain the original point cloud matrix and the model point cloud matrix, and then the eigenvalues ​​of the original point cloud and the model point cloud are calculated respectively;

[0020] For all points in the original point cloud, the eigenvalues ​​of the points considered and those not considered are calculated respectively, and compared with the eigenvalues ​​of the model point cloud;

[0021] If the original point cloud eigenvalue of the point is far away from the eigenvalue of the model point cloud, and the original point cloud eigenvalue of the point is not close to the eigenvalue of the model point cloud, it is regarded as an interference noise point and is removed from the original point cloud domain.

[0022] Preferably, in S2, the spatial filtering comprises the following steps:

[0023] Select the coordinates of the outermost points in each direction of the point cloud image to construct the point cloud space. According to all the points in the point cloud space, find the coordinates of the center of gravity in this point cloud space. The formula is as follows:

[0024] (2)

[0025] In the formula, ( X, Y, Z ) represents the centroid coordinates of the point cloud space, ( x a ,y a ,z a ) is the coordinate of each point cloud in the point cloud space;

[0026] Set point cloud distance threshold D max , calculate the distance from each point in the point cloud space to the centroid coordinates D a ,like D max > D a , then the points in the original point cloud a The invalid interference noise is removed from the original point cloud domain.

[0027] Preferably, in S4, the iterative closest point registration method comprises the following steps:

[0028] The original point cloud of the original point cloud domain A With model point cloud M A , visual point cloud domain SWith model point cloud M S , Wavelet Point Cloud Domain H With model point cloud M H As input to the iterative closest point algorithm;

[0029] Model point cloud M A For the previously established electric vehicle charging port point cloud, M A Based on knowledge, data, and spatial filtering M S , M S Perform wavelet transformation to obtain M H , A Based on knowledge, data, and spatial filtering S , S Perform wavelet transformation to obtain H .

[0030] Preferably, consider J Point cloud transformation method, j The rotation matrix and displacement matrix of the point cloud transformation method are expressed as Q J and U J , for points in the model point cloud in different point cloud domains m a , m s , m j , in the original point cloud A , Visual Point Cloud S , Wavelet point cloud H Find the point with the closest Euclidean distance to it a , s , h As corresponding points, they form a corresponding point set;

[0031] Calculate the normal vector angle of the corresponding point pair. If it is greater than the set angle threshold β , then remove this point pair, update the corresponding point set, and use the j A point cloud transformation method for the original point cloud A point a For rotation and displacement changes, the formula is as follows:

[0032] (3)

[0033] In the formula, m a,j For point a Adopt thej The points obtained by rotating and displacing the point cloud using the point cloud transformation method;

[0034] Point a With point m a,j Make a visual point cloud domain change and get the point a With point m a,j Corresponding points on the visual point cloud s With point m s,j , point a With point m a,j Do the wavelet point cloud domain change to get the point a With point m a,j Corresponding points on the wavelet point cloud h With point m h,j;

[0035] The calculation is based on j Registration errors in the original point cloud domain, visual point cloud domain and wavelet point cloud domain of a point cloud transformation method , and score the registration result based on the error and the preset weight. The formula is as follows:

[0036] (4)

[0037] In the formula, are the scoring weights of the original point cloud domain, visual point cloud domain, and wavelet point cloud domain respectively;

[0038] The point cloud change method with the highest registration result score is selected as the optimal point cloud change method, and the corresponding rotation matrix and displacement matrix are used as the final results of the registration to achieve accurate and reliable point cloud registration.

[0039] Compared with the prior art, the present invention has the following advantages:

[0040] This application performs knowledge, data, and space filtering operations on the original point cloud, and completes the preprocessing of the original point cloud data based on historical point cloud information, data feature correlation, and point cloud space to obtain an accurate visual point cloud suitable for further processing. While retaining the features, it can effectively remove interference noise in the original point cloud, thereby facilitating the accurate identification and positioning of the charging port.

[0041] This application obtains a wavelet point cloud domain by performing wavelet transformation on the visual point cloud domain. The original point cloud domain, visual point cloud domain, and wavelet point cloud domain are respectively aligned based on the iterative nearest point. The alignment results are scored based on the errors and weights of different domains, and the best point cloud transformation method is selected to obtain the best rotation matrix and the best displacement matrix, which effectively reduces the point cloud operation error rate of electric vehicle automatic charging technology, solves the problem of accurate identification and positioning of charging ports in complex environments, and realizes reliable point cloud registration. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0043] Figure 1 It is the overall flow chart of the present invention. DETAILED DESCRIPTION

[0044] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0045] like Figure 1 As shown, this embodiment proposes a multi-domain point cloud registration automatic identification and positioning method for an electric vehicle charging port, including the following steps:

[0046] S1: Establish the point cloud domain of the electric vehicle charging port model, and the sensor collects the original point cloud domain of the charging port;

[0047] S2: Based on the historical point cloud information, the knowledge filtering of the original point cloud domain is realized, based on the data feature correlation, the data filtering of the original point cloud domain is realized, and based on the point cloud space, the spatial filtering of the original point cloud domain is realized to obtain the visual point cloud domain;

[0048] S3: Perform wavelet transformation on the visual point cloud domain to obtain the wavelet point cloud domain;

[0049] S4: Perform registration based on iterative closest points (ICP) in the original point cloud domain, visual point cloud domain, and wavelet point cloud domain, and score the registration results based on the errors and weights of different domains;

[0050] S5: Select the best point cloud change method, and the corresponding rotation matrix and displacement matrix as the final result of the registration. Output the best rotation matrix and the best displacement matrix to achieve accurate and reliable point cloud registration.

[0051] Due to the limitations of sensors, the inherent noise of acquisition equipment, the lighting or reflective properties of surfaces or artifacts in the scene, the raw point clouds collected by these sensors are inevitably contaminated by noise and contain interfering noise points. At the same time, due to the huge amount of point cloud data obtained, there is a great impact on the subsequent processing, storage and operation of the point cloud.

[0052] Therefore, it is necessary to perform filtering operations on the original point cloud, and complete the preprocessing of the point cloud data based on the historical interference noise information, data feature correlation and point cloud space, so as to obtain an accurate point cloud suitable for further processing, and remove the interference noise in the point cloud while retaining the features. A points, the set is represented by .

[0053] In S2, knowledge filtering includes the following steps:

[0054] Knowledge filtering builds a historical knowledge base based on the interference noise information of historical point clouds. The historical knowledge base contains the sparsity of interference noise in different historical point clouds. q The average Euclidean distance and Euclidean distance variance of each noise point are used to calculate the sparsity of the interference noise points. The formula is as follows:

[0055] (1)

[0056] In the formula, is the sparseness of interference noise points predicted based on the historical knowledge base, Based on l The sparseness of the interference noise points predicted by the historical point cloud, , Respectively l Interference noise points in historical point cloud k and its neighbors q The average Euclidean distance and Euclidean distance variance of the internal noise points, For the l Interference noise points in historical point cloud k Its neighboring areas i The Euclidean distance of points, For the l The prediction adjustment parameters of the sparsity of interference noise points in the historical point cloud;

[0057] The sparsity represents the information contained in the point cloud. L Historical point clouds, l There are historical point clouds K L The sparseness of interference noise points is predicted based on the historical knowledge base.

[0058] Calculate the sparsity of each point in the original point cloud ,like , then the points in the original point cloud a The invalid interference noise is removed from the original point cloud domain.

[0059] In S2, data filtering removes interference noise points in the point cloud based on the distribution characteristics of the point cloud data. Data filtering includes the following steps:

[0060] The original point cloud and the model point cloud are meshed to obtain the original point cloud matrix and the model point cloud matrix, and then the eigenvalues ​​of the original point cloud and the model point cloud are calculated respectively;

[0061] For all points in the original point cloud, the eigenvalues ​​of the points considered and those not considered are calculated respectively, and compared with the eigenvalues ​​of the model point cloud;

[0062] If the original point cloud eigenvalue of the point is far away from the eigenvalue of the model point cloud, and the original point cloud eigenvalue of the point is close to the eigenvalue of the model point cloud, it is regarded as an interference noise point and is removed from the original point cloud domain to ensure the accuracy and reliability of subsequent processing.

[0063] In S2, spatial filtering includes the following steps:

[0064] Spatial filtering selects the coordinates of the outermost points in each direction of the point cloud image to construct the point cloud space. According to all the points in the point cloud space, the coordinates of the center of gravity in this point cloud space are found. The formula is as follows:

[0065] (2)

[0066] In the formula, ( X, Y, Z ) represents the centroid coordinates of the point cloud space, ( x a ,y a ,z a ) is the coordinate of each point cloud in the point cloud space;

[0067] Set point cloud distance threshold D max , calculate the distance from each point in the point cloud space to the centroid coordinates D a ,like D max > D a , then the points in the original point cloud a The invalid interference noise is removed from the original point cloud domain.

[0068] In S4, the iterative closest point registration method includes the following steps:

[0069] The original point cloud of the original point cloud domainA With model point cloud M A , visual point cloud domain S With model point cloud M S , Wavelet Point Cloud Domain H With model point cloud M H As input to the iterative closest point algorithm;

[0070] Model point cloud M A For the previously established electric vehicle charging port point cloud, M A Based on knowledge, data, and spatial filtering M S , M S Perform wavelet transformation to obtain M H , A Based on knowledge, data, and spatial filtering S , S Perform wavelet transformation to obtain H .

[0071] consider J Point cloud transformation method, j The rotation matrix and displacement matrix of the point cloud transformation method are expressed as Q J and U J , for points in the model point cloud in different point cloud domains m a , m s , m j , in the original point cloud A , Visual Point Cloud S , wavelet point cloud H Find the point with the closest Euclidean distance to it a , s , h As corresponding points, they form a corresponding point set;

[0072] Calculate the normal vector angle of the corresponding point pair. If it is greater than the set angle threshold β , then remove this point pair, update the corresponding point set, and use the j A point cloud transformation method for the original point cloud A point a For rotation and displacement changes, the formula is as follows:

[0073] (3)

[0074] In the formula, m a,j For point a Adopt the j The points obtained by rotating and displacing the point cloud using the point cloud transformation method;

[0075] Point a With point m a,j Make a visual point cloud domain change and get the point a With point m a,j Corresponding points on the visual point cloud s With point m s,j , point a With point m a,j Do the wavelet point cloud domain change to get the point a With point m a,j Corresponding points on the wavelet point cloud h With point m h,j;

[0076] The calculation is based on j Registration errors in the original point cloud domain, visual point cloud domain and wavelet point cloud domain of a point cloud transformation method , and score the registration result based on the error and the preset weight. The formula is as follows:

[0077] (4)

[0078] In the formula, are the scoring weights of the original point cloud domain, visual point cloud domain, and wavelet point cloud domain respectively;

[0079] The point cloud change method with the highest registration result score is selected as the optimal point cloud change method, and the corresponding rotation matrix and displacement matrix are used as the final results of the registration to achieve accurate and reliable point cloud registration.

[0080] By performing knowledge, data, and spatial filtering operations on the original point cloud, the original point cloud data is preprocessed based on historical point cloud information, data feature correlation, and point cloud space to obtain an accurate visual point cloud suitable for further processing. While retaining the features, the interference noise in the original point cloud can be effectively removed, facilitating the accurate identification and positioning of the charging port.

[0081] The wavelet point cloud domain is obtained by performing wavelet transformation on the visual point cloud domain. The original point cloud domain, visual point cloud domain, and wavelet point cloud domain are respectively registered based on the iterative nearest point. The registration results are scored based on the errors and weights of different domains, and the best point cloud transformation method is selected to obtain the best rotation matrix and the best displacement matrix, which effectively reduces the point cloud operation error rate of electric vehicle automatic charging technology, solves the problem of accurate identification and positioning of charging ports in complex environments, and realizes reliable point cloud registration.

[0082] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A multi-domain point cloud registration automatic identification and positioning method for electric vehicle charging ports, characterized in that: The following steps are involved: S1: Establish the point cloud domain of the electric vehicle charging port model, and the sensor collects the original point cloud domain of the charging port; S2: Based on the historical point cloud information, the knowledge filtering of the original point cloud domain is realized, based on the data feature correlation, the data filtering of the original point cloud domain is realized, and based on the point cloud space, the spatial filtering of the original point cloud domain is realized to obtain the visual point cloud domain; S3: Perform wavelet transformation on the visual point cloud domain to obtain the wavelet point cloud domain; S4: Perform registration based on iterative closest point in the original point cloud domain, visual point cloud domain, and wavelet point cloud domain, and score the registration results based on the errors and weights of different domains; S5: Select the best point cloud change method, and the corresponding rotation matrix and displacement matrix as the final result of the registration, output the best rotation matrix and the best displacement matrix, and realize accurate and reliable registration of the point cloud; In S2, knowledge filtering includes the following steps: A historical knowledge base is built based on the interference noise information of historical point clouds. The historical knowledge base contains the sparsity of interference noise in different historical point clouds. q The average Euclidean distance and Euclidean distance variance of each noise point are used to calculate the sparsity of the interference noise points. The formula is as follows: (1); In the formula, is the sparseness of interference noise points predicted based on the historical knowledge base, Based on l The sparseness of the interference noise points predicted by the historical point cloud, , Respectively l Interference noise points in historical point cloud k and its neighbors q The average Euclidean distance and Euclidean distance variance of the internal noise points, For the l Interference noise points in historical point cloud k Its neighboring areas i The Euclidean distance of points, For the l The prediction adjustment parameters of the sparsity of interference noise points in the historical point cloud; Calculate the sparsity of each point in the original point cloud ,like , then the points in the original point cloud a The noise points are considered as invalid interference points and are removed from the original point cloud domain; In S2, data filtering includes the following steps: The original point cloud and the model point cloud are meshed to obtain the original point cloud matrix and the model point cloud matrix, and then the eigenvalues ​​of the original point cloud and the model point cloud are calculated respectively; For all points in the original point cloud, the eigenvalues ​​of the points considered and those not considered are calculated respectively, and compared with the eigenvalues ​​of the model point cloud; If the original point cloud eigenvalue of the point is far away from the eigenvalue of the model point cloud, and the original point cloud eigenvalue of the point is not close to the eigenvalue of the model point cloud, it is regarded as an interference noise point and is removed from the original point cloud domain; In S4 and S5, consider J Point cloud transformation method, j The rotation matrix and displacement matrix of the point cloud transformation method are expressed as Q j and U j , for points in the model point cloud in different point cloud domains m a , m s , m h , in the original point cloud A , Visual Point Cloud S , wavelet point cloud H Find the point with the closest Euclidean distance to it a , s , h As corresponding points, they form a corresponding point set; Calculate the normal vector angle of the corresponding point pair. If it is greater than the set angle threshold β , then remove this point pair, update the corresponding point set, and use the j A point cloud transformation method for the original point cloud A point a The formula for rotation and displacement is as follows: (3); In the formula, m a,j For point a Adopt the j The points obtained by rotating and displacing the point cloud using the point cloud transformation method; Point a With point m a,j Make a visual point cloud domain change and get the point a With point m a,j Corresponding points on the visual point cloud s With point m s,j , point a With point m a,j Do wavelet point cloud domain change to get points a With point m a,j Corresponding points on the wavelet point cloud h With point m h,j; The calculation is based on j Registration errors in the original point cloud domain, visual point cloud domain and wavelet point cloud domain of a point cloud transformation method , and score the registration result based on the error and the preset weight. The formula is as follows: (4); In the formula, are the scoring weights of the original point cloud domain, visual point cloud domain, and wavelet point cloud domain respectively; The point cloud change method with the highest registration result score is selected as the optimal point cloud change method, and the corresponding rotation matrix and displacement matrix are used as the final results of the registration to achieve accurate and reliable point cloud registration.

2. The multi-domain point cloud registration automatic identification and positioning method for electric vehicle charging ports according to claim 1 is characterized in that: In S2, the spatial filtering includes the following steps: Select the coordinates of the outermost points in each direction of the point cloud image to construct the point cloud space. According to all the points in the point cloud space, find the coordinates of the center of gravity in this point cloud space. The formula is as follows: (2); In the formula, ( X, Y, Z ) represents the centroid coordinates of the point cloud space, ( x a ,y a ,z a ) is the coordinate of each point cloud in the point cloud space; Set point cloud distance threshold D max , calculate the distance from each point in the point cloud space to the centroid coordinates D a ,like D max > D a , then the points in the original point cloud a The invalid interference noise is removed from the original point cloud domain.

3. The multi-domain point cloud registration automatic identification and positioning method for electric vehicle charging port according to claim 1 is characterized in that: In S4, the iterative closest point registration method comprises the following steps: The original point cloud of the original point cloud domain A With model point cloud M A , visual point cloud domain S With model point cloud M S , Wavelet Point Cloud Domain H With model point cloud M H As input to the iterative closest point algorithm; Model point cloud M A For the previously established electric vehicle charging port point cloud, M A Based on knowledge, data, and spatial filtering M S , M S Perform wavelet transformation to obtain M H , A Based on knowledge, data, and spatial filtering S , S Perform wavelet transformation to obtain H .

Citation Information

Patent Citations

  • Point cloud registration method based on topological structure and multi-scale features

    CN113674332A

  • Point cloud registration method based on point neighborhood condition constraint

    CN118037788A

  • Point cloud data extraction method, device, system and equipment and storage medium

    CN114898100A

  • Hyperspectral image classification method based on adaptive filter and feature complementation network

    CN117218402A

  • Method for automatically detecting building structure and generating 3D model based on laser radar

    WO2019242174A1