CN-122020209-A - Method and device for global alignment of coordinate data oriented to inspection scene

CN122020209ACN 122020209 ACN122020209 ACN 122020209ACN-122020209-A

Abstract

The embodiment of the application provides a method and a device for globally aligning coordinate data facing a patrol scene, which are used for constructing a feature description vector set through filtering noise reduction and local curvature key point screening, gradually screening seed corresponding relations by combining space distance constraint and normal vector consistency constraint, constructing a geometric compatibility matrix, extracting a global consistency weight sequence through eigenvalue decomposition to construct a weighted covariance matrix, executing singular value decomposition to solve a rotation matrix and a translation vector to finish corridor coordinate alignment, and finally outputting over-limit vegetation positioning information for clipping operation and calling, thereby effectively solving the defects of the traditional technology in aspects of coordinate data preprocessing and feature extraction, mismatching Lu Bangshai selection and geometric compatibility modeling, global consistency weight solving and alignment result output and the like, and providing technical guarantee for intelligent global alignment of coordinate data and over-limit vegetation accurate positioning in the power transmission unmanned aerial vehicle patrol scene.

Inventors

HU YARONG
LIU YANG
YANG WENNA
MA XIAOLIN
LIN XUMING
MA HONGXIA
Yang Xunfa
ZHU HONG
HAN ZHONGFU
WU YANHUA
MA QIAN

Assignees

国网甘肃省电力公司临夏供电公司

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (8)

1. The method for globally aligning coordinate data oriented to a patrol scene is characterized by comprising the following steps: Acquiring a source coordinate data set and a reference coordinate data set which are acquired by an unmanned aerial vehicle on a power transmission corridor, performing filtering noise reduction and downsampling on the source coordinate data set and the reference coordinate data set to obtain a standardized corridor coordinate data set, screening key data points from the standardized corridor coordinate data set based on a local curvature threshold condition, and calculating a feature descriptor to obtain a feature description vector set; Performing cross-data set similarity matching on the feature description vector set to obtain an initial corresponding relation set, sequentially performing space distance constraint screening and normal vector consistency constraint screening on the initial corresponding relation set to obtain a seed corresponding relation set, and calculating compatibility scores for each matching pair in the seed corresponding relation set based on rigid transformation distance maintenance properties to obtain a geometric compatibility matrix; And performing eigenvalue decomposition on the geometric compatibility matrix, extracting a main eigenvector to obtain a global consistency weight sequence, constructing a weighted covariance matrix by taking the global consistency weight sequence as a coefficient, performing singular value decomposition to obtain a rotation matrix and a translation vector, applying the rotation matrix and the translation vector to the source coordinate data set to finish corridor coordinate alignment, and outputting out-of-limit vegetation positioning information for cutting operation call.
2. The global alignment method of coordinate data for a patrol scene according to claim 1, wherein the acquiring a source coordinate data set and a reference coordinate data set acquired by an unmanned aerial vehicle on a power transmission corridor, performing filtering noise reduction and downsampling on the source coordinate data set and the reference coordinate data set to obtain a standardized corridor coordinate data set, comprises: Calculating the average space distance between each data point in the source coordinate data set and the reference coordinate data set and the nearest neighbor data point in the neighborhood to obtain a neighborhood distance value, comparing the neighborhood distance value with the global average distance, judging a noise point based on a standard deviation threshold condition, and eliminating the data point judged as the noise point from the source coordinate data set and the reference coordinate data set to obtain a noise reduction coordinate data set; dividing the three-dimensional space covered by the noise reduction coordinate data set into uniform voxel grid units, calculating mass center coordinates of all data points in each voxel grid unit to obtain grid mass center points, and converging each grid mass center point to form a standardized corridor coordinate data set.
3. The method of global alignment of coordinate data for a patrol scene as recited in claim 1, wherein said screening key data points from said standardized corridor coordinate data set and computing feature descriptors based on local curvature threshold conditions to obtain a feature description vector set comprises: extracting a neighborhood point set from each data point in the standardized corridor coordinate data set, constructing a neighborhood covariance matrix, performing eigenvalue decomposition on the neighborhood covariance matrix to obtain a minimum eigenvalue and a maximum eigenvalue, calculating the ratio of the minimum eigenvalue to the maximum eigenvalue to obtain a local curvature value, comparing the local curvature value with a preset local curvature threshold condition, and screening the data points with curvature values meeting the threshold condition as key data points; establishing a local reference coordinate system by taking each key data point as a center, dividing a neighborhood space into a plurality of subareas, obtaining a subarea statistical histogram by counting radial distance distribution, normal vector angular distribution and relative height distribution of the data points in each subarea, splicing each subarea statistical histogram according to a fixed sequence to form feature description vectors, and converging the feature description vectors corresponding to all the key data points to obtain a feature description vector set.
4. The global alignment method of coordinate data for a patrol scene according to claim 1, wherein said performing cross-dataset similarity matching on the feature description vector set to obtain an initial correspondence set, sequentially performing spatial distance constraint screening and normal vector consistency constraint screening on the initial correspondence set to obtain a seed correspondence set, includes: Searching the feature description vector of each key data point in the source coordinate data set for the feature description vector with the minimum Euclidean distance in the feature description vector set of the reference coordinate data set to obtain nearest neighbor matching and secondary neighbor matching, calculating the ratio of the distance between the nearest neighbor matching and the secondary neighbor matching to obtain a distance ratio, comparing the distance ratio with a preset ratio inspection threshold condition, and reserving the matching pair meeting the threshold condition to form an initial corresponding relation set; And (3) calculating the space Euclidean distance between the source data point and the target data point for each matching pair in the initial corresponding relation set to obtain a matching space distance, comparing the matching space distance with a preset maximum displacement threshold condition, removing the matching pair exceeding the threshold condition to obtain a distance-screened set, calculating the included angle between the normal vector of the source data point and the normal vector of the target data point for each matching pair in the distance-screened set to obtain a normal vector included angle value, comparing the normal vector included angle value with the preset included angle threshold condition, and reserving the matching pair meeting the threshold condition to form a seed corresponding relation set.
5. The global alignment method of coordinate data for a patrol scene according to claim 1, wherein calculating compatibility scores for each matching pair in the set of seed correspondences based on rigid body transformation distance preserving properties to obtain a geometric compatibility matrix comprises: Calculating the space distance between two corresponding points in a source coordinate data set to obtain a source end distance, calculating the space distance between two corresponding points in a reference coordinate data set to obtain a target end distance, and taking an absolute value as a distance difference value by taking the difference between the source end distance and the target end distance; And inputting the distance difference value into a kernel function which monotonically decreases along with the distance difference, calculating to obtain a compatibility score, writing all compatibility scores between every two matched pairs in the seed corresponding relation set into a symmetric matrix according to a row-column index to obtain a geometric compatibility matrix, and setting diagonal elements of the geometric compatibility matrix as maximum values to represent that each matched pair is completely compatible with the self.
6. The global alignment method of coordinate data for a patrol scene according to claim 1, wherein said performing eigenvalue decomposition on the geometric compatibility matrix and extracting a principal eigenvector to obtain a global consistency weight sequence, constructing a weighted covariance matrix with the global consistency weight sequence as a coefficient and performing singular value decomposition to obtain a rotation matrix and a translation vector, comprises: Performing eigenvalue decomposition on the geometric compatibility matrix to obtain an eigenvalue set and a corresponding eigenvector set, extracting a maximum eigenvalue from the eigenvalue set, reading an eigenvector corresponding to the maximum eigenvalue as a main eigenvector, and normalizing each component in the main eigenvector to a zero-to-one interval to obtain a global consistency weight sequence; And taking each weight value in the global consistency weight sequence as a coefficient of a corresponding matching pair, constructing a weighted cross covariance matrix based on a source data point coordinate and a target data point coordinate in a seed corresponding relation set by the coefficient, executing singular value decomposition on the weighted cross covariance matrix to obtain a rotation matrix, and calculating based on the rotation matrix and the weighted centroid coordinate to obtain a translation vector.
7. The global alignment method of coordinate data for a patrol scene according to claim 1, wherein said applying the rotation matrix and the translation vector to the source coordinate data set completes corridor coordinate alignment and outputs out-of-limit vegetation positioning information for a crop operation call, comprising: Performing matrix multiplication operation on each data point coordinate in the source coordinate data set and the rotation matrix to obtain a rotated coordinate, performing vector addition operation on the rotated coordinate and the translation vector to obtain a transformed coordinate, and converging all the transformed coordinates to form an aligned corridor coordinate data set; Identifying vegetation data points and lead data points from the aligned corridor coordinate data set, calculating the space distance between each vegetation data point and the nearest lead data point to obtain a tree line distance value, comparing the tree line distance value with a preset safety distance threshold condition, screening vegetation data points smaller than the threshold condition as overrun vegetation points, extracting the space coordinates of each overrun vegetation point to form overrun vegetation positioning information, and outputting the overrun vegetation positioning information for planning and calling of an unmanned aerial vehicle cutting operation path.
8. A global alignment device for coordinate data of a patrol scene, the device comprising: The data screening module is used for acquiring a source coordinate data set and a reference coordinate data set which are acquired by the unmanned aerial vehicle on a power transmission corridor, performing filtering noise reduction and downsampling on the source coordinate data set and the reference coordinate data set to obtain a standardized corridor coordinate data set, screening key data points from the standardized corridor coordinate data set based on a local curvature threshold condition, and calculating a feature descriptor to obtain a feature description vector set; The data alignment module is used for performing cross-data set similarity matching on the characteristic description vector set to obtain an initial corresponding relation set, sequentially performing space distance constraint screening and normal vector consistency constraint screening on the initial corresponding relation set to obtain a seed corresponding relation set, and calculating compatibility scores on each matching pair in the seed corresponding relation set based on rigid transformation distance maintenance properties to obtain a geometric compatibility matrix; And the pruning operation module is used for executing eigenvalue decomposition on the geometric compatibility matrix, extracting a main eigenvector to obtain a global consistency weight sequence, constructing a weighted covariance matrix by taking the global consistency weight sequence as a coefficient, executing singular value decomposition to obtain a rotation matrix and a translation vector, acting the rotation matrix and the translation vector on the source coordinate data set to finish corridor coordinate alignment, and outputting out-of-limit vegetation positioning information for the calling of the pruning operation.

Description

Method and device for global alignment of coordinate data oriented to inspection scene Technical Field The application relates to the field of electronic digital data processing, in particular to a method and a device for global alignment of coordinate data for a patrol scene. Background The existing global alignment method for the coordinates data of the inspection scene has obvious defects. The traditional system is poor in preprocessing and feature extraction of unmanned aerial vehicle acquired data, generally lacks the capability of performing systematic filtering noise reduction and downsampling on a source coordinate data set and a reference coordinate data set, fails to effectively screen key data points from a standardized corridor coordinate data set based on local curvature threshold conditions and calculate feature descriptors to obtain a feature description vector set, causes insufficient noise suppression capability and key feature point extraction precision of the coordinate data under the complex ground object background of a power transmission corridor, and is difficult to provide a high-quality feature expression basis for subsequent cross-data set matching. In addition, the prior art has a bottleneck in terms of initial correspondence screening and geometric compatibility modeling. Most systems lack the capability of sequentially executing space distance constraint screening and normal vector consistency constraint screening on an initial corresponding relation set to obtain a seed corresponding relation set, calculating compatibility scores for all matching pairs based on rigid transformation distance retention properties and constructing a geometric compatibility matrix, so that mismatching causes larger interference on subsequent pose estimation, the robustness and accuracy of global alignment are affected, and the system is particularly more prominent in actual inspection scenes with dense repetitive structures of power transmission corridors and unstable point cloud overlapping rate. The existing system has a technical short plate in the aspects of global consistency weight solving and coordinate alignment output based on a geometric consistency matrix. The method comprises the steps of performing eigenvalue decomposition on a geometric compatibility matrix to extract a main eigenvector to obtain a global consistency weight sequence, constructing a weighted covariance matrix based on the overall consistency weight sequence, performing singular value decomposition to solve a rotation matrix and a translation vector, further completing corridor coordinate alignment and outputting out-of-limit vegetation positioning information for a complete closed loop alignment execution mechanism called by cutting operation, and influencing the coordination capability of a patrol system on the accurate positioning of out-of-limit vegetation targets and subsequent intelligent cutting operation. The method has important significance for improving the accuracy and the intelligent level of global alignment of the coordinate data in the inspection scene of the unmanned aerial vehicle in the power transmission corridor. Disclosure of Invention Aiming at the problems in the prior art, the application provides a global alignment method and device for coordinate data oriented to a patrol scene, which can effectively solve the defects of the traditional technology in aspects of coordinate data preprocessing, feature extraction, mismatching Lu Bangshai selection, geometric compatibility modeling, global consistency weight solving, alignment result output and the like, and provide technical guarantee for intelligent global alignment of the coordinate data and accurate positioning of overrun vegetation in the patrol scene of a power transmission corridor unmanned aerial vehicle. In order to solve at least one of the problems, the application provides the following technical scheme: in a first aspect, the present application provides a method for global alignment of coordinate data for a patrol scene, including: Acquiring a source coordinate data set and a reference coordinate data set which are acquired by an unmanned aerial vehicle on a power transmission corridor, performing filtering noise reduction and downsampling on the source coordinate data set and the reference coordinate data set to obtain a standardized corridor coordinate data set, screening key data points from the standardized corridor coordinate data set based on a local curvature threshold condition, and calculating a feature descriptor to obtain a feature description vector set; Performing cross-data set similarity matching on the feature description vector set to obtain an initial corresponding relation set, sequentially performing space distance constraint screening and normal vector consistency constraint screening on the initial corresponding relation set to obtain a seed corresponding relation set, and calculating compatibility scores for ea