Search

CN-122023705-A - Unmanned plane point cloud fusion refined modeling method for slope structural surface

CN122023705ACN 122023705 ACN122023705 ACN 122023705ACN-122023705-A

Abstract

The invention relates to the technical field of slope engineering modeling and discloses a slope structural plane unmanned aerial vehicle point cloud fusion refined modeling method which comprises the steps of acquiring slope global point cloud data and optical images through an unmanned aerial vehicle carrying high-precision positioning system, combining local point cloud data acquired by a handheld laser scanner, determining a point cloud normal vector based on a local plane fitting technology after calibration and preprocessing, extracting structural boundary information through a normal direction difference boundary detection algorithm, completing structural plane clustering segmentation through a region growing algorithm, finally generating a triangular grid model, performing texture mapping and level simplification, and realizing the refined modeling of a slope structural plane. The invention improves the modeling precision and efficiency, and provides reliable model support for stability analysis and risk assessment of slope engineering.

Inventors

  • CHEN KEJING

Assignees

  • 南昌航空大学

Dates

Publication Date
20260512
Application Date
20251224

Claims (10)

  1. 1. The unmanned aerial vehicle point cloud fusion refined modeling method for the slope structural surface is characterized by comprising the following steps of: carrying a high-precision positioning system by an unmanned aerial vehicle, scanning a slope area according to a preset route to obtain global point cloud data and an optical image of the slope area, and carrying out fine scanning on the slope area by a handheld laser scanner to obtain local point cloud data of the slope area; Calibrating the global point cloud data and the point cloud data to obtain point cloud data of a slope region, and preprocessing the point cloud data, wherein the preprocessing comprises denoising processing, point cloud matching processing and point cloud simplifying processing; Determining a normal vector of each point in the point cloud data based on a local plane fitting technology, determining structural boundary information of a side slope region based on a normal difference boundary detection algorithm, and carrying out cluster segmentation on a structural surface of the side slope region by adopting a region growing algorithm according to the structural boundary information; based on the point cloud data of the segmented structural surface, generating a triangular mesh model, performing texture mapping on the optical image and the triangular mesh model, establishing a pixel-level corresponding relation between the optical image and the triangular mesh model, and performing hierarchical simplification on the triangular mesh model after mapping to realize modeling of the slope structural surface.
  2. 2. The unmanned aerial vehicle point cloud fusion refined modeling method of the slope structural surface according to claim 1, wherein the global point cloud data and the local point cloud data are calibrated to obtain point cloud data of a slope region, specifically: The high-precision positioning system comprises an optical camera and a laser radar; And (3) adopting a standard target calibration field, and reversely calculating the mounting postures and the range errors of the laser radar and the handheld laser scanner according to the known target positions by utilizing the reflection target array to realize calibration.
  3. 3. The unmanned aerial vehicle point cloud fusion refined modeling method for the side slope structural surface according to claim 1 is characterized in that when denoising processing is carried out, the radius of the field is set, the point density of each point in the neighborhood is calculated, if the point number of a certain point is lower than a point density threshold value, the corresponding point is judged to be a noise point and deleted, and the radius of the field is 0.5 meter.
  4. 4. The unmanned aerial vehicle point cloud fusion refined modeling method for the slope structural surface according to claim 1, wherein when point cloud matching processing is carried out, an iterative nearest point algorithm is adopted, an initial corresponding point set is selected, point cloud characteristic points of not less than 50 pairs are randomly sampled, the maximum step length of each iteration is limited to 0.1 meter, the convergence error margin is set to 1 millimeter, and global optimization is carried out once every 5 iterations.
  5. 5. The unmanned aerial vehicle point cloud fusion refined modeling method of the slope structural surface according to claim 1, wherein when the point cloud compaction is carried out, a voxel grid method is adopted to carry out point cloud compaction, the size of a voxel grid is set to be 0.2m x 0.2m, a point cloud space is divided into uniform cube units, and only one representative point is reserved in each voxel; and combining a curvature criterion, calculating the curvature by adopting a quadric surface fitting method, wherein the curvature threshold is divided into three grades, namely a high grade, a medium grade and a low grade, and the three grades respectively correspond to different compaction proportions so as to realize point cloud compaction.
  6. 6. The method for accurately modeling the unmanned aerial vehicle point cloud fusion of the slope structural surface according to claim 1 is characterized by determining the normal vector of each point in the point cloud data based on a local plane fitting technology, specifically, selecting a neighborhood point set in a certain radius range by taking a target point as the center, wherein the initial value of the neighborhood radius of the local plane fitting is 1 meter, decreasing according to an exponential function along with the increase of curvature, and the minimum is reduced to 0.3 meter, fitting a required plane through a least square method, and setting the normal vector of the plane as the normal vector of the target point.
  7. 7. The unmanned aerial vehicle point cloud fusion refined modeling method for the side slope structural surface is characterized by adopting an area growing algorithm to conduct clustering segmentation on the structural surface of the side slope area according to structural boundary information, specifically adopting an area growing algorithm to conduct clustering segmentation on the structural surface according to the structural boundary information and the spatial connectivity and similarity of point clouds, gradually bringing adjacent points meeting the conditions into the same structural surface area according to a preset growing criterion from seed points, recalculating the statistical characteristics of the area once every 20 points are added in the growing process, updating the statistical characteristics of the area in real time, and correcting the growing direction, wherein the distance threshold of the growing criterion is 0.3 meter, and the maximum included angle deviation of normal vectors is 15 degrees.
  8. 8. The unmanned aerial vehicle point cloud fusion refined modeling method for the side slope structural surface according to claim 1 is characterized in that a triangular mesh model is generated based on the point cloud data of the structural surface after segmentation, specifically, the triangular mesh model is generated by applying a delusie triangulation algorithm based on the point cloud data of the structural surface after segmentation, and constraint conditions of the delusie triangulation algorithm comprise empty circle characteristics and side length limitation, wherein the length of the longest side cannot exceed 3 times of the shortest side.
  9. 9. The unmanned aerial vehicle point cloud fusion refined modeling method for the side slope structural surface is characterized by comprising the steps of performing texture mapping on an optical image and a triangular mesh model, and establishing a pixel-level corresponding relation between the optical image and the triangular mesh model, wherein texture coordinates of the optical image are back projected to the surface of the triangular mesh model, seamless fusion and correction are performed on textures by using a Poisson editing or Laplace weight image processing technology, joint and blurring phenomena are eliminated, the solution precision of a Poisson editing Laplace equation is controlled within 0.001, and a Laplace weight coefficient is set to be 1.
  10. 10. The unmanned aerial vehicle point cloud fusion refined modeling method for the side slope structural surface according to claim 1 is characterized in that the mapped triangular mesh model is subjected to hierarchical simplification, specifically, a side folding algorithm is adopted, folding priority is determined according to the length of a side, the normal direction difference of two end points and the area change of an adjacent triangle, weight distribution is respectively set to 40%, 30% and 30%, and redundant triangles are combined; The vertex caching technology is adopted, the vertex storage sequence is organized, the cache block size is set to be 16 bytes, the cache content is managed by adopting the LRU replacement strategy, invalid data is cleaned regularly, so that the cache hit rate during graphic rendering is improved, and the model loading and display speed is accelerated.

Description

Unmanned plane point cloud fusion refined modeling method for slope structural surface Technical Field The invention relates to the technical field of slope engineering modeling, in particular to a point cloud fusion refined modeling method for a slope structural surface unmanned plane. Background The geometric form of the slope structural surface is a key factor influencing the stability of the slope, and accurately obtaining the three-dimensional model of the slope structural surface has important significance for the design, construction and risk assessment of slope engineering. The traditional slope modeling method mostly adopts contact type measuring equipment such as total stations, level gauges and the like, so that the measuring efficiency is low, the labor intensity is high, and the measuring precision is difficult to guarantee for slope areas which are complex in terrain and difficult to reach. Along with the development of unmanned aerial vehicle technology and laser scanning technology, unmanned aerial vehicle point cloud measurement technology is widely applied to the slope modeling field because of the advantages of high efficiency, non-contact, large-scale coverage and the like. However, the problem of insufficient precision of single unmanned aerial vehicle point cloud data on local details is difficult to accurately reflect the fine characteristics of the slope structural surface, and the handheld laser scanner can acquire high-precision local point cloud data, but has limited measurement range and cannot realize rapid scanning of the whole slope area. In addition, how to effectively remove noise, realize accurate matching and compaction of point cloud, and how to accurately segment a structural plane and generate a high-quality three-dimensional model based on point cloud data in the processing process of the point cloud data still remains a main challenge faced by the current slope modeling technology. Therefore, a method for realizing the fine modeling of the side slope structural surface by integrating the advantages of the global point cloud of the unmanned aerial vehicle and the handheld laser scanning local point cloud data is needed, so as to solve the problems that the measurement precision and the efficiency are difficult to consider, the model detail expression is insufficient and the like in the prior art. Disclosure of Invention The invention aims to provide a point cloud fusion refined modeling method for a slope structural plane unmanned aerial vehicle, which aims to solve one or more of the problems. The invention provides a point cloud fusion refined modeling method of a slope structural plane unmanned aerial vehicle, which comprises the following steps: carrying a high-precision positioning system by an unmanned aerial vehicle, scanning a slope area according to a preset route to obtain global point cloud data and an optical image of the slope area, and carrying out fine scanning on the slope area by a handheld laser scanner to obtain local point cloud data of the slope area; Calibrating the global point cloud data and the point cloud data to obtain point cloud data of a slope region, and preprocessing the point cloud data, wherein the preprocessing comprises denoising processing, point cloud matching processing and point cloud simplifying processing; Determining a normal vector of each point in the point cloud data based on a local plane fitting technology, determining structural boundary information of a side slope region based on a normal difference boundary detection algorithm, and carrying out cluster segmentation on a structural surface of the side slope region by adopting a region growing algorithm according to the structural boundary information; based on the point cloud data of the segmented structural surface, generating a triangular mesh model, performing texture mapping on the optical image and the triangular mesh model, establishing a pixel-level corresponding relation between the optical image and the triangular mesh model, and performing hierarchical simplification on the triangular mesh model after mapping to realize modeling of the slope structural surface. Preferably, the global point cloud data and the local point cloud data are calibrated to obtain point cloud data of a slope region, specifically: The high-precision positioning system comprises an optical camera and a laser radar; And (3) adopting a standard target calibration field, and reversely calculating the mounting postures and the range errors of the laser radar and the handheld laser scanner according to the known target positions by utilizing the reflection target array to realize calibration. Preferably, when denoising processing is carried out, setting the radius of the field, calculating the dot density of each dot in the neighborhood, and if the dot number of a certain dot is lower than the dot density threshold value, judging the corresponding dot as a noise dot and deleting the noise dot