CN-121767892-B - Unmanned aerial vehicle monitoring method and system for high slope displacement

CN121767892BCN 121767892 BCN121767892 BCN 121767892BCN-121767892-B

Abstract

The invention relates to the technical field of image processing, in particular to an unmanned aerial vehicle monitoring method and system for high slope displacement, which solve the technical problems that in the prior art, point cloud density is increased rapidly and geometric characteristic distortion occurs due to high slope displacement areas, mismatching of point pairs is easy to occur, and accuracy of point cloud matching in the high slope displacement monitoring process is affected. The method comprises the steps of obtaining point cloud data of a high slope area to be monitored in a plurality of periods, analyzing laser reflection intensity and geometric form characteristics of a plurality of points in the high slope area according to the point cloud data of the plurality of periods, determining displacement complexity of the high slope area, wherein the displacement complexity is used for representing the scattering degree of the point cloud caused by displacement in the area, adjusting the learning rate of a point cloud matching algorithm in an iterative process according to the displacement complexity, matching the point cloud of the high slope area in the plurality of periods by using the point cloud matching algorithm with the learning rate adjusted, and outputting a displacement monitoring result of the high slope area according to the matching result.

Inventors

REN WEI
YANG CHUNYAN
Qi Rusong
WANG ZISHUAI
JIA HONGFEI
HE ZIGUANG
Gao Dejuan
WEI ZHENJIE
KONG XIANGTAI
LI GUOJUN
ZHU LIZHEN
BAN KUN

Assignees

中科斌港建设集团有限公司

Dates

Publication Date: 20260508
Application Date: 20260304

Claims (7)

1. The unmanned aerial vehicle monitoring method for the high slope displacement is characterized by comprising the following steps of: Acquiring point cloud data of a high slope region to be monitored in a plurality of periods; Determining local intensity, local density and local divergence of each point according to corresponding neighborhood point clouds aiming at each point in a high side slope region, wherein the local intensity is used for representing the discrete degree of laser reflection intensity in a neighborhood, the local density is used for representing the distribution density degree of the points in the neighborhood in a geometric form feature, and the local divergence is used for representing the change intensity degree of a displacement field in the neighborhood in the geometric form feature; Clustering a plurality of points in a high slope area according to the three-dimensional coordinates, the local intensity, the local density and the local divergence of each point, and dividing the points into at least one displacement area and at least one non-displacement area; according to the local intensity in the displacement region, a local intensity sequence is formed according to the spatial position sequence to extract frequency spectrum characteristics, and the first characteristics are determined by combining the local intensity difference between the displacement region and the non-displacement region; Carrying out nonlinear feature extraction according to the local intensity, the local density and the local divergence in the displacement region to obtain a second feature; determining the displacement complexity of the high slope region according to the first characteristic and the second characteristic, wherein the displacement complexity is used for representing the point cloud scattering degree caused by displacement in the region; Combining the proportional relation between the current iteration times and the total iteration times of the point cloud matching algorithm, dynamically adjusting the learning rate of the point cloud matching algorithm according to the displacement complexity, wherein the adjusted learning rate is reduced along with the increase of the point cloud scattering degree represented by the displacement complexity, and the dynamic adjustment formula is as follows: representing that the point cloud matching algorithm is at the first The learning rate after adjustment in the time of iteration, Representing that the point cloud matching algorithm is at the first The original learning rate at the time of the iteration, Representing the number of current iterations and, Representing the total number of iterations, e represents a natural exponential function, Representing the displacement complexity; And matching the point clouds of the high slope region in a plurality of periods by using a point cloud matching algorithm after the learning rate is adjusted, and outputting a displacement monitoring result of the high slope region according to the matching result.
2. The unmanned aerial vehicle monitoring method of claim 1, wherein for each point in the high-side slope region, determining the local intensity, local density, and local divergence of each point from the corresponding neighborhood point cloud comprises: determining the local intensity of each point according to the laser reflection intensity of a plurality of points in the neighborhood; Determining the local density of each point according to the distance between a plurality of points in the neighborhood; and carrying out preliminary matching on the point clouds in a plurality of periods, obtaining the displacement vector of each point in the high side slope region, and determining the local divergence of each point according to the fitting relation between the three-dimensional coordinates of a plurality of points in the neighborhood and the displacement vector.
3. The unmanned aerial vehicle monitoring method of claim 1, wherein the nonlinear feature extraction according to the local intensity, local density and local divergence in the displacement region, to obtain the second feature, comprises: Constructing a feature matrix of each displacement region according to the normalization results of the local density, the local divergence and the local intensity of the midpoint of each displacement region; extracting nonlinear characteristics of a characteristic matrix of each displacement area to obtain an embedded matrix, and determining the global scattering degree of the embedded matrix, wherein the global scattering degree comprises any one of a variation coefficient of a distance between points in the matrix, a distribution entropy or a difference degree with a preset uniform distribution matrix; and determining the average value of the global scattering degree corresponding to at least one displacement region as the second characteristic.
4. The unmanned aerial vehicle monitoring method of claim 1, wherein clustering the plurality of points in the high slope region into at least one displaced region and at least one non-displaced region according to the three-dimensional coordinates, the local intensity, the local density, and the local divergence of each point comprises: Clustering according to the normalized results of the three-dimensional coordinates, the local intensity, the local density and the local divergence of each point to obtain a plurality of clustering clusters; And determining an integrated threshold according to the normalization results of the local densities and the local divergences of the points in the plurality of clusters, so as to divide each cluster into a displacement area or an undisplaced area.
5. The unmanned aerial vehicle monitoring method of claim 1, wherein obtaining point cloud data for a high slope area to be monitored over a plurality of time periods comprises: flying along a preset path by using an unmanned aerial vehicle carrying a laser radar in each period, and collecting the original data of a high slope area; And filtering and denoising the acquired original data to obtain point cloud data.
6. The unmanned aerial vehicle monitoring method of claim 5, wherein outputting the displacement monitoring result of the high slope region according to the matching result comprises: Obtaining translation vectors among each group of matching points in different periods, and determining the modular length of the translation vectors; If the modulus of the translation vector is larger than a preset displacement threshold, determining that the position corresponding to the matching point is subjected to dumping damage; If the module length of the translation vector is smaller than or equal to a preset displacement threshold value, determining that the position corresponding to the matching point is not toppled and damaged.
7. The unmanned aerial vehicle monitoring system for the high slope displacement is characterized by comprising a data acquisition module, an algorithm optimization module and a displacement monitoring module; The data acquisition module is used for acquiring point cloud data of the high slope area to be monitored in a plurality of periods; The algorithm optimization module is used for determining local intensity, local density and local divergence of each point in a high slope area according to corresponding neighborhood point clouds, wherein the local intensity is used for representing the discrete degree of laser reflection intensity in the neighborhood, the local density is used for representing the distribution density degree of the points in the neighborhood in geometric form characteristics, the local divergence is used for representing the change intensity degree of a displacement field in the neighborhood in the geometric form characteristics, clustering is carried out on a plurality of points in the high slope area according to the three-dimensional coordinates, the local intensity, the local density and the local divergence of each point, the clustering is divided into at least one displacement area and at least one non-displacement area, a local intensity sequence is formed according to the local intensity in the displacement area in a sequence of spatial positions, spectral feature extraction is carried out, the local intensity difference between the displacement area and the non-displacement area is combined, a first feature is determined, nonlinear feature extraction is carried out according to the local intensity, the local density and the local divergence in the displacement area, a second feature is obtained, the displacement complexity of the high slope area is determined according to the first feature and the second feature, and the displacement complexity of the displacement area is used for representing the displacement degree of the points caused by disorder of the displacement area; The algorithm optimization module is further used for dynamically adjusting the learning rate of the point cloud matching algorithm according to the displacement complexity by combining the proportional relation between the current iteration times and the total iteration times of the point cloud matching algorithm, wherein the adjusted learning rate is reduced along with the increase of the point cloud scattering degree represented by the displacement complexity, and the dynamic adjustment formula is as follows: representing that the point cloud matching algorithm is at the first The learning rate after adjustment in the time of iteration, Representing that the point cloud matching algorithm is at the first The original learning rate at the time of the iteration, Representing the number of current iterations and, Representing the total number of iterations, e represents a natural exponential function, Representing the displacement complexity; the displacement monitoring module is used for matching point clouds of the high slope area in a plurality of periods by using a point cloud matching algorithm after the learning rate is adjusted, and outputting a displacement monitoring result of the high slope area according to the matching result.

Description

Unmanned aerial vehicle monitoring method and system for high slope displacement Technical Field The invention relates to the technical field of image processing, in particular to an unmanned aerial vehicle monitoring method and system for high slope displacement. Background With the rapid development of economy, the construction process of public infrastructures such as highways, bridges and the like is continuously accelerated, frequent geological activities such as landslide and debris flow of mountain bodies in high-mountain dense construction areas bring remarkable influence to the operation of the rocky high-slope, and the safety guarantee of constructors and the protection of building materials face serious challenges. Therefore, before and during construction, accurate displacement monitoring and stability analysis are required for the rock high slope, so that dangerous information such as deformation, looseness and the like of the rock high slope can be timely obtained, disaster early warning is sent out, and loss of manpower and material resources is avoided. In the concrete technical practice of rock high slope displacement monitoring, one of the key links is to compare and analyze three-dimensional point cloud data of a slope in different periods, and point cloud matching is generally achieved by minimizing total distances between all point pairs by adopting ICP series algorithms such as iteration closest point (ITERATIVE CLOSEST POINT, ICP), normal iteration closest point (normal iterative closest point, NICP) and the like. However, as the rock high slope displacement area often presents special deformation characteristics such as structural surface tilting, a large amount of new point cloud data is generated by broken and loosened rocks, the density of the point cloud is increased rapidly, and geometrical characteristic distortions such as curved surfaces and straight lines are generated when a plane is changed, so that a local optimal solution is easily trapped in the point cloud matching process, namely one point of the displacement area of the source point cloud is likely to be erroneously matched with another new point which is generated due to collapse in the target point cloud, and the point is closest to the new point but the new point is not really corresponding to the point. Finally, accuracy of point cloud matching in the three-dimensional point cloud data comparison process is affected, and accuracy and reliability of monitoring results are further affected. Disclosure of Invention In order to solve the technical problems that in the prior art, point cloud density is increased rapidly and geometric characteristics are distorted due to high slope displacement areas, error point pair matching is easy to occur, and accuracy of point cloud matching in a high slope displacement monitoring process is affected, the invention aims to provide an unmanned aerial vehicle monitoring method and system for high slope displacement, and the adopted technical scheme is as follows: According to the method, point cloud data of a high slope area to be monitored in multiple periods are obtained, laser reflection intensity and geometric form features of multiple points in the high slope area are analyzed according to the point cloud data in the multiple periods, displacement complexity of the high slope area is determined, the displacement complexity is used for representing the point cloud scattering degree caused by displacement in the area, the learning rate of a point cloud matching algorithm in an iterative process is adjusted according to the displacement complexity, the point cloud matching algorithm after the learning rate is adjusted is used for matching the point cloud of the high slope area in the multiple periods, and the displacement monitoring result of the high slope area is output according to the matching result. Based on the technical scheme, in the unmanned aerial vehicle monitoring method for high slope displacement, provided by the invention, the displacement complexity capable of quantitatively representing the scattering degree of point clouds caused by displacement is constructed by analyzing the laser reflection intensity and the geometric form characteristics in the multi-period point cloud data, and the iterative learning rate of a point cloud matching algorithm is dynamically adjusted based on the complexity, so that the technical barrier that the algorithm is easy to fall into a local optimal solution when matching high-density and high-scattering point clouds generated by slope displacement is effectively overcome, and the accuracy and the convergence stability of point cloud matching and the accuracy and the reliability of a final displacement monitoring result are obviously improved. With reference to the first aspect, in one possible implementation manner, the method for determining the displacement complexity of the high slope area by analyzing the laser reflection inten