CN-122024080-A - Intelligent monitoring method and system for slope surface deformation based on unmanned aerial vehicle oblique photography

CN122024080ACN 122024080 ACN122024080 ACN 122024080ACN-122024080-A

Abstract

The invention discloses an intelligent monitoring method and system for slope surface deformation based on unmanned aerial vehicle oblique photography, and relates to the technical field of slope surface deformation monitoring and image recognition. The method comprises the steps of S1, acquiring multi-stage unmanned aerial vehicle oblique photography multi-view images of a side slope and generating corresponding multi-stage three-dimensional point clouds, and S2, preprocessing the multi-stage three-dimensional point clouds, namely carrying out point cloud registration according to the characteristics of a rock mass structural surface and dividing the point clouds into a plurality of monitoring subareas according to the two dimensions of stratum layering and structural surface blocking. According to the invention, a slope multi-period image is acquired and a three-dimensional point cloud is generated through an unmanned aerial vehicle oblique photography technology, and the problem of limited coverage range of a traditional monitoring means is solved by combining layered variable angle route planning and double-frequency positioning deviation compensation.

Inventors

FAN JIANGTAO
ZHAO JING
LI YUANSHU
XIANG XINYI
XIE MINGHAO
TIAN TIAN
ZHANG CHEN

Assignees

西安航空学院

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (10)

1. Intelligent monitoring method for slope surface deformation based on unmanned aerial vehicle oblique photography is characterized by comprising the following steps: S1, acquiring a multi-stage unmanned aerial vehicle oblique photography multi-view image of a side slope, and generating a corresponding multi-stage three-dimensional point cloud; S2, preprocessing the multi-period three-dimensional point cloud, namely registering the point cloud according to the characteristics of the structural surface of the rock mass, dividing the point cloud into a plurality of monitoring subareas according to the two dimensions of layering and structural surface blocking, and binding lithology labels for the monitoring subareas; s3, extracting geometric features, texture features and time sequence change features aiming at each monitoring subarea to form a three-dimensional visual feature set, wherein the time sequence change features comprise space offset vectors of point clouds in adjacent time periods, and the space offset vectors are decomposed into a slope surface normal component and a slope surface tangential component; S4, calling geomechanical priori parameters corresponding to the monitoring subareas, carrying out mechanical rationality verification on space offset vectors in the three-dimensional visual feature set, removing feature values which do not accord with rock deformation mechanical logic, and generating feature vectors containing geometric, texture, time sequence and mechanical four-dimensional information; S5, inputting the four-dimensional feature vectors into a pre-trained deep learning classification model, and identifying and distinguishing the unstructured wear and the structured slipping deformation.
2. The intelligent monitoring method for slope surface deformation based on unmanned aerial vehicle oblique photography according to claim 1, wherein the acquiring the multi-stage unmanned aerial vehicle oblique photography image of the slope comprises: planning a layered variable angle flight route according to the slope height and the slope of the slope, wherein the layered variable angle flight route comprises a low-altitude near-view layer route and a high-altitude panoramic layer route, and the inclination angle of a lens of the low-altitude near-view layer route is linearly adjusted along with the slope; controlling an unmanned aerial vehicle carrying the five-lens oblique photography tripod head to fly along a layered variable-angle flight route, and synchronously triggering exposure of all lenses to obtain multi-view images; And acquiring the space attitude information of the image by using the double-frequency positioning unit, and carrying out positioning deviation compensation on the space attitude information by combining the known coordinates of the slope periphery datum point.
3. The intelligent monitoring method for slope surface deformation based on unmanned aerial vehicle oblique photography according to claim 1, wherein the preprocessing of the multi-stage three-dimensional point cloud comprises the following steps: Matching the unmanned aerial vehicle oblique photography multi-view images by adopting an improved SIFT feature matching algorithm, wherein the improved SIFT feature matching algorithm increases directional constraint on texture features of rock mass joint and crack areas, and preferentially matches feature points of engineering key areas to generate single-period three-dimensional point clouds, and the multi-period three-dimensional point clouds are composed of a plurality of groups of single-period three-dimensional point clouds generated in different monitoring periods; denoising the single-period three-dimensional point cloud by adopting a mixed denoising algorithm combining statistical filtering and region growth, and reserving continuous point clouds of a rock mass structural plane; and registering the multi-period three-dimensional point cloud by adopting a characteristic point guided ICP registration optimization algorithm, wherein the characteristic point guided ICP registration optimization algorithm takes rock mass structural plane characteristic points extracted from the point cloud of different periods as registration anchor points.
4. The intelligent monitoring method for slope surface deformation based on unmanned aerial vehicle oblique photography according to claim 1, wherein the extracting geometrical features comprises: Calculating the curvature of the point cloud by adopting a weighted neighborhood surface fitting algorithm; calculating elevation variation coefficients of the neighborhood points to represent local elevation dispersion; The neighborhood radius used for calculation is set differently according to lithology labels of the monitoring subareas.
5. The intelligent monitoring method for slope surface deformation based on unmanned aerial vehicle oblique photography according to claim 1, wherein the extraction of the texture features comprises: and (3) adopting a local binary pattern algorithm to encode the texture of the rock mass surface, strengthening the abrupt change characteristics of the texture of the crack and the peeling area, and filtering the ineffective texture of the vegetation coverage area.
6. The intelligent monitoring method for the deformation of the side slope surface based on unmanned aerial vehicle oblique photography according to claim 1, wherein the geomechanical prior parameters comprise the shear strength of the rock mass and the occurrence of structural surface, and the mechanical rationality check comprises the step of judging whether the direction of a spatial offset vector is matched with the occurrence of structural surface and whether the accumulated displacement of the rock mass relative to a datum point in an adjacent monitoring period exceeds the allowable range of the shear strength of the corresponding rock mass.
7. The intelligent monitoring method for the deformation of the side slope surface based on unmanned aerial vehicle oblique photography is characterized in that the deep learning classification model is an improved CNN-LSTM hybrid classification model, a convolution layer is used for extracting spatial features, a long-term memory layer is used for capturing time sequence deformation rules, and the deep learning classification model introduces a difficult sample mining strategy in a training stage so as to strengthen the distinguishing capability of weak slippage and weathering exfoliation.
8. The intelligent monitoring method for deformation of a side slope surface based on unmanned aerial vehicle oblique photography according to claim 1, wherein the identifying and distinguishing between unstructured wear and structured slip deformation further comprises: and calling a Bayesian updated dynamic threshold generation algorithm aiming at the monitoring subarea generating structural sliding deformation, and generating a real-time deformation rate threshold and a cumulative offset threshold through Bayesian posterior probability calculation based on a normal probability distribution function of rock mass mechanical parameters and historical deformation data of the subarea.
9. The intelligent monitoring method for slope surface deformation based on unmanned aerial vehicle oblique photography according to claim 1, wherein the method further comprises: Continuously tracking the deformation rate and the accumulated offset of the monitoring subarea with structural sliding deformation, comparing the deformation rate threshold with the accumulated offset threshold in real time, and executing grading early warning operation according to the comparison result, wherein the grading early warning operation comprises the steps of pushing a deformation monitoring report, triggering an engineering inspection instruction or starting an emergency response flow, and outputting the three-dimensional coordinates and the deformation vector direction of the deformation area.
10. The intelligent monitoring system for the deformation of the side slope surface based on unmanned aerial vehicle oblique photography is applied to the intelligent monitoring method for the deformation of the side slope surface based on unmanned aerial vehicle oblique photography as claimed in any one of claims 1 to 9, and is characterized by comprising the following steps: the unmanned aerial vehicle oblique photography data acquisition module is used for acquiring multi-period unmanned aerial vehicle oblique photography images of the side slope and generating corresponding multi-period three-dimensional point clouds; The side slope three-dimensional point cloud preprocessing module is used for preprocessing multi-stage three-dimensional point cloud, and comprises the steps of registering point cloud according to the characteristics of a rock mass structural surface, dividing the point cloud into a plurality of monitoring subareas according to the two dimensions of stratum layering and structural surface blocking, and binding lithology labels for the monitoring subareas; The multi-feature fusion analysis module is used for extracting geometric features, texture features and time sequence change features for each monitoring subarea to form a three-dimensional visual feature set, calling geomechanical prior parameters corresponding to the monitoring subareas, and carrying out mechanical rationality verification on space offset vectors in the three-dimensional visual feature set to generate feature vectors containing geometric, texture, time sequence and mechanical four-dimensional information; and the deformation intelligent recognition and early warning module is used for inputting four-dimensional feature vectors into a pre-trained deep learning classification model to recognize and distinguish non-structural abrasion and structural slippage deformation.

Description

Intelligent monitoring method and system for slope surface deformation based on unmanned aerial vehicle oblique photography Technical Field The invention relates to the technical field of slope surface deformation monitoring and image recognition, in particular to an intelligent slope surface deformation monitoring method and system based on unmanned aerial vehicle oblique photography. Background Slope surface deformation monitoring is a key link for guaranteeing engineering safety of highways, railways, mines and the like. Traditional monitoring means include total stations and GPS technology, which rely on manual field acquisition of data, and although single-point accurate measurement can be provided, complex slope surfaces are difficult to comprehensively cover. In recent years, unmanned aerial vehicle oblique photography technology rapidly develops, and efficiency and range of side slope surface data acquisition are remarkably improved by shooting high-definition images from different angles and generating a three-dimensional model. The technology can reconstruct the fine geometric form of the slope and provide a rich data base for deformation analysis. However, when slope deformation analysis is performed by using dense three-dimensional point clouds generated by unmanned aerial vehicle oblique photography, the conventional method faces a core problem that geometric shape changes generated by natural weathering or local spalling of the slope surface have signal strength which is highly similar to that of real potential sliding deformation, so that an algorithm based on simple geometric feature comparison is difficult to distinguish stably. The method can easily misjudge the abrasion of the unstructured surface as dangerous deformation, thereby triggering false alarms and interfering engineering judgment. Due to the lack of effective fusion of the geomechanical background of the side slope and the deformation physical mechanism, the existing monitoring means have the defect of identifying early substantial slippage signs, so that deformation signals with engineering significance are difficult to accurately capture in conventional surface fluctuation, and the accuracy and timeliness of early warning are limited. Disclosure of Invention Aiming at the defects of the prior art, the invention provides an intelligent monitoring method and system for slope surface deformation based on unmanned aerial vehicle oblique photography, which solve the problems of limited coverage, easy misjudgment of unstructured wear and structural slippage, high false alarm rate and early warning hysteresis in the prior art. In order to achieve the purpose, the intelligent monitoring method for the slope surface deformation based on unmanned aerial vehicle oblique photography comprises the following steps: S1, acquiring a multi-stage unmanned aerial vehicle oblique photography multi-view image of a side slope, and generating a corresponding multi-stage three-dimensional point cloud; S2, preprocessing the multi-period three-dimensional point cloud, namely registering the point cloud according to the characteristics of the structural surface of the rock mass, dividing the point cloud into a plurality of monitoring subareas according to the two dimensions of layering and structural surface blocking, and binding lithology labels for the monitoring subareas; s3, extracting geometric features, texture features and time sequence change features aiming at each monitoring subarea to form a three-dimensional visual feature set, wherein the time sequence change features comprise space offset vectors of point clouds in adjacent time periods, and the space offset vectors are decomposed into a slope surface normal component and a slope surface tangential component; S4, calling geomechanical priori parameters corresponding to the monitoring subareas, carrying out mechanical rationality verification on space offset vectors in the three-dimensional visual feature set, removing feature values which do not accord with rock deformation mechanical logic, and generating feature vectors containing geometric, texture, time sequence and mechanical four-dimensional information; S5, inputting the four-dimensional feature vectors into a pre-trained deep learning classification model, and identifying and distinguishing the unstructured wear and the structured slipping deformation. Preferably, the acquiring the oblique photographic image of the multi-stage unmanned aerial vehicle of the side slope includes: planning a layered variable angle flight route according to the slope height and the slope of the slope, wherein the layered variable angle flight route comprises a low-altitude near-view layer route and a high-altitude panoramic layer route, and the inclination angle of a lens of the low-altitude near-view layer route is linearly adjusted along with the slope; controlling an unmanned aerial vehicle carrying the five-lens oblique photography tripod head to