Open-pit mine slope feature analysis method and system based on three-dimensional point cloud and unmanned aerial vehicle

Abstract

The application relates to the technical field of open-pit mine safety monitoring, and discloses an open-pit mine slope feature analysis method and system based on three-dimensional point clouds and unmanned aerial vehicles. The method obtains original three-dimensional point cloud data through an unmanned aerial vehicle and carries out standardized processing, determines a slope surface sheet boundary line and carries out segmentation based on spatial geometric features, extracts geometric feature parameters, determines a deformation area and a displacement vector through registration and differential calculation of multi-time phase point cloud data, and generates a stability evaluation result in combination with a slope mechanical constraint condition. The application can realize accurate monitoring and stability evaluation of slope deformation, and improves the accuracy of open-pit mine slope safety early warning.

Description

Technical Field

[0001] This invention relates to the field of open-pit mine safety monitoring technology, and in particular to a method and system for analyzing the characteristics of open-pit mine slopes based on three-dimensional point clouds and unmanned aerial vehicles (UAVs). Background Technology

[0002] Slope stability analysis is a crucial step in ensuring production safety during open-pit mining. Traditional slope monitoring methods mainly rely on manual on-site measurements or fixed monitoring equipment, which are limited by factors such as measurement frequency, coverage area, and working environment. With the development of UAV technology and 3D point cloud data acquisition technology, using UAVs equipped with ranging sensors to collect 3D point cloud data of open-pit mine slopes provides a new technical means for slope feature analysis.

[0003] Currently, various technical solutions exist for open-pit mine slope feature analysis, mainly including optical image-based analysis methods, ground-based lidar scanning methods, and traditional manual measurement methods. While these methods have achieved some success in slope feature extraction and stability assessment, they still have some technical defects and limitations. Traditional slope feature extraction methods struggle to automatically segment slope patches, typically relying on manual intervention to determine the geometric boundaries of these patches. This not only increases workload but also easily introduces subjective judgment errors, failing to guarantee the consistency and accuracy of the analysis results. Existing technologies lack effective registration and difference calculation methods for multi-temporal point cloud data processing, making it difficult to accurately identify and quantitatively analyze slope deformation areas, especially under complex terrain conditions, where the accuracy and reliability of deformation detection are low. Existing slope stability assessment methods often focus on single geometric parameters or empirical judgments, lacking a comprehensive analytical framework that combines slope geometric features, deformation monitoring results, and slope mechanical constraints, making it difficult to provide comprehensive assessment results including key information such as stability level and potential slip direction. Summary of the Invention

[0004] The present invention provides a method and system for analyzing the features of open-pit mine slopes based on 3D point clouds and UAVs, which can at least solve some of the problems existing in the prior art.

[0005] A first aspect of this invention provides a method for analyzing the features of open-pit mine slopes based on 3D point clouds and unmanned aerial vehicles (UAVs), comprising:

[0006] The raw 3D point cloud data of the open-pit mine slope area is acquired by a ranging sensor carried by a drone, and the raw 3D point cloud data is normalized to obtain normalized 3D point cloud data.

[0007] Based on the spatial geometric features of the standardized three-dimensional point cloud data, the geometric boundary lines between slope patches are determined, and the standardized three-dimensional point cloud data is divided into multiple slope patches according to the geometric boundary lines.

[0008] For each slope patch, geometric feature parameters characterizing the slope morphology are extracted. Based on these geometric feature parameters, the deformation regions in the slope patch are determined through spatial registration and difference calculation of multi-temporal point cloud data, and the displacement vectors of the deformation regions are calculated.

[0009] Based on the geometric characteristic parameters and the displacement vector, combined with the slope mechanical constraints, a slope stability assessment result including stability level and potential slip direction is generated.

[0010] The raw 3D point cloud data of the open-pit mine slope area is acquired using a ranging sensor mounted on a drone. The raw 3D point cloud data is then normalized to obtain normalized 3D point cloud data, including:

[0011] The original three-dimensional point cloud data of the open-pit mine slope area is acquired by a ranging sensor carried by a drone. A local neighborhood is established for each point in the original three-dimensional point cloud data, and noise points are identified by calculating the point cloud density distribution characteristics within the local neighborhood.

[0012] The original 3D point cloud data after removing the noise points is subjected to elevation layering. The point cloud data is divided into multiple elevation intervals according to the elevation direction. Local minimum points are extracted in each elevation interval, and the set of local minimum points of the multiple elevation intervals is used as the bottom boundary points.

[0013] Based on the spatial distribution characteristics of the bottom boundary points, the bottom boundary points are obtained by calculating the distance relationship and elevation change characteristics between adjacent bottom boundary points, eliminating isolated points and abrupt change points.

[0014] The topographic reference surface of the slope area is obtained by performing surface fitting on the selected bottom boundary points. The original three-dimensional point cloud data after removing the noise points is then subjected to coordinate transformation using the topographic reference surface as a reference to obtain the normalized three-dimensional point cloud data.

[0015] Based on the spatial geometric features of the standardized 3D point cloud data, geometric boundaries between slope patches are determined. The standardized 3D point cloud data is then divided into multiple slope patches according to these geometric boundaries, including:

[0016] For each point in the normalized 3D point cloud data, a normal vector is calculated to determine the point cloud normal vector field, which is used to represent the spatial geometric features of the normalized 3D point cloud data;

[0017] Based on the point cloud normal vector field, the angle between the normal vectors of adjacent points is calculated, and point pairs whose angle exceeds the normal vector consistency constraint are identified and marked as candidate boundary points.

[0018] Based on the Euclidean distance and topological adjacency between the candidate boundary points, candidate boundary points whose distances satisfy the continuity constraint and have topological adjacency are connected to obtain a candidate boundary point chain;

[0019] For each candidate boundary chain, a length evaluation and a direction consistency test are performed. Candidate boundary chains whose length meets the validity constraint and whose direction changes meet the smoothness constraint are selected as valid boundary chain chains.

[0020] Curve fitting is performed on the effective boundary point chain to generate geometric boundary lines between slope patches. Based on the geometric boundary lines, the normalized three-dimensional point cloud data is divided into multiple slope patches.

[0021] For each slope patch, geometric feature parameters characterizing the slope morphology are extracted. Based on these geometric feature parameters, the deformation regions in the slope patch are determined through spatial registration and difference calculation of multi-temporal point cloud data. The displacement vectors of these deformation regions are then calculated, including:

[0022] For each slope patch, local surface fitting is performed on the point cloud data within the slope patch, and geometric feature parameters characterizing the slope morphology are calculated based on the fitted surface.

[0023] Acquire multi-temporal point cloud data collected at different times, and extract the point cloud data corresponding to the same slope surface from the multi-temporal point cloud data into a reference temporal point cloud and a target temporal point cloud respectively;

[0024] Using the geometric feature parameters of the reference time-phase point cloud as registration constraints, the spatial correspondence between the reference time-phase point cloud and the target time-phase point cloud is determined, and a spatial registration transformation is performed on the target time-phase point cloud to obtain the registered target time-phase point cloud;

[0025] The point-to-point distance between the registered target temporal point cloud and the reference temporal point cloud is calculated to generate point cloud difference results. Based on the point cloud difference results, the deformation areas in the slope surface that have undergone deformation are determined.

[0026] For each point within the deformation region, the spatial displacement between the corresponding point position in the reference time phase point cloud and the position in the registered target time phase point cloud is calculated to obtain the displacement vector of the deformation region.

[0027] Calculate the point-to-point distance between the registered target temporal point cloud and the reference temporal point cloud to generate point cloud difference results. Based on the point cloud difference results, determine the deformation areas in the slope surface that have undergone deformation, including:

[0028] For each point in the registered target temporal point cloud, search for the nearest corresponding point in the reference temporal point cloud, and calculate the Euclidean distance between the point and the corresponding point as the point-to-point distance;

[0029] Point cloud difference results are obtained based on the point-to-point distances of all points within the slope patch, and these point cloud difference results represent the spatial variation distribution of the slope patch over time.

[0030] Spatial clustering analysis was performed on the point cloud difference results. By calculating the spatial distribution density of point-to-point distances and neighborhood similarity characteristics, multiple clusters were obtained.

[0031] Statistical analysis is performed on each cluster. Based on the statistical feature value of the point-to-point distance within the cluster, clusters whose statistical feature values ​​satisfy the deformation detection constraints are selected as candidate deformation regions.

[0032] Boundary extraction and regional connectivity verification are performed on the candidate deformation regions. Isolated regions whose areas do not meet the validity constraints are eliminated, and the remaining candidate deformation regions are determined as the deformation regions in the slope surface that have undergone deformation.

[0033] Based on the geometric characteristic parameters and the displacement vector, combined with the slope mechanical constraints, a slope stability assessment result is generated, including the stability level and potential slip direction, comprising:

[0034] The critical slip angle of each slope patch is calculated based on the slope angle and normal vector in the geometric feature parameters, and the displacement rate and cumulative displacement of the deformation area are calculated based on the direction and amplitude of the displacement vector.

[0035] By comparing and analyzing the slope angle and the critical slip angle among the geometric characteristic parameters, and combining the displacement rate and cumulative displacement, the evaluation index of the slope's mechanical state is determined.

[0036] The slope mechanical state evaluation index is weighted and integrated with the slope material strength constraint and slope geometric shape constraint in the slope mechanical constraint conditions to obtain the comprehensive slope stability index.

[0037] The stability level of the slope section is determined by comparing the comprehensive slope stability index with a preset stability grading threshold.

[0038] Based on the spatial distribution characteristics of the displacement vector and the normal vector in the geometric feature parameters, the dominant displacement direction of the deformation area is determined as the potential slip direction by calculating the projection component and tangential component of the displacement vector in the direction of the normal vector of the slope patch;

[0039] The stability level is combined with the potential slip direction to obtain the slope stability assessment result.

[0040] By comparing and analyzing the slope angle and the critical slip angle among the geometric characteristic parameters, and combining the displacement rate and cumulative displacement, the evaluation indicators of the slope's mechanical state are determined, including:

[0041] Calculate the angle difference between the slope angle and the critical slip angle, and perform geometric morphology correction on the angle difference based on the slope height and slope length of the slope patch to obtain the geometric safety margin of the slope patch;

[0042] The displacement rate time series composed of displacement rates at multiple times is subjected to difference operation to calculate the rate change between adjacent displacement rates, and the displacement acceleration sequence is determined based on the rate change.

[0043] Trend features are extracted from the displacement acceleration sequence. By analyzing the monotonicity and volatility of the displacement acceleration sequence, the stage features of different stages are identified and quantified into changing trend features.

[0044] The deformation evolution index is obtained by nonlinearly mapping the change trend characteristics to the cumulative displacement, applying a preset time-sensitive weight to the change trend characteristics and coupling it with the cumulative displacement;

[0045] The geometric safety margin and the deformation evolution index are coupled and calculated to obtain the slope mechanical state evaluation index.

[0046] A second aspect of the present invention provides an open-pit mine slope feature analysis system based on 3D point clouds and unmanned aerial vehicles (UAVs), comprising:

[0047] The first unit is used to acquire raw three-dimensional point cloud data of the open-pit mine slope area through a ranging sensor mounted on a UAV, and to perform normalization processing on the raw three-dimensional point cloud data to obtain normalized three-dimensional point cloud data;

[0048] The second unit is used to determine the geometric boundary lines between slope patches based on the spatial geometric features of the standardized three-dimensional point cloud data, and to divide the standardized three-dimensional point cloud data into multiple slope patches according to the geometric boundary lines.

[0049] The third unit is used to extract geometric feature parameters that characterize the slope morphology for each slope patch. Based on the geometric feature parameters, the deformation areas in the slope patch are determined by spatial registration and difference calculation of multi-temporal point cloud data, and the displacement vector of the deformation areas is calculated.

[0050] The fourth unit is used to generate slope stability assessment results, including stability level and potential slip direction, based on the geometric feature parameters and the displacement vector, combined with slope mechanical constraints.

[0051] A third aspect of the present invention provides an electronic device, comprising:

[0052] processor;

[0053] Memory used to store processor-executable instructions;

[0054] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0055] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0056] This invention acquires 3D point cloud data using a drone equipped with a ranging sensor, overcoming the dangers and heavy workload of traditional manual measurement and improving the safety and efficiency of data acquisition. Based on the spatial geometric features of standardized 3D point cloud data, this invention determines the geometric boundaries between slope patches, achieving automatic slope segmentation and avoiding the subjectivity and inconsistency of manual segmentation, thus improving the accuracy and objectivity of segmentation. By extracting the geometric feature parameters of each slope patch and combining spatial registration and differential calculation of multi-temporal point cloud data, this invention can accurately identify deformation areas and calculate displacement vectors, significantly improving the accuracy and spatial resolution of deformation monitoring compared to traditional methods. This invention innovatively combines geometric feature parameters, displacement vectors, and slope mechanical constraints to generate assessment results including stability levels and potential slip directions, achieving fully automated analysis from data acquisition to risk assessment, providing more scientific and comprehensive technical support for open-pit mine slope stability assessment. This invention forms a complete slope feature analysis technology system, significantly improving the intelligence level of open-pit mine slope monitoring and assessment, and has important practical value for preventing slope instability disasters. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating the open-pit mine slope feature analysis method based on 3D point cloud and UAV according to an embodiment of the present invention.

[0058] Figure 2 This is a schematic diagram illustrating the process of determining the deformation region in a slope surface area according to an embodiment of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0061] Figure 1 This is a flowchart illustrating the open-pit mine slope feature analysis method based on 3D point cloud and UAV according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0062] The raw 3D point cloud data of the open-pit mine slope area is acquired by a ranging sensor carried by a drone, and the raw 3D point cloud data is normalized to obtain normalized 3D point cloud data.

[0063] Based on the spatial geometric features of the standardized three-dimensional point cloud data, the geometric boundary lines between slope patches are determined, and the standardized three-dimensional point cloud data is divided into multiple slope patches according to the geometric boundary lines.

[0064] For each slope patch, geometric feature parameters characterizing the slope morphology are extracted. Based on these geometric feature parameters, the deformation regions in the slope patch are determined through spatial registration and difference calculation of multi-temporal point cloud data, and the displacement vectors of the deformation regions are calculated.

[0065] Based on the geometric characteristic parameters and the displacement vector, combined with the slope mechanical constraints, a slope stability assessment result including stability level and potential slip direction is generated.

[0066] In one optional implementation, raw 3D point cloud data of the open-pit mine slope area is acquired using a ranging sensor mounted on a UAV. The raw 3D point cloud data is then normalized to obtain normalized 3D point cloud data, including:

[0067] The original three-dimensional point cloud data of the open-pit mine slope area is acquired by a ranging sensor carried by a drone. A local neighborhood is established for each point in the original three-dimensional point cloud data, and noise points are identified by calculating the point cloud density distribution characteristics within the local neighborhood.

[0068] The original 3D point cloud data after removing the noise points is subjected to elevation layering. The point cloud data is divided into multiple elevation intervals according to the elevation direction. Local minimum points are extracted in each elevation interval, and the set of local minimum points of the multiple elevation intervals is used as the bottom boundary points.

[0069] Based on the spatial distribution characteristics of the bottom boundary points, the bottom boundary points are obtained by calculating the distance relationship and elevation change characteristics between adjacent bottom boundary points, eliminating isolated points and abrupt change points.

[0070] The topographic reference surface of the slope area is obtained by performing surface fitting on the selected bottom boundary points. The original three-dimensional point cloud data after removing the noise points is then subjected to coordinate transformation using the topographic reference surface as a reference to obtain the normalized three-dimensional point cloud data.

[0071] Raw 3D point cloud data of the open-pit mine slope area is acquired using a ranging sensor mounted on a drone. This ranging sensor can be a lidar or structured light sensor. During data acquisition, the drone flies above the slope area along a preset flight path. The lidar measures the position of ground points by emitting laser beams and receiving their reflected signals, forming a 3D point cloud dataset containing spatial coordinate information. During data acquisition, the drone's flight altitude is typically maintained between 40 and 60 meters above the ground to ensure the accuracy and coverage of the data acquisition.

[0072] For each point in the original 3D point cloud data, a local neighborhood is established. This local neighborhood can be determined using either the K-nearest neighbor (KNN) or spherical neighborhood method. Taking the KNN method as an example, the 30 points with the closest spatial distance to each point are selected as its local neighborhood. When using the spherical neighborhood method, all points within a sphere with a radius of 0.5 meters, centered on the current point, constitute the local neighborhood. After determining the local neighborhood, the point cloud density distribution characteristics within that neighborhood are calculated, including the spatial uniformity of point distribution and the local point density. Density calculation uses a kernel density estimation method. If the local density of a point is significantly lower than 20% of the average density within its neighborhood, or if its average distance to its neighboring points is significantly greater than twice the overall average, then that point is identified as a noise point and removed.

[0073] The original 3D point cloud data, after noise point removal, undergoes elevation layering. First, the elevation range of the entire dataset along the Z-axis is determined, and then this range is equally divided into multiple elevation intervals. The interval between elevation intervals is typically set to 1 to 2 meters to ensure that effective terrain features can be extracted within each interval. For each elevation interval, by comparing the Z-coordinates of all points within the interval, the set of points with the smallest Z-value is extracted as the local minimum points of that interval. These points typically represent low-lying areas or the bottom contour of slopes within that elevation interval. The local minimum points of all elevation intervals are merged to form a set of bottom boundary points, which initially delineate the bottom contour of the slope.

[0074] Based on the spatial distribution characteristics of the bottom boundary points, the Euclidean distance and elevation difference between adjacent bottom boundary points are calculated. For each pair of adjacent points, if the distance exceeds a preset threshold (usually 3 to 5 meters) or the elevation change rate exceeds twice the normal slope of the slope, it is identified as an anomaly. Further analysis of the spatial connectivity of these anomaly points reveals that if the average distance between a point and its five nearest bottom boundary points is significantly greater than the overall average, it is identified as an isolated point; if the elevation gradient formed by a point and its adjacent points is significantly greater than the average gradient within the local area, it is identified as a mutation point. After removing these isolated and mutation points, a filtered set of bottom boundary points is obtained, which can more accurately represent the topographic contour of the slope bottom.

[0075] Surface fitting is performed on the selected bottom boundary points. The RANSAC (Random Sample Consensus) algorithm combined with a quadratic polynomial model can be used to fit the bottom boundary points, resulting in a mathematical model representing the topographic reference surface of the slope area. This reference surface can be represented as Z = f(X,Y), where f is the fitted quadratic polynomial function. During the fitting process, the number of iterations is set to 200, and the interior point threshold is 0.1 meters to ensure a good fit of the model to the bottom boundary points.

[0076] Using the topographic reference plane as a reference, a coordinate transformation is performed on the original 3D point cloud data after noise point removal. This transformation includes two steps: translation and rotation. First, the point cloud data is translated to a coordinate system with the center point of the topographic reference plane as the origin. Then, a rotation matrix is ​​calculated based on the normal vectors of the topographic reference plane and the horizontal plane, rotating the point cloud to a position where the topographic reference plane is parallel to the horizontal plane. Through this coordinate transformation, the slope point cloud data is standardized into a unified reference coordinate system, facilitating subsequent analysis and processing.

[0077] In practical application, this method was used to process point cloud data acquired by UAVs in a slope monitoring project at an open-pit coal mine. The original data contained approximately 3 million points, and after noise point removal, approximately 2.8 million valid points were retained. Elevation stratification was performed at 1.5-meter intervals, dividing the data into 12 elevation zones, and more than 2,000 bottom boundary points were extracted. After screening for isolated and abrupt change points, approximately 1,800 bottom boundary points were ultimately retained for datum surface fitting. The average fitting error of the fitted topographic datum surface to the bottom boundary points was less than 0.08 meters, indicating good fitting accuracy. In the normalized point cloud data after coordinate transformation, the bottom of the slope was basically parallel to the horizontal plane, providing an effective data foundation for subsequent slope stability analysis and deformation monitoring.

[0078] Through the above processing procedures, the point cloud data of open-pit mine slope areas has been standardized, providing a unified spatial reference framework and a high-quality data foundation for slope monitoring and analysis.

[0079] In one optional implementation, based on the spatial geometric features of the normalized 3D point cloud data, geometric boundaries between slope patches are determined, and the normalized 3D point cloud data is divided into multiple slope patches according to the geometric boundaries, including:

[0080] For each point in the normalized 3D point cloud data, a normal vector is calculated to determine the point cloud normal vector field, which is used to represent the spatial geometric features of the normalized 3D point cloud data;

[0081] Based on the point cloud normal vector field, the angle between the normal vectors of adjacent points is calculated, and point pairs whose angle exceeds the normal vector consistency constraint are identified and marked as candidate boundary points.

[0082] Based on the Euclidean distance and topological adjacency between the candidate boundary points, candidate boundary points whose distances satisfy the continuity constraint and have topological adjacency are connected to obtain a candidate boundary point chain;

[0083] For each candidate boundary chain, a length evaluation and a direction consistency test are performed. Candidate boundary chains whose length meets the validity constraint and whose direction changes meet the smoothness constraint are selected as valid boundary chain chains.

[0084] Curve fitting is performed on the effective boundary point chain to generate geometric boundary lines between slope patches. Based on the geometric boundary lines, the normalized three-dimensional point cloud data is divided into multiple slope patches.

[0085] In the field of slope safety monitoring, accurate segmentation of three-dimensional point cloud data is a key step in achieving slope stability assessment. Based on the spatial geometric characteristics of standardized three-dimensional point cloud data, the geometric boundaries between slope patches can be determined, and the standardized three-dimensional point cloud data can be segmented into multiple slope patches accordingly.

[0086] First, the normal vector is calculated for each point in the normalized 3D point cloud data to determine the point cloud normal vector field. For each point p in the point cloud, its K nearest neighbors N(p) are selected. Typically, the value of K is set between 30 and 50 to ensure sufficient local information for normal vector estimation. Based on these neighboring points, the local covariance matrix C is calculated:

[0087] Eigenvalue decomposition is performed on the covariance matrix C to obtain three eigenvalues ​​λ1≥λ2≥λ3 and their corresponding eigenvectors v1, v2, and v3. The eigenvector v3 corresponding to the smallest eigenvalue λ3 is the normal vector at point p. By repeating this process for all points in the point cloud, a complete point cloud normal vector field is formed, which can accurately represent the spatial geometric features of the normalized 3D point cloud data.

[0088] Next, based on the point cloud normal vector field, the angle between the normal vectors of adjacent points is calculated, and point pairs whose normal vector angle exceeds the normal vector consistency constraint are identified. For each point p, each point q in its neighborhood point set N(p) is checked, and the angle between the normal vectors of the two points is calculated:

[0089] If the angle θ between the normal vectors exceeds a preset threshold θth (usually set between 15° and 20°), then a potential geometric feature change exists between points p and q, and the point pair (p,q) is marked as a candidate boundary point. The output of this step is a set of candidate boundary points located at the intersection of different slope patches.

[0090] Then, based on the Euclidean distance and topological adjacency between candidate boundary points, candidate boundary points that satisfy the continuity constraint and have a topological adjacency relationship are connected to form a candidate boundary point chain. For any two candidate boundary points pi and pj, if their Euclidean distance d(pi,pj) is less than a preset threshold dth (usually set to 1.5 to 2 times the average point spacing of the point cloud), and they have a connecting path in the topological structure of the point cloud, then these two points are considered to be connected as part of the candidate boundary point chain. A region growing algorithm is used to gradually add adjacent candidate boundary points that meet the conditions to the current boundary point chain, starting from any candidate boundary point, until no new points that meet the conditions can be found. This process is repeated until all candidate boundary points are processed, forming a set of candidate boundary point chains.

[0091] For each candidate boundary point chain, a length evaluation and direction consistency test are performed. Candidate boundary point chains whose length meets the validity constraint and whose direction change meets the smoothness constraint are selected as valid boundary point chains. For each candidate boundary point chain C, the number of points len(C) it contains is first calculated. If len(C) is less than the preset minimum length threshold lenmin (usually set to 0.5% to 1% of the total number of points in the point cloud), the chain is considered too short, is noise or a false detection, and is removed.

[0092] For candidate boundary point chains that meet the length requirement, the smoothness of their direction changes is further examined. Local tangent vectors are calculated along the boundary point chain, and the angle change between adjacent tangent vectors is evaluated. If the maximum angle change exceeds a preset smoothness threshold αmax (usually set to 30° to 45°), the boundary point chain is considered to have excessive direction change and does not meet the smoothness constraint, and is therefore discarded. Through this screening step, a set of effective boundary point chains with suitable length and smooth shape is obtained.

[0093] Finally, curve fitting is performed on the effective boundary point chains to generate geometric boundaries between slope patches. Based on these geometric boundaries, the normalized 3D point cloud data is divided into multiple slope patches. For each effective boundary point chain, a B-spline curve fitting method is used for smoothing. B-spline curves have the characteristics of good local controllability and numerical stability, and can effectively eliminate noise and irregularities in the point chains. The smoothed curve obtained by fitting is the geometric boundary between the slope patches.

[0094] Based on the generated set of geometric boundaries, the entire normalized 3D point cloud data can be divided into multiple independent slope patch regions. A region growing algorithm is then used, starting with non-boundary points as seed points, to expand outwards until a boundary line or a marked point is encountered. In this way, each point in the point cloud is assigned to a specific slope patch, completing the slope patch segmentation based on geometric features.

[0095] In practical applications, such as a mountain highway slope monitoring project, the above method was used to process the collected point cloud data, successfully dividing the complex slope into more than 20 independent patches, providing a reliable foundation for subsequent displacement monitoring and stability analysis, and effectively improving the accuracy and efficiency of slope monitoring.

[0096] In one optional implementation, geometric feature parameters characterizing the slope morphology are extracted for each slope patch. Based on these geometric feature parameters, the deformation regions in the slope patch that have undergone deformation are determined through spatial registration and difference calculation of multi-temporal point cloud data, and the displacement vectors of the deformation regions are calculated, including:

[0097] For each slope patch, local surface fitting is performed on the point cloud data within the slope patch, and geometric feature parameters characterizing the slope morphology are calculated based on the fitted surface.

[0098] Acquire multi-temporal point cloud data collected at different times, and extract the point cloud data corresponding to the same slope surface from the multi-temporal point cloud data into a reference temporal point cloud and a target temporal point cloud respectively;

[0099] Using the geometric feature parameters of the reference time-phase point cloud as registration constraints, the spatial correspondence between the reference time-phase point cloud and the target time-phase point cloud is determined, and a spatial registration transformation is performed on the target time-phase point cloud to obtain the registered target time-phase point cloud;

[0100] The point-to-point distance between the registered target temporal point cloud and the reference temporal point cloud is calculated to generate point cloud difference results. Based on the point cloud difference results, the deformation areas in the slope surface that have undergone deformation are determined.

[0101] For each point within the deformation region, the spatial displacement between the corresponding point position in the reference time phase point cloud and the position in the registered target time phase point cloud is calculated to obtain the displacement vector of the deformation region.

[0102] When extracting geometric feature parameters representing the slope morphology for each slope patch, different slope patches are first segmented from the acquired point cloud data. These patches can be segmented using a region growing method, grouping adjacent points with similar normal vectors into the same patch. During segmentation, a point is randomly selected as a seed point. Then, its neighboring points are checked. If the angle between the normal vector of a neighboring point and the normal vector of the seed point is less than a preset threshold (e.g., 10 degrees), that neighboring point is added to the current patch and used as a new seed point to continue expanding until no neighboring points satisfying the condition can be found.

[0103] For each segmented slope patch, local surface fitting is performed on the point cloud data within the patch. The fitting process uses the least squares method, with a quadratic surface as the basic model. The fitting equation can be expressed as ax² + by² + cxy + dx + ey + f = z. By solving for the coefficients a, b, c, d, e, and f, the best-fitting surface is obtained. Based on the fitted surface, geometric characteristic parameters representing the slope morphology are calculated. These parameters include, but are not limited to, the principal curvature values ​​and directions of the surface, Gaussian curvature, mean curvature, slope, aspect, and roughness of the patch. These characteristic parameters can effectively characterize the geometric morphological features of the slope.

[0104] After acquiring multi-temporal point cloud data collected at different times, the point cloud data corresponding to the same slope patch are extracted into a baseline time phase point cloud and a target time phase point cloud. During the extraction process, the aforementioned slope patch segmentation method can be used to ensure that the extracted patches correspond to the slope area at the same geographical location. The baseline time phase is usually selected from the point cloud data at the initial monitoring time, while the target time phase is the point cloud data at subsequent monitoring times.

[0105] Using the geometric feature parameters of the reference time-phase point cloud as registration constraints, spatial registration transformation of the target time-phase point cloud is performed by determining the spatial correspondence between the reference and target time-phase point clouds. The registration process employs the Iterative Closest Point (ICP) algorithm, but unlike traditional ICP, it introduces constraints based on geometric feature parameters. Specifically, firstly, points with distinct features are selected as control points in the reference time-phase point cloud. Then, candidate matching points with similar geometric feature parameters to these control points are searched in the target time-phase point cloud to establish an initial correspondence. Subsequently, based on these correspondences, a rigid body transformation matrix (including the rotation matrix R and translation vector t) is calculated to transform the target time-phase point cloud into the coordinate system of the reference time-phase point cloud. By iteratively optimizing the transformation parameters, the distance error between the transformed target time-phase point cloud and the reference time-phase point cloud is minimized, thus obtaining the registered target time-phase point cloud.

[0106] Point cloud difference results are generated by calculating the point-to-point distance between the registered target temporal point cloud and the reference temporal point cloud. For each point in the reference temporal point cloud, the nearest point in the registered target temporal point cloud is found, and the Euclidean distance between them is calculated to obtain the point-to-point distance value. These distance values ​​constitute the point cloud difference results, which can intuitively reflect the deformation of each area on the slope surface. When determining the deformation areas in the slope surface based on the point cloud difference results, a threshold (e.g., 3 cm) is set. When the point-to-point distance is greater than this threshold, the area where the point is located is considered to have undergone significant deformation. Through cluster analysis, adjacent points that have both undergone significant deformation are divided into the same deformation area, thereby identifying the deformation areas in the slope surface.

[0107] For each point within the defined deformation area, the spatial displacement between its corresponding position in the reference time-phase point cloud and its position in the registered target time-phase point cloud is calculated, yielding the displacement vector of the deformation area. The displacement vector calculation is based on the previously established point correspondence. For each point p_base (a point in the reference time-phase point cloud) within the deformation area, its corresponding point p_target in the registered target time-phase point cloud is found. Then, the displacement vector v = p_target - p_base. These displacement vectors contain not only the magnitude of the displacement but also its direction, comprehensively describing the spatial characteristics of the slope deformation.

[0108] In practical applications, the calculated displacement vectors can be overlaid onto a 3D model using color coding for visualization. For example, different colors can represent displacements of different magnitudes and directions, with red indicating outward movement (expansion) and blue indicating inward movement (contraction). Furthermore, based on the spatial distribution characteristics of the displacement vectors, the patterns and trends of slope deformation can be further analyzed, providing a scientific basis for slope stability assessment and early warning.

[0109] In one optional implementation, the point-to-point distance between the registered target temporal point cloud and the reference temporal point cloud is calculated to generate a point cloud difference result. Based on the point cloud difference result, the deformation area in the slope patch that has undergone deformation is determined, including:

[0110] For each point in the registered target temporal point cloud, search for the nearest corresponding point in the reference temporal point cloud, and calculate the Euclidean distance between the point and the corresponding point as the point-to-point distance;

[0111] Point cloud difference results are obtained based on the point-to-point distances of all points within the slope patch, and these point cloud difference results represent the spatial variation distribution of the slope patch over time.

[0112] Spatial clustering analysis was performed on the point cloud difference results. By calculating the spatial distribution density of point-to-point distances and neighborhood similarity characteristics, multiple clusters were obtained.

[0113] Statistical analysis is performed on each cluster. Based on the statistical feature value of the point-to-point distance within the cluster, clusters whose statistical feature values ​​satisfy the deformation detection constraints are selected as candidate deformation regions.

[0114] Boundary extraction and regional connectivity verification are performed on the candidate deformation regions. Isolated regions whose areas do not meet the validity constraints are eliminated, and the remaining candidate deformation regions are determined as the deformation regions in the slope surface that have undergone deformation.

[0115] In the implementation of point cloud-based slope monitoring technology, the deformation area of ​​the slope can be effectively detected by calculating the point-to-point distance between the registered target time-phase point cloud and the reference time-phase point cloud.

[0116] Figure 2 This is a schematic flowchart illustrating the process of determining the deformation region in a slope surface area according to an embodiment of the present invention. Figure 2 As shown, firstly, each point in the registered target temporal point cloud is processed. Specifically, for each point Pi in the target temporal point cloud, the nearest corresponding point Qi in the reference temporal point cloud is searched, and the kd-tree algorithm is used to accelerate the nearest neighbor search process. For each point Pi, the point-to-point distance is obtained by calculating the Euclidean distance d(Pi, Qi) = sqrt[(xi-xqi)² + (yi-yqi)² +(zi-zqi)²], where xi, yi, and zi represent the three-dimensional coordinates of point Pi, and xqi, yqi, and zqi represent the three-dimensional coordinates of point Qi. In this way, a distance value can be obtained for each point in the target temporal point cloud, forming a point cloud difference dataset D.

[0117] Based on the point-to-point distances of all points within the slope patch calculated above, point cloud difference results are constructed. These difference distances are then mapped onto the point cloud according to a color gradient, such as a blue-to-red gradient spectrum, where blue represents smaller variations (e.g., 0-5 mm) and red represents larger variations (e.g., greater than 30 mm). This visualization intuitively shows the spatial distribution of slope patch variations over time, providing a foundation for subsequent analysis.

[0118] Next, spatial clustering analysis is performed on the point cloud difference results using the density-based clustering algorithm DBSCAN. This algorithm is based on two key parameters: ε (defining the neighborhood radius) and MinPts (defining the minimum number of neighbors required to become a core point). In practical applications, ε can be set to 2-3 times the average spacing of the point cloud, and MinPts can range from 5 to 10, depending on the density of the slope point cloud. During the clustering process, the density of each point is calculated, i.e., the number of points in the neighborhood of ε radius, and the similarity of the difference values ​​of the neighboring points is analyzed, thereby dividing the data into multiple clusters C = {C1, C2, ..., Cn}.

[0119] Statistical analysis was performed on each cluster to extract statistical feature values ​​for deformation region identification. For each cluster Ci, the following statistical feature values ​​were calculated: mean μi, standard deviation σi, maximum value maxi, minimum value mini, median median, and number of points counti. Based on these statistical feature values, deformation detection constraints were set: (1) μi is greater than threshold T1 (e.g., 10 mm); (2) counti is greater than threshold T2 (e.g., 50 points); (3) the difference between maxi and mini is less than threshold T3 (e.g., 30 mm), ensuring that the differences within the deformation region are moderate. Clusters that meet the above conditions are marked as candidate deformation regions R = {R1, R2, ..., Rm}, where m≤n.

[0120] Boundary extraction and connectivity verification are performed on candidate deformation regions. First, the boundary of each candidate deformation region is extracted using the convex hull algorithm or the α-shape algorithm, resulting in a boundary point set Bi = {b1, b2, ..., bk}. Then, the area Ai of each candidate region is calculated, which can be achieved by projecting the region onto the principal plane and then using a polygon area calculation method. An area validity constraint threshold T4 (e.g., 0.5 square meters) is set, and isolated regions with an area smaller than T4 are eliminated. Simultaneously, the connectivity between adjacent candidate regions is checked. If the minimum distance between two regions is less than the threshold T5 (e.g., 5 times the average spacing of slope point clouds) and the difference values ​​are similar (difference less than the threshold T6, e.g., 5 mm), then these two regions are merged into a single complete deformation region.

[0121] In practical slope monitoring applications, three main deformation areas were detected on a mountain slope after two scans 30 days apart, using the method described above. The maximum deformation reached 28 mm, and the total area of ​​the deformation areas was approximately 15 square meters, mainly distributed in the upper and middle parts of the slope. This method can not only accurately locate the deformation areas but also quantify the degree of deformation, providing a scientific basis for subsequent slope stability assessments and disaster prevention and mitigation measures.

[0122] After the above processing steps, the finally determined deformation area represents the area in the slope surface where significant deformation has occurred. These areas need to be closely monitored and continuously monitored to prevent the risk of slope instability.

[0123] In one optional implementation, based on the geometric characteristic parameters and the displacement vector, combined with the slope mechanical constraints, a slope stability assessment result including stability level and potential slip direction is generated, including:

[0124] The critical slip angle of each slope patch is calculated based on the slope angle and normal vector in the geometric feature parameters, and the displacement rate and cumulative displacement of the deformation area are calculated based on the direction and amplitude of the displacement vector.

[0125] By comparing and analyzing the slope angle and the critical slip angle among the geometric characteristic parameters, and combining the displacement rate and cumulative displacement, the evaluation index of the slope's mechanical state is determined.

[0126] The slope mechanical state evaluation index is weighted and integrated with the slope material strength constraint and slope geometric shape constraint in the slope mechanical constraint conditions to obtain the comprehensive slope stability index.

[0127] The stability level of the slope section is determined by comparing the comprehensive slope stability index with a preset stability grading threshold.

[0128] Based on the spatial distribution characteristics of the displacement vector and the normal vector in the geometric feature parameters, the dominant displacement direction of the deformation area is determined as the potential slip direction by calculating the projection component and tangential component of the displacement vector in the direction of the normal vector of the slope patch;

[0129] The stability level is combined with the potential slip direction to obtain the slope stability assessment result.

[0130] Based on the geometric characteristic parameters and displacement vectors, combined with the slope mechanical constraints, slope stability assessment results including stability level and potential slip direction can be generated.

[0131] First, obtain the geometric characteristic parameters of the slope surface, including spatial geometric quantities such as slope angle, normal vector, area, and shape features, as well as the displacement vector data of the slope surface, including displacement direction and magnitude. This data can be obtained through techniques such as radar interferometry, 3D laser scanning, or UAV photogrammetry.

[0132] The critical slip angle for each slope segment is calculated based on the slope angle and normal vector obtained from the geometric feature parameters. The critical slip angle θc can be calculated by combining the slope angle α and the internal friction angle φ of the slope material: θc = arctan(tanφ / cosα). For each slope segment, when the actual slip angle exceeds the critical slip angle, there is a risk of instability and slippage in that area.

[0133] Simultaneously, the displacement rate and cumulative displacement of the deformation area are calculated based on the direction and amplitude of the displacement vector. The displacement rate can be obtained by dividing the displacement difference between two adjacent monitoring points by the time interval. The cumulative displacement is obtained by integrating displacement data from multiple time periods. For example, for slope patch i, if the time interval between two monitoring points is Δt, and the displacement vectors are d1 and d2 respectively, then the displacement rate v = |d2 - d1| / Δt, and the cumulative displacement is the vector sum of all displacement amplitudes during the monitoring period.

[0134] By comparing and analyzing the slope angle and critical slip angle among the geometric characteristic parameters, and combining the displacement rate and cumulative displacement, an evaluation index for the mechanical state of the slope is determined. Specifically, a slip risk index SRI can be constructed as: SRI = (α / θc) × (v / v0) × (D / D0), where v0 is the warning displacement rate threshold, D is the cumulative displacement, and D0 is the warning cumulative displacement threshold. When the SRI value is greater than 1, it indicates that the area is in a potentially unstable state.

[0135] Next, the slope mechanical state evaluation index is weighted and fused with the slope material strength constraint and slope geometric shape constraint in the slope mechanical constraints to obtain the slope comprehensive stability index SSI. The weighted fusion can be expressed as: SSI = w1×SRI + w2×MSI + w3×GSI, where MSI is the material strength index, reflecting the strength characteristics of the soil and rock mass; GSI is the geometric shape index, reflecting the influence of slope geometry on stability; w1, w2, and w3 are the corresponding weight coefficients, and w1+w2+w3=1.

[0136] The stability level of a slope section is determined by comparing the slope comprehensive stability index (SSI) with a preset stability grading threshold. Stability can be divided into four levels: stable when SSI < 0.3; basically stable when 0.3 ≤ SSI < 0.6; understability when 0.6 ≤ SSI < 0.9; and unstable when SSI ≥ 0.9.

[0137] Based on the spatial distribution characteristics of the displacement vector and the normal vector in the geometric characteristic parameters, the dominant displacement direction of the deformation area is determined as the potential sliding direction by calculating the projection component and tangential component of the displacement vector in the direction of the normal vector of the slope patch. For slope patch i, if its normal vector is n and its displacement vector is d, then the normal component dn = (d×n)×n, and the tangential component dt = d - dn. The dominant displacement direction can be obtained by clustering the tangential components of all patches in the region, and the direction of the cluster center is the potential sliding direction.

[0138] Finally, the stability level is combined with the potential slip direction to obtain the slope stability assessment results. The assessment results may include: stability level distribution map of each area of ​​the slope, potential slip direction vector map, high-risk area identification and early warning information.

[0139] In practical applications, taking a slope in an open-pit mine as an example, the geometric parameters and displacement data of the slope surface were obtained through radar monitoring. In the northern area, the slope angle is 42°, the critical slip angle is 38°, the displacement rate is 2.3 mm / day, and the cumulative displacement reaches 78 mm. The calculated SRI value is 1.35, indicating that the area is in a potentially unstable state. Considering that the lithology of this area is weathered granite with a material strength index (MSI) of 0.58 and a geometric morphology index (GSI) of 0.72, and using weighting coefficients w1=0.5, w2=0.3, and w3=0.2, the calculated comprehensive stability index (SSI) is 0.95, classifying it as unstable.

[0140] Analysis of the displacement vector revealed that the normal component of the displacement vector in this area is relatively small, indicating primarily tangential movement. The potential slip direction points towards the toe of the slope, forming a 65° angle with the slope's strike. The comprehensive assessment indicates a high-risk landslide hazard in the northern area, requiring immediate reinforcement measures and increased monitoring frequency.

[0141] In one optional implementation, the slope angle and the critical slip angle in the geometric characteristic parameters are compared and analyzed, and the slope mechanical state evaluation index is determined by combining the displacement rate and the cumulative displacement, including:

[0142] Calculate the angle difference between the slope angle and the critical slip angle, and perform geometric morphology correction on the angle difference based on the slope height and slope length of the slope patch to obtain the geometric safety margin of the slope patch;

[0143] The displacement rate time series composed of displacement rates at multiple times is subjected to difference operation to calculate the rate change between adjacent displacement rates, and the displacement acceleration sequence is determined based on the rate change.

[0144] Trend features are extracted from the displacement acceleration sequence. By analyzing the monotonicity and volatility of the displacement acceleration sequence, the stage features of different stages are identified and quantified into changing trend features.

[0145] The deformation evolution index is obtained by nonlinearly mapping the change trend characteristics to the cumulative displacement, applying a preset time-sensitive weight to the change trend characteristics and coupling it with the cumulative displacement;

[0146] The geometric safety margin and the deformation evolution index are coupled and calculated to obtain the slope mechanical state evaluation index.

[0147] The slope mechanical state evaluation process includes two main stages: data acquisition and calculation of slope mechanical state evaluation indicators. In the data acquisition stage, three-dimensional deformation data of the slope is obtained using synthetic aperture radar interferometry, and geometric characteristic parameters of the slope are extracted using a digital elevation model. In the slope mechanical state evaluation indicator calculation stage, the slope angle and critical slip angle are compared and analyzed, and the displacement rate and cumulative displacement are combined to determine the slope mechanical state evaluation indicators.

[0148] First, the geometric parameters of a slope include the slope angle, slope height, and slope length. The slope angle reflects the angle between the slope surface and the horizontal plane and is calculated using a digital elevation model. The critical slip angle is a function of the slope material properties and is determined through geotechnical mechanics tests or empirical formulas. It represents the critical angle value at which the slope has a risk of instability under specific material conditions.

[0149] The angle difference between the slope angle and the critical slip angle is calculated. This difference reflects the safety margin of the slope geometry. Taking a real slope as an example, if the slope angle of a certain slope section is 35 degrees and its critical slip angle is 42 degrees, then the angle difference is 7 degrees, which initially indicates that the slope section is in a relatively stable state. However, the angle difference needs to be corrected for geometric correlation with the slope height and slope length of the slope section, because slopes with the same angle difference but different slope heights and slope lengths will have significantly different actual stability.

[0150] The specific correction process uses the slope height-slope length correction coefficient method. The correction coefficient K is calculated as the ratio of slope height to slope length multiplied by an adjustment factor. When the slope height is large and the slope length is small, the K value is large, indicating that the slope geometry is not conducive to stability; conversely, the K value is small, and the slope geometry is conducive to stability. Dividing the angle difference by the correction coefficient K yields the geometrical safety margin (GSM). For example, if the slope height of the above slope section is 60 meters and the slope length is 120 meters, and the adjustment factor is 1.5, then K = 60 / 120 × 1.5 = 0.75, and the geometrical safety margin (GSM) = 7 / 0.75 = 9.33.

[0151] Next, based on the displacement data acquired from multiple moments by radar monitoring, a displacement rate time series is formed. The series is then subjected to differential calculation to determine the rate change between adjacent moments, i.e., the displacement acceleration. For example, if the displacement rates at moments t1, t2, and t3 are 2 mm / day, 3 mm / day, and 5 mm / day, respectively, then the acceleration from t1 to t2 is 1 mm / day², and the acceleration from t2 to t3 is 2 mm / day², thus forming the displacement acceleration sequence [1,2] mm / day².

[0152] Trend features are extracted from the displacement acceleration sequence to analyze its monotonicity and volatility. Monotonicity reflects whether the acceleration sequence continues to increase or decrease, while volatility reflects the degree of oscillation in the sequence. The trend characteristics are quantified by calculating the rate of change, volatility coefficient, and trend index of the acceleration sequence. The rate of change is calculated as the ratio of the last value to the first value in the sequence; the volatility coefficient is the ratio of the standard deviation to the mean; and the trend index is obtained by linearly fitting the sequence to obtain its slope.

[0153] The deformation evolution index (DEI) is obtained by nonlinearly mapping the trend characteristics to the cumulative displacement. Specifically, a preset time-sensitive weight is applied to the trend characteristics and coupled with the cumulative displacement. The time-sensitive weight reflects the fact that recent deformation has a greater impact on the slope condition and is typically designed using an exponential decay function. The cumulative displacement needs to be normalized to match the dimensions of the trend characteristics. Assuming the trend characteristic value is 2.5, the normalized cumulative displacement is 0.8, and the time-sensitive weight is 1.2, then the deformation evolution index DEI = 2.5 × 0.8 × 1.2 = 2.4.

[0154] Finally, the Geometric Safety Margin (GSM) and Deformation Evolution Index (DEI) are coupled to calculate the Slope Mechanical State Evaluation Index (SMI). The coupling calculation uses a weighted average method, with weights determined based on the slope type and monitoring target. For example, for deformation-sensitive slopes, the weight of DEI can be set higher; for geometrically controlled slopes, the weight of GSM can be set higher. If GSM = 9.33, DEI = 2.4, and the weights are 0.4 and 0.6 respectively, then SMI = 9.33 × 0.4 + 2.4 × 0.6 = 5.17.

[0155] The Slope Mechanical State Index (SMI) can be divided into multiple levels. For example, 0-3 is defined as a high-risk area, 3-6 as a medium-risk area, and 6 and above as a low-risk area. Based on this classification standard, the SMI value of the slope in the above example is 5.17, which belongs to the medium-risk area. Strengthened monitoring and appropriate protective measures are required.

[0156] This invention combines static geometric features with dynamic deformation features to comprehensively reflect the mechanical state of slopes, which helps guide slope monitoring and disaster prevention and mitigation work, and has strong practicality and promotion value.

[0157] The open-pit mine slope feature analysis system based on 3D point cloud and UAV in this embodiment of the invention includes:

[0158] The first unit is used to acquire raw three-dimensional point cloud data of the open-pit mine slope area through a ranging sensor mounted on a UAV, and to perform normalization processing on the raw three-dimensional point cloud data to obtain normalized three-dimensional point cloud data;

[0159] The second unit is used to determine the geometric boundary lines between slope patches based on the spatial geometric features of the standardized three-dimensional point cloud data, and to divide the standardized three-dimensional point cloud data into multiple slope patches according to the geometric boundary lines.

[0160] The third unit is used to extract geometric feature parameters that characterize the slope morphology for each slope patch. Based on the geometric feature parameters, the deformation areas in the slope patch are determined by spatial registration and difference calculation of multi-temporal point cloud data, and the displacement vector of the deformation areas is calculated.

[0161] The fourth unit is used to generate slope stability assessment results, including stability level and potential slip direction, based on the geometric feature parameters and the displacement vector, combined with slope mechanical constraints.

[0162] A third aspect of the present invention provides an electronic device, comprising:

[0163] processor;

[0164] Memory used to store processor-executable instructions;

[0165] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0166] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0167] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing the characteristics of open-pit mine slopes based on 3D point clouds and UAVs, characterized in that, include: The raw 3D point cloud data of the open-pit mine slope area is acquired by a ranging sensor carried by a drone, and the raw 3D point cloud data is normalized to obtain normalized 3D point cloud data. Based on the spatial geometric features of the standardized three-dimensional point cloud data, the geometric boundary lines between slope patches are determined, and the standardized three-dimensional point cloud data is divided into multiple slope patches according to the geometric boundary lines. For each slope patch, geometric feature parameters characterizing the slope morphology are extracted. Based on these geometric feature parameters, the deformation regions in the slope patch are determined through spatial registration and difference calculation of multi-temporal point cloud data, and the displacement vectors of the deformation regions are calculated. Based on the geometric feature parameters and the displacement vector, combined with the slope mechanical constraints, a slope stability assessment result including stability level and potential slip direction is generated. Based on the spatial geometric features of the standardized 3D point cloud data, geometric boundaries between slope patches are determined, and the standardized 3D point cloud data is divided into multiple slope patches according to the geometric boundaries, including: For each point in the normalized 3D point cloud data, a normal vector is calculated to determine the point cloud normal vector field, which is used to represent the spatial geometric features of the normalized 3D point cloud data; Based on the point cloud normal vector field, the angle between the normal vectors of adjacent points is calculated, and point pairs whose angle exceeds the normal vector consistency constraint are identified and marked as candidate boundary points. Based on the Euclidean distance and topological adjacency between the candidate boundary points, candidate boundary points whose distances satisfy the continuity constraint and have topological adjacency are connected to obtain a candidate boundary point chain; For each candidate boundary chain, a length evaluation and a direction consistency test are performed. Candidate boundary chains whose length meets the validity constraint and whose direction changes meet the smoothness constraint are selected as valid boundary chain chains. Curve fitting is performed on the effective boundary point chain to generate geometric boundary lines between slope patches. Based on the geometric boundary lines, the normalized three-dimensional point cloud data is divided into multiple slope patches.

2. The method according to claim 1, characterized in that, The raw 3D point cloud data of the open-pit mine slope area is acquired using a ranging sensor mounted on a drone. The raw 3D point cloud data is then normalized to obtain normalized 3D point cloud data, including: The original three-dimensional point cloud data of the open-pit mine slope area is acquired by a ranging sensor carried by a drone. A local neighborhood is established for each point in the original three-dimensional point cloud data, and noise points are identified by calculating the point cloud density distribution characteristics within the local neighborhood. The original 3D point cloud data after removing the noise points is subjected to elevation layering. The point cloud data is divided into multiple elevation intervals according to the elevation direction. Local minimum points are extracted in each elevation interval, and the set of local minimum points of the multiple elevation intervals is used as the bottom boundary points. Based on the spatial distribution characteristics of the bottom boundary points, the bottom boundary points are obtained by calculating the distance relationship and elevation change characteristics between adjacent bottom boundary points, eliminating isolated points and abrupt change points. The topographic reference surface of the slope area is obtained by performing surface fitting on the selected bottom boundary points. The original three-dimensional point cloud data after removing the noise points is then subjected to coordinate transformation using the topographic reference surface as a reference to obtain the normalized three-dimensional point cloud data.

3. The method according to claim 1, characterized in that, For each slope patch, geometric feature parameters characterizing the slope morphology are extracted. Based on these geometric feature parameters, the deformation regions in the slope patch are determined through spatial registration and difference calculation of multi-temporal point cloud data. The displacement vectors of these deformation regions are then calculated, including: For each slope patch, local surface fitting is performed on the point cloud data within the slope patch, and geometric feature parameters characterizing the slope morphology are calculated based on the fitted surface. Acquire multi-temporal point cloud data collected at different times, and extract the point cloud data corresponding to the same slope surface from the multi-temporal point cloud data into a reference temporal point cloud and a target temporal point cloud respectively; Using the geometric feature parameters of the reference time-phase point cloud as registration constraints, the spatial correspondence between the reference time-phase point cloud and the target time-phase point cloud is determined, and a spatial registration transformation is performed on the target time-phase point cloud to obtain the registered target time-phase point cloud; The point-to-point distance between the registered target temporal point cloud and the reference temporal point cloud is calculated to generate point cloud difference results. Based on the point cloud difference results, the deformation areas in the slope surface that have undergone deformation are determined. For each point within the deformation region, the spatial displacement between the corresponding point position in the reference time phase point cloud and the position in the registered target time phase point cloud is calculated to obtain the displacement vector of the deformation region.

4. The method according to claim 3, characterized in that, Calculate the point-to-point distance between the registered target temporal point cloud and the reference temporal point cloud to generate point cloud difference results. Based on the point cloud difference results, determine the deformation areas in the slope surface that have undergone deformation, including: For each point in the registered target temporal point cloud, search for the nearest corresponding point in the reference temporal point cloud, and calculate the Euclidean distance between the point and the corresponding point as the point-to-point distance; Point cloud difference results are obtained based on the point-to-point distances of all points within the slope patch, and these point cloud difference results represent the spatial variation distribution of the slope patch over time. Spatial clustering analysis was performed on the point cloud difference results. By calculating the spatial distribution density of point-to-point distances and neighborhood similarity characteristics, multiple clusters were obtained. Statistical analysis is performed on each cluster. Based on the statistical feature value of the point-to-point distance within the cluster, clusters whose statistical feature values ​​satisfy the deformation detection constraints are selected as candidate deformation regions. Boundary extraction and regional connectivity verification are performed on the candidate deformation regions. Isolated regions whose areas do not meet the validity constraints are eliminated, and the remaining candidate deformation regions are determined as the deformation regions in the slope surface that have undergone deformation.

5. The method according to claim 1, characterized in that, Based on the geometric characteristic parameters and the displacement vector, combined with the slope mechanical constraints, a slope stability assessment result is generated, including the stability level and potential slip direction, comprising: The critical slip angle of each slope patch is calculated based on the slope angle and normal vector in the geometric feature parameters, and the displacement rate and cumulative displacement of the deformation area are calculated based on the direction and amplitude of the displacement vector. By comparing and analyzing the slope angle and the critical slip angle among the geometric characteristic parameters, and combining the displacement rate and cumulative displacement, the evaluation index of the slope's mechanical state is determined. The slope mechanical state evaluation index is weighted and integrated with the slope material strength constraint and slope geometric shape constraint in the slope mechanical constraint conditions to obtain the comprehensive slope stability index. The stability level of the slope section is determined by comparing the comprehensive slope stability index with a preset stability grading threshold. Based on the spatial distribution characteristics of the displacement vector and the normal vector in the geometric feature parameters, the dominant displacement direction of the deformation area is determined as the potential slip direction by calculating the projection component and tangential component of the displacement vector in the direction of the normal vector of the slope patch; The stability level is combined with the potential slip direction to obtain the slope stability assessment result.

6. The method according to claim 5, characterized in that, By comparing and analyzing the slope angle and the critical slip angle among the geometric characteristic parameters, and combining the displacement rate and cumulative displacement, the evaluation indicators of the slope's mechanical state are determined, including: Calculate the angle difference between the slope angle and the critical slip angle, and perform geometric morphology correction on the angle difference based on the slope height and slope length of the slope patch to obtain the geometric safety margin of the slope patch; The displacement rate time series composed of displacement rates at multiple times is subjected to difference operation to calculate the rate change between adjacent displacement rates, and the displacement acceleration sequence is determined based on the rate change. Trend features are extracted from the displacement acceleration sequence. By analyzing the monotonicity and volatility of the displacement acceleration sequence, the stage features of different stages are identified and quantified into changing trend features. The deformation evolution index is obtained by nonlinearly mapping the change trend characteristics to the cumulative displacement, applying a preset time-sensitive weight to the change trend characteristics and coupling it with the cumulative displacement; The geometric safety margin and the deformation evolution index are coupled and calculated to obtain the slope mechanical state evaluation index.

7. An open-pit mine slope feature analysis system based on 3D point cloud and UAV, used to implement the method as described in any one of claims 1-6, characterized in that, include: The first unit is used to acquire raw three-dimensional point cloud data of the open-pit mine slope area through a ranging sensor mounted on a UAV, and to perform normalization processing on the raw three-dimensional point cloud data to obtain normalized three-dimensional point cloud data; The second unit is used to determine the geometric boundary lines between slope patches based on the spatial geometric features of the standardized three-dimensional point cloud data, and to divide the standardized three-dimensional point cloud data into multiple slope patches according to the geometric boundary lines. The third unit is used to extract geometric feature parameters that characterize the slope morphology for each slope patch. Based on the geometric feature parameters, the deformation areas in the slope patch are determined by spatial registration and difference calculation of multi-temporal point cloud data, and the displacement vector of the deformation areas is calculated. The fourth unit is used to generate slope stability assessment results, including stability level and potential slip direction, based on the geometric feature parameters and the displacement vector, combined with slope mechanical constraints.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images