Steep cliff three-dimensional modeling method and system based on structural plane constraint

Abstract

The invention discloses a steep cliff three-dimensional modeling method and system based on structural plane constraint, and relates to the technical field of three-dimensional geographic information system.The steep cliff three-dimensional modeling method comprises the steps that steep cliff area geological data including steep cliff multi-view images and steep cliff surface dense point cloud data are collected, and a geological prior knowledge base is constructed; an image recognition technology is used for processing a steep cliff multi-view image, point cloud geometric analysis is combined, and the structural plane spatial orientation including a rock stratum interface and a joint plane is recognized. According to the method, geological priori knowledge is fused, the geometric distortion problem of a traditional digital elevation model in a cliff area is effectively solved, the spatial orientation of a rock stratum interface and a joint surface is extracted by using an image recognition technology, and the spatial orientation is used as a hard constraint condition to construct a control triangulation network framework; the problems of forced smoothing errors of a regular grid DEM and a suspended surface of a triangular irregular network are avoided, and the rock stratum structure, the crack distribution and the dangerous rock body form of the cliff can be precisely restored through the self-adaptive triangulation algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of three-dimensional geographic information system, in particular to a cliff three-dimensional modeling method and system based on structural plane constraint. BACKGROUND

[0002] The cliff is a high-risk area of landslide, collapse, rockfall and other geological disasters. Three-dimensional modeling can accurately restore the geometric shape, rock structure and crack distribution of the cliff. Combined with geomechanical analysis, it not only helps complex mountain engineering planning, but also helps geological disaster prevention. Through simulating the disaster occurrence process, data support is provided for the early warning system. After the disaster occurs, the three-dimensional model can quickly evaluate the damage range and assist in developing repair schemes to reduce the risk of secondary disasters.

[0003] In the prior art, the digital elevation model (DEM) is used to process the three-dimensional modeling process of the cliff. However, the digital elevation model has a serious geometric distortion problem in the >60° cliff area. The regular grid DEM causes calculation error of the dangerous rock volume due to forced smoothing. The traditional irregular triangular mesh model generates a large number of suspended triangular faces, which destroys the terrain continuity. In addition, commercial software (such as ContextCapture) does not introduce geological constraints, and cannot express the spatial distribution of rock joint surfaces. Therefore, how to solve the geometric distortion problem of the cliff based on the three-dimensional modeling technology that integrates geological prior knowledge, and ensure the accuracy of the three-dimensional modeling of the cliff by combining with the image recognition technology, is a problem to be solved by the present application. Therefore, the present application proposes a cliff three-dimensional modeling method and system based on structural plane constraint. SUMMARY

[0004] The present application aims to provide a cliff three-dimensional modeling method and system based on structural plane constraint to solve the problems raised in the background.

[0005] To solve the above technical problems, the technical solution adopted by the present application is as follows: In a first aspect, the present application provides a cliff three-dimensional modeling method based on structural plane constraint, comprising the following steps: Step 1: Collecting cliff area geological data including cliff multi-view images and cliff surface dense point cloud data, and constructing a geological prior knowledge base; Step 2: Using image recognition technology to process the cliff multi-view images, and combining point cloud geometry analysis to identify the spatial orientation of the structural plane including the rock interface and joint surface; Step 3: Taking the identified structural plane as a hard constraint condition, constructing a control triangular mesh skeleton to ensure that the rock occurrence and joint surface spatial distribution comply with the geological law; Step 4: Combining point cloud density and structural plane constraint, using an adaptive triangular subdivision algorithm to generate a terrain surface, eliminating suspended triangular faces and forced smoothing distortion, and constructing a digital elevation model; Step 5, based on the topological relationship of structural surface, the consistency of the triangular net is checked, the abnormal surface piece in the digital elevation model is removed, the terrain discontinuous area is repaired, and the geometric integrity is ensured; Step 6, by comparing the measured profile line with the point cloud data, the model error is compared, and the structural surface constraint parameter is iteratively optimized; Step 7, the rock texture and joint surface attribute are integrated, and the three-dimensional model of the cliff is output, which meets the accuracy requirement, accurately expresses the cliff characteristics, and fuses the geological constraint.

[0006] Further improvement of the technical scheme of the application is that in step 1, the process of constructing the geological prior knowledge base is: According to the range of the cliff region and the modeling accuracy requirement, a data acquisition plan is formulated, at the same time, a unmanned aerial vehicle for collecting multi-view images of the cliff and a ground laser scanner for collecting point cloud data of the cliff surface are selected, and debugging and calibration are completed to ensure normal operation of the equipment, so as to collect geological data of the cliff region including multi-view images of the cliff and dense point cloud data of the cliff surface. According to the predetermined route and station, the cliff images are shot from different angles by using the unmanned aerial vehicle tilt photography, so that the coverage is comprehensive and there is no serious obstruction, multi-view image data of the cliff are obtained, and the dense point cloud data of the cliff surface are obtained by using the ground laser scanner, and accurate spatial coordinate information is recorded. The collected multi-view images of the cliff and the dense point cloud data of the cliff surface are preprocessed to remove noise and outliers, and the preprocessed data are classified and stored, and a geological prior knowledge base is constructed by integrating and analyzing geological literature materials.

[0007] Further improvement of the technical scheme of the application is that in step 2, the process of identifying the spatial orientation of the structural surface including the rock interface and the joint surface is: The multi-view image data of the cliff are subjected to semantic segmentation, the rock interface and the joint linear feature are identified, the image recognition algorithm of edge detection is used to strengthen the geometric contour of the structural surface, the crack information is enhanced by using high-contrast image, the SfM sparse point cloud of the image is calculated synchronously, and the three-dimensional spatial distribution of the feature points is preliminarily obtained; Based on the dense point cloud data of the cliff surface, the rock interface and the joint surface are divided by using the normal vector estimation and the clustering algorithm, the rock occurrence, i.e. the inclination or the dip angle, is calculated by using the principal component analysis, the two-dimensional feature is extracted from the image, the joint linear feature is mapped to the three-dimensional point cloud by means of spatial registration, the geometric consistency of the structural surface is verified, and the abnormal interpretation result is removed; The structural surface parameters of the cliff region geological data are integrated, the dominant direction and the interval distribution of the joint surface are analyzed by using the pole density diagram, the spatial topological relationship of the structural surface is established, the control geometric network is generated, and the three-dimensional vector model of the rock occurrence and the joint distribution is output.

[0008] The further improvement of the technical scheme of the present application is that in the step 3, the process of constructing the control triangular net framework is: The extracted rock stratum interface and joint surface are converted into three-dimensional polygon boundaries as hard constraint conditions of Delaunay triangulation, so that the triangular net edge is consistent with the structure surface trend, the sharp features of the geological structure are retained, and the control points are inserted through the constraint boundary to ensure that the triangular net forms a topologically correct node at the joint intersection and avoids geometric conflicts; The grid size is dynamically adjusted based on the structure surface density and the point cloud distribution, the triangular net grid is encrypted in the dense area of the joint surface, and the triangular net grid is appropriately relaxed in the gentle area, the Ruppert algorithm is used to optimize the triangular quality, the long and narrow triangle with a minimum angle greater than or equal to 25 degrees is eliminated, and the Laplacian smoothing weight is constrained to prevent the rock stratum occurrence distortion, wherein the dense area of the joint surface is an area with a spacing of less than 0.5 m, and the triangular net grid is encrypted to make the edge length less than or equal to 0.1 m; The continuity of the structure surface intersection line is verified, the self-intersecting or suspended triangular surface is removed, the Alpha Shape algorithm is used to identify the hole and abnormal surface, the discontinuous area is repaired by combining the radial basis function interpolation, and the control triangular net framework meeting the geological rules and retaining the structure details is output.

[0009] The further improvement of the technical scheme of the present application is that in the step 4, the process of constructing the digital elevation model is: The steep cliff surface dense point cloud data is integrated, the point cloud density distribution characteristics are analyzed, the data density degree of different areas is determined, the structure surface information is extracted and converted into a constraint condition; According to the point cloud density and the structure surface constraint, the grid size is dynamically adjusted by using an adaptive triangulation algorithm, the triangular grid is generated by following the constraint at the structure surface, the terrain surface is constructed, and the suspended surface is preliminarily eliminated; The forced smoothing distortion problem existing in the terrain surface is checked and processed, the terrain characteristics are accurately expressed by local adjustment and optimization algorithm, and the high-fidelity digital elevation model meeting the requirements is output.

[0010] The further improvement of the technical scheme of the present application is that in the step 5, the process of removing the abnormal surface in the digital elevation model and repairing the terrain discontinuous area is: A spatial topological network is constructed based on the structure surface intersection line, the topological relationship of the structure surface is extracted, the intersection, parallel or adjacent relationship between the structure surfaces is determined, and the triangular net and the structure surface topological relationship are mapped to mark the associated structure surface of each triangular surface; According to the structure surface topological relationship, whether the triangular net surface meets the consistency rule is checked, the abnormal surface not meeting the rule is marked and removed, and the local rationality of the triangular net is ensured; The position and characteristics of the terrain discontinuous area after the abnormal facet is removed are analyzed, the interpolation and smoothing method is used to repair the terrain discontinuous area in combination with the topological relationship of the surrounding structural surface and the triangulation information, and the geometric integrity of the digital elevation model is ensured.

[0011] The further improvement of the technical scheme of the present application is that the process of marking and removing the abnormal facet not meeting the rules comprises: The structural surface topological network data including structural surface nodes and intersection nodes and topological graph structure are loaded, the triangulation data including vertex coordinates, edge connection relationship and facet index and the like information are loaded, all the facets in the triangulation are traversed, the matching of the geometric properties and the geological constraints is checked, the deviation of the facet normal vector and the associated rock layer occurrence (dip / tilt tolerance is less than or equal to 5 o ), the fitting degree of the boundary edge and the structural surface intersection (distance threshold is less than or equal to 0.02m) and whether there is self-intersection or suspended facet are verified; Whether it meets the requirements is judged through the spatial position association and the normal vector comparison, then the attribute identification bit is set to mark the abnormal facet deviating from the constraint condition, the triangulation is ensured to strictly follow the spatial distribution law of the structural surface, and the position, associated structural surface information and specific non-compliance items of the abnormal facet are recorded in detail; The data structure operation of the adjacency matrix is used, all the marked abnormal facets are found according to the attribute identification bit, the abnormal facets are removed from the triangulation, the topological connection relationship between the remaining facets is maintained at the same time, after the abnormal facets are removed, the connectivity of the triangulation is analyzed, the isolated facets or missing areas are found, the local incomplete area generated is identified, the geometric adjustment and optimization are carried out, and the triangulation is ensured to be reasonable in structure and smooth in connection in the local range, so that the needs of subsequent analysis and application are met.

[0012] The further improvement of the technical scheme of the present application is that the step 6 specifically comprises: The target cliff region is measured on site to obtain profile line data, point cloud data is synchronously collected, the two types of data are subjected to denoising and filtering pretreatment operation, and the coordinate systems are unified to ensure data format compatibility; The measured profile line and the elevation value of the corresponding position in the digital elevation model are compared point by point, the root mean square error and the maximum deviation are calculated, the error distribution law is identified through spatial statistical analysis, the area where the structural surface constraint is insufficient or excessive is located, the error heat map is formed to guide parameter adjustment, the structural surface constraint parameter is iteratively optimized according to the error analysis result, the model form is adjusted, and the deviation between the model and the measured data is gradually reduced; The optimized model and the measured data are compared again to verify whether the error is within the allowable range.

[0013] The further improvement of the technical scheme of the present application is that the step 7 specifically comprises: Based on the rock interface and joint linear features obtained from the cliff area geological data, the rock texture and joint surface properties are analyzed, the rock texture is mapped to the digital elevation model with iterative optimization of the structural surface constraint parameters, the color and geometry are aligned to ensure the initial geological constraint framework is constructed; Based on the structural surface recognition result, the joint linear feature is given the joint surface properties including spacing, continuity and roughness, and the joint surface properties are converted into geometric constraints, the grid topology is adjusted through parameterized modeling to make the structural surface continuous in the model and comply with the geological rules, and the cliff three-dimensional model with semantic labels is formed; The cliff three-dimensional model is exported, compatible with the model format of the geological analysis software, containing the rock texture map, joint attribute table and structural surface topological relationship, ensuring that the cliff three-dimensional model has high-fidelity topographic expression and geological semantic information, and can be directly used for stability analysis or visualization application.

[0014] In the second aspect, the structural surface constraint-based cliff three-dimensional modeling system is used to implement the structural surface constraint-based cliff three-dimensional modeling method described above, comprising: The geological data acquisition and preprocessing module is used to acquire multi-view images and ground laser scanning point cloud data of the cliff area, and construct a geological prior knowledge base combined with geological literature, to provide high-precision and unobstructed cliff area geological data; The structural surface recognition and analysis module is used to recognize the structural surface spatial orientation of the rock interface and joint surface by using image recognition technology and point cloud geometric analysis, accurately recognize the spatial distribution of the structural surface, and provide hard constraint conditions for modeling to avoid geometric distortion; The constrained Delaunay triangulation module is used to construct a control triangular network skeleton by taking the recognized structural surface as a hard constraint condition, to ensure that the occurrence of the rock layer and the spatial distribution of the joint surface comply with the geological rules, and to generate a control triangular network skeleton that complies with the geological structure constraint; The digital elevation model generation and optimization module is used to generate a terrain surface by using an adaptive triangulation algorithm in combination with point cloud density and structural surface constraint, to construct a digital elevation model, to eliminate the geometric distortion problem of the traditional DEM in the cliff area, and to improve the model elevation accuracy; The model iterative optimization module is used to compare the model error by comparing the measured profile line and the point cloud data, to iteratively optimize the structural surface constraint parameters, to generate a cliff three-dimensional model with geometric precision and semantic labels, and to ensure the model accuracy through iterative optimization to reduce the deviation from the measured data.

[0015] Due to the adoption of the above technical solutions, the present application has the following technical progress compared with the prior art: 1. The application provides a steep cliff three-dimensional modeling method and system based on structural plane constraint, which effectively solves the geometric distortion problem of traditional digital elevation model in steep cliff area by fusing geological prior knowledge, uses image recognition technology to extract the spatial orientation of rock layer interface and joint surface, and constructs a control triangular network skeleton as a hard constraint condition, avoids the forced smoothing error of regular grid DEM and the suspended surface problem of irregular triangular network, and through an adaptive triangular division algorithm, can accurately restore the rock structure, crack distribution and dangerous rock body shape of the steep cliff, and provides reliable three-dimensional terrain data support for complex mountain engineering planning.

[0016] 2. The application provides a steep cliff three-dimensional modeling method and system based on structural plane constraint, which constructs a high-fidelity digital elevation model by eliminating suspended triangular surface and forced smoothing distortion based on the modeling technology of structural plane constraint, meets the accuracy requirements of engineering design, can dynamically adjust the grid size, encrypts the triangular network in the joint dense area, relaxes the grid in the gentle area, balances the calculation efficiency and modeling accuracy, so that the model can be directly used for engineering site selection and stability analysis, reduces the engineering rework caused by terrain error, and optimizes the resource utilization efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0018] Figure 1 The working flowchart of the steep cliff three-dimensional modeling method and system based on structural plane constraint of the present application; Figure 2 The method flowchart of the steep cliff three-dimensional modeling method based on structural plane constraint of the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0020] Embodiment 1, as shown in Figure 1 , Figure 2 The present application provides a steep cliff three-dimensional modeling method based on structural plane constraint, which comprises the following steps: Step 1: Collect geological data of the steep cliff area, including multi-view images and dense point cloud data of the steep cliff surface, and construct a geological prior knowledge base. Based on the scope of the steep cliff area and the modeling accuracy requirements, formulate a data collection plan. At the same time, select a UAV for collecting multi-view images of the steep cliff and a ground laser scanner for collecting point cloud data, and complete debugging and calibration to ensure that the equipment is operating normally. Collect geological data of the steep cliff area, including multi-view images and dense point cloud data of the steep cliff surface. According to the predetermined route and station, use UAV oblique photography to take steep cliff images from different perspectives to ensure comprehensive coverage and no serious obstruction, and obtain multi-view image data of the steep cliff. Use a ground laser scanner to obtain dense point cloud data of the steep cliff surface, record accurate spatial coordinate information, preprocess the collected multi-view images and dense point cloud data of the steep cliff surface to remove noise and outliers, and classify and store the preprocessed data. Combine with geological literature, integrate and analyze to construct a geological prior knowledge base. The specific work content includes: Based on the specific scope of the steep cliff area and the required accuracy standards for modeling, developing a data acquisition plan, clarifying the data acquisition tasks, time nodes, and personnel division of labor at different stages to ensure the orderly progress of the acquisition work; simultaneously, conducting equipment selection work, choosing suitable UAVs for the multi-view image acquisition needs of the steep cliff, considering factors such as flight stability and image capture quality; for ground laser scanning point cloud data acquisition, selecting a high-precision ground laser scanner, focusing on its scanning range, accuracy, and speed; after completing the equipment selection, conducting comprehensive debugging and calibration, precisely setting and calibrating the UAV's flight parameters, image capture parameters, and the ground laser scanner's scanning mode and coordinate system to ensure the equipment can operate normally and acquire accurate data in subsequent acquisition work; and planning the UAV flight route and ground laser scanning stations according to the pre-established data acquisition plan, with operators... Unmanned aerial vehicles (UAVs) were used to conduct oblique photography along a predetermined route, capturing images of the steep cliff from multiple different perspectives to ensure comprehensive coverage of the cliff area without severe obstruction. This yielded multi-view image data of the cliff. Simultaneously, ground-based scanning stations were deployed, and ground-based laser scanners were used to scan the cliff surface, acquiring dense point cloud data containing accurate spatial coordinates of each point on the cliff surface. Preprocessing was performed on the acquired multi-view images and dense point cloud data. For the multi-view images, image processing algorithms were used to remove noise and outliers, improving image clarity and quality. For the dense point cloud data, point cloud filtering methods were used to remove noise and outliers, ensuring the accuracy of the point cloud data. The preprocessed image and point cloud data were categorized and stored, and combined with geological literature, the stored data was integrated and analyzed to construct a geological prior knowledge base. Step 2, processing the cliff multi-view image by using image recognition technology, combining point cloud geometry analysis, identifying the spatial orientation of structural surfaces including rock interface and joint surface, performing semantic segmentation on the cliff multi-view image data, identifying rock interface and joint linear features, and using an edge detection image recognition algorithm to strengthen the geometric profile of the structural surface, combining high-contrast image to enhance crack information, synchronously calculating SfM sparse point cloud of the image, preliminarily obtaining the three-dimensional spatial distribution of the feature points, based on the dense point cloud data of the cliff surface, dividing the rock interface and joint surface by normal vector estimation and clustering algorithm, calculating the rock occurrence by principal component analysis, i.e. dip or inclination, combining the two-dimensional features extracted from the image, mapping the joint linear features to the three-dimensional point cloud through spatial registration, verifying the geometric consistency of the structural surface and removing abnormal interpretation results, integrating the structural surface parameters of the cliff area geological data, analyzing the dominant orientation and spacing distribution of the joint surface by using the pole density diagram, establishing the spatial topological relationship of the structural surface, generating the control geometric network, and outputting the three-dimensional vector model of the rock occurrence and joint distribution; Specific work content: perform semantic segmentation on the cliff multi-view image data, identify the rock interface and joint linear features in the image, and use an edge detection image recognition algorithm to strengthen the geometric profile of the structural surface, making the features clearer, at the same time, combine high-contrast images to enhance crack information in the image, improve the ability to identify fine cracks, at the same time, calculate the sparse point cloud of the image based on the structure from motion (SfM) technology, preliminarily obtain the three-dimensional spatial distribution of the feature points; based on the obtained dense point cloud data of the cliff surface, calculate the local geometric features by normal vector estimation (RANSAC), and divide the rock interface and joint surface by using the clustering algorithm (K-means), determine the distribution range of different structural surfaces, calculate the rock occurrence (dip, inclination) by principal component analysis (PCA), determine its spatial orientation, combine the two-dimensional joint linear features extracted from the image, map the joint linear features to the three-dimensional point cloud through spatial registration (ICP algorithm), verify the geometric consistency of the structural surface, remove abnormal interpretation results (pseudo-joints caused by noise or vegetation interference), and ensure the accuracy of structural surface identification; integrate the structural surface parameters of the cliff area geological data, analyze the dominant orientation and spacing distribution of the joint surface by using the pole density diagram, master the spatial distribution rule of the joint surface, and establish the spatial topological relationship between the structural surfaces, construct the control geometric network, present the mutual relationship between the structural surfaces, and then output the three-dimensional vector model of the rock occurrence and joint distribution, directly and accurately display the geological structure characteristics of the cliff area; Step 3: The identified structural plane is taken as a hard constraint condition to construct a control triangular mesh skeleton, to ensure that the occurrence of the rock stratum and the spatial distribution of the joint plane conform to the geological law, to convert the extracted rock stratum interface and joint plane into a three-dimensional polygon boundary, to take the Delaunay triangulation as a hard constraint condition, to force the triangular mesh edge to be consistent with the strike of the structural plane, to retain the sharp features of the geological structure, to insert a control point through the constraint boundary, to ensure that the triangular mesh forms a topologically correct node at the joint intersection, to avoid geometric conflicts, to dynamically adjust the grid size based on the structural plane density and the point cloud distribution, to densify the triangular mesh in the dense area of the joint plane, to appropriately relax in the gentle area, to optimize the triangular quality by using the Ruppert algorithm, to eliminate the long and narrow triangle with a minimum angle ≥ 25°, to constrain the Laplacian smoothing weight at the same time, to prevent the distortion of the occurrence of the rock stratum, wherein the dense area of the joint plane is an area with a spacing < 0.5 m, the triangular mesh is densified to make the edge length ≤ 0.1 m, the continuity of the intersection line of the structural plane is verified, the self-intersecting or suspended triangular plane is removed, the Alpha Shape algorithm is used to identify the hole and the abnormal plane, the discontinuous area is repaired by combining the radial basis function interpolation, and a control triangular mesh skeleton that meets the geological law and retains the structural details is output; Specific work content is: the extracted rock stratum interface and joint surface is converted into three-dimensional closed polygon boundary, as the hard constraint condition of Delaunay triangulation, ensure that the triangulation network strictly adhere to the spatial distribution of geological structure surface, and through the insertion of forced constraint boundary, triangulation network generates nodes that comply with the topological rules at joint intersection, avoid geometric conflict or face overlap, in addition, in order to retain the sharp features of the cliff area (strata fault, joint cutting, etc.), the constraint boundary does not participate in smoothing optimization, maintain the original geometric shape, at the same time, insert control points in the area with varying structure surface density, dynamically adjust the local grid resolution, so that the triangulation network has higher geometric expression ability in complex structure area; Based on the spatial distribution density of structure surface and the local characteristics of point cloud data, the adaptive size field is used to control the triangulation generation, the triangulation grid is encrypted in the dense area of joint surface (spacing <0.5m) (edge length ≤0.1m), and it is relaxed to 0.5m in the flat area, balance the calculation efficiency and modeling accuracy, and optimize the quality of triangular unit through Ruppert algorithm, eliminate narrow triangle (minimum angle ≥25°), ensure numerical stability, in the smoothing stage, constrain the weight coefficient of Laplacian operator, only moderate smoothing for non-structure surface area, prevent the strata occurrence (dip / dip angle) from being distorted due to excessive smoothing, maintain the original geometric characteristics of geological structure; Verify the spatial continuity of structure surface intersection, remove self-intersecting or suspended triangular surface through geometric collision detection, and identify holes and abnormal surface (isolated triangular surface with area mutation) by using Alpha Shape algorithm, judge whether it is a real feature or noise according to the geological law (joint set extension), for discontinuous area, based on radial basis function (RBF), local interpolation reconstruction is carried out, which repairs the terrain fracture while maintaining the sharpness of the structure surface, and then outputs the control triangular network framework that follows the geological structure constraint and ensures the geometric and topological correctness of the rock stratum interface, joint surface and their intersection area; Step 4, combined with point cloud density and structure surface constraint, adaptive triangulation algorithm is used to generate terrain surface, eliminate suspended triangular surface and forced smoothing distortion, build digital elevation model, integrate cliff surface dense point cloud data, analyze point cloud density distribution characteristics, and determine the data density in different areas, at the same time, extract structure surface information and convert it into constraint condition, according to the point cloud density and structure surface constraint, use adaptive triangulation algorithm to dynamically adjust the grid size, generate triangular grid that fits the terrain and preliminarily eliminates suspended surface at structure surface, build terrain surface, check and process the forced smoothing distortion problem existing in the terrain surface, through local adjustment and optimization algorithm, make the terrain features accurately expressed, output the high-fidelity digital elevation model that meets the requirements; The specific work content is: based on the cliff surface dense point cloud data, analyzing the spatial distribution characteristics of the point cloud density, quantifying the density of different regions, and identifying the point cloud sparse region and the point cloud dense region by calculating the number of points in the unit area, for the point cloud sparse region, there are problems such as insufficient data acquisition or shielding; the point cloud dense region may be caused by repeated acquisition or complex terrain; after completing the point cloud density analysis, the structural surface information is extracted from the point cloud dense region, the clustering analysis, normal vector estimation and other methods are used to identify the structural surface point set with similar characteristics, and the structural surface point set is converted into a constraint condition for guiding the terrain surface construction, according to the point cloud density and the structural surface constraint, the adaptive triangulation algorithm is used to dynamically adjust the grid size, in the region with dense point cloud and complex structural surface, smaller grid size is used to accurately express the terrain details; in the region with sparse point cloud and gentle terrain, the grid size is appropriately increased to improve the calculation efficiency, and the constraint condition is strictly followed at the structural surface to ensure that the generated triangular mesh fits the terrain and preliminarily eliminates the suspended surface, and the preliminary terrain surface is constructed; the geometric quality of the preliminary generated terrain surface is checked, local distortion caused by sparse data or constraint conflict is identified and processed, local adjustment and optimization algorithm is used to finely process the distortion region, the node position is adjusted, the triangular shape is optimized and the like, so that the terrain features are accurately expressed, after repeated checking and optimization, it is ensured that the terrain surface meets the actual terrain features and the accuracy requirements, finally, the high-fidelity digital elevation model meeting the requirements is output, which ensures that the digital elevation model meets the elevation accuracy of the point cloud data and follows the spatial distribution law of the structural surface; Step 5, based on the structural surface topological relationship, the consistency of the triangular mesh is checked, the abnormal surface patch in the digital elevation model is removed, the terrain discontinuous region is repaired, and the geometric integrity is ensured; Step 6, by comparing the model error of the measured profile line and the point cloud data, the structural surface constraint parameters are iteratively optimized; Step 7, integrating the rock texture and joint surface attributes, the cliff three-dimensional model meeting the accuracy requirements, accurately expressing the cliff characteristics and fusing the geological constraints is output.

[0021] As shown in the embodiment 1, the application provides a technical solution based on the embodiment 1: Figure 1 , Figure 2 Preferably, in step 5, the process of removing the abnormal surface patch in the digital elevation model and repairing the terrain discontinuous region is: Based on the intersection line of structural plane to construct spatial topology network, extract the topological relationship of structural plane, clear the intersection, parallel or adjacent relationship between structural planes, and map the triangular net and the topological relationship of structural plane, mark the associated structural plane of each triangular facet, check whether the facet in the triangular net meets the consistency rule according to the topological relationship of structural plane, mark and remove the abnormal facet which does not meet the rule, ensure the local rationality of the triangular net, analyze the position and characteristics of the terrain discontinuous area after removing the abnormal facet, repair the terrain discontinuous area by using interpolation and smoothing method combined with the topological relationship of surrounding structural plane and the triangular net information, and guarantee the geometric integrity of the digital elevation model; In addition, the process of marking and removing the abnormal facet which does not meet the rule is as follows: Load the topological network data of structural plane, including the nodes of structural plane and intersection line and the topological graph structure, and load the triangular net data, including the vertex coordinates, edge connection relationship and facet index and other information, traverse all the facets in the triangular net, check the matching of its geometric properties and geological constraints, verify the deviation of the facet normal vector and the associated rock layer occurrence (dip / tilt tolerance ≤5 o], the fit of the boundary edge to the intersection line of the structural plane (distance threshold ≤ 0.02 m), and whether there is self-intersection or hanging surface patch, wherein, for the deviation check of the surface patch normal vector and the associated rock stratum occurrence, the cross product of the vector formed by each surface patch three vertices in the triangular network is calculated to obtain the normal vector, and normalization processing is performed, the coordinates of the surface patch center point are used to search for the closest structural plane node in the structural plane topological network to determine the associated structural plane, the deviation of the surface patch normal vector and the associated rock stratum occurrence is calculated, the surface patch normal vector is converted into a dip and an inclination representation, and is compared with the occurrence of the associated rock stratum to calculate the deviation value, if the deviation exceeds the tolerance range, the surface patch is marked as a normal vector deviation anomaly; for the fit of the boundary edge to the intersection line of the structural plane, the shortest distance from the point on the boundary edge to the intersection line of the structural plane is calculated to obtain the distance to the intersection line of the structural plane, if the distance exceeds the set threshold (0.02m), the face patch is marked as abnormal boundary fit; for self-intersection or hanging face patch checking, in self-intersection checking, a spatial index is used to accelerate the search process, edges and face patches of the triangular net are traversed to check whether there is an edge intersecting a face patch, if a self-intersection condition is found, the relevant face patch is marked as abnormal self-intersection, in hanging face patch checking, it is checked whether a face patch has a connection relationship with a structural face or other face patches, if a face patch is not connected with any structural face or adjacent face patch, or the connection relationship is unreasonable, the face patch is marked as abnormal hanging, through spatial position association and normal vector comparison, it is judged whether it meets the requirements, and then an attribute identification bit is set to mark the abnormal face patch deviating from the constraint condition, to ensure that the triangular net strictly follows the spatial distribution law of the structural face, and the position of the abnormal face patch, associated structural face information and specific non-compliance items are recorded in detail, wherein, an attribute identification bit is set for each triangular face patch, which is used to mark whether the face patch meets the requirements of geometric attributes and geological constraints, the attribute identification bit of the face patch not meeting the requirements is set, the position information of the abnormal face patch is recorded in detail, including the coordinates of the face patch center point, the structural face information associated with the abnormal face patch is recorded, including the structural face number, plane equation, rock layer occurrence, etc., the specific non-compliance items are recorded, including the normal vector deviation value, the distance between the boundary edge and the structural face intersection line, the specific conditions of self-intersection or hanging, etc., using the data structure operation of the adjacency matrix, according to the attribute identification bit, all face patches marked as abnormal are found and removed from the triangular net, when the abnormal face patches are removed, the topological connection relationship between the remaining face patches is maintained, after the abnormal face patches are removed, the connectivity of the triangular net is analyzed to find isolated face patches or missing areas, local incomplete areas are identified, geometric adjustment and optimization are performed to ensure that the triangular net is reasonable in structure and smooth in connection within a local range, meeting the needs of subsequent analysis and application, wherein, the local incomplete area includes a boundary incomplete area and a missing face patch area, for the boundary incomplete area, the position and direction of the boundary edge are adjusted according to the extension trend of the surrounding structural face and the geometric characteristics of the adjacent face patch, so that it naturally transitions with the structural face and the adjacent face patch, for the missing face patch area, an interpolation method is used to generate new vertices, and the face patch is reconstructed to fill the missing area. The specific work involves: constructing a spatial topology network based on the intersection lines of structural surfaces; extracting the spatial connection relationships between structural surfaces, including intersecting, parallel, or adjacent geometric relationships; establishing a topology graph structure by calculating the continuity of the intersection lines and the endpoint connection status, where nodes represent structural surfaces or intersection lines, and edges represent their connection relationships; mapping the triangular mesh to the topology network; and labeling the structural surface type of each triangular facet by spatial location association, ensuring that the vertices and edges of the triangular mesh strictly match the boundaries of the structural surfaces to avoid geometric misalignment. The resulting data structure should simultaneously contain the geometric information of the triangular mesh and the attribute labels of the structural surfaces. After the topology mapping is completed, checking whether the triangular mesh faces conform to the geometric consistency rules of the structural surface constraints, specifically including: whether the facet normal vector is consistent with the attitude of the associated rock strata (dip / dip tolerance ≤ 5°), whether the boundary strictly fits the intersection lines of the structural surfaces, and whether there are any invalid geometric shapes such as overhangs or self-intersections. An automated algorithm is used to traverse all faces and label them. Record non-compliant anomalous areas, such as isolated patches, sharp and distorted triangles, or units that deviate excessively from structural surfaces. Remove the marked anomalous patches and create data holes in the corresponding areas to ensure that the triangulation network conforms to the geological structure within a local range and avoids geometric distortion caused by data noise or constraint conflicts. After removing anomalous patches, repair local discontinuities in the terrain surface by combining the topological relationship of the structural surfaces and the information of the surrounding triangulation network. Identify the boundary features of discontinuous areas, such as fracture lines, steep slopes, or holes, and infer reasonable terrain trends based on the extension trend of adjacent structural surfaces. Use interpolation methods (radial basis functions) to generate elevation points in the hole areas to ensure that the repaired surface transitions naturally with the surrounding structural surfaces. At the same time, perform restricted smoothing on the repaired area, allowing only moderate adjustment of the vertices of non-structural surface areas to maintain the sharp features of rock strata and joints. The output digital elevation model must pass geometric integrity verification to ensure that the terrain is continuous and strictly follows the spatial distribution law of structural surfaces. Step 6 specifically includes: In the target steep cliff area, profile data was obtained through field measurements, and point cloud data was collected simultaneously. Both types of data underwent denoising and filtering preprocessing. A unified coordinate system was implemented to ensure data format compatibility. The measured profile lines were compared point-by-point with the corresponding elevation values ​​in the digital elevation model. The root mean square error and maximum deviation were calculated. Spatial statistical analysis was used to identify error distribution patterns, pinpointing areas with insufficient or excessive structural surface constraints. An error heatmap was generated to guide parameter adjustments. Based on the error analysis results, the structural surface constraint parameters were iteratively optimized to adjust the model shape and gradually reduce the deviation between the model and the measured data. The optimized model was then compared with the measured data again to verify whether the error was within the allowable range. The specific work content is: using total station to measure the target cliff area, obtain high-precision profile line data, collect point cloud data of the area, denoise and filter the two types of data respectively, remove outliers and smooth noise, ensure data quality, unify coordinate system, convert profile line data and point cloud data to the same spatial reference framework to ensure data format compatibility; compare the elevation values of the measured profile line and the corresponding position of the digital elevation model point by point, calculate the key indicators including root mean square error and maximum deviation, quantify the difference between the model and the measured data, generate error heat map through spatial statistical analysis, identify error distribution law, locate the area where the structural plane constraint is insufficient or excessive, based on the error analysis results, adjust the structural plane constraint parameters, iteratively optimize the model shape, gradually reduce the deviation between the model and the measured data, until the error converges to the allowed range, ensure that the model accurately reflects the true geometric characteristics of the cliff; after completing the parameter optimization, compare the optimized model and the measured data again to verify whether the error meets the accuracy requirements, the model must follow the spatial distribution law of the structural plane to ensure geometric consistency and geological rationality; Step 7 specifically includes: Based on the rock layer interface and joint line features obtained from the geological data of the cliff area, analyze the rock texture and joint surface properties, map the rock texture to the digital elevation model with iterative optimization of the structural plane constraint parameters, ensure color and geometry alignment, build an initial geological constraint framework, based on the structural plane recognition result, give the joint line feature a joint surface property, including spacing, continuity and roughness, and convert the joint surface property to geometric constraints, adjust the grid topology through parameterized modeling, make the structural plane continuous in the model and comply with the geological law, form a cliff three-dimensional model with semantic labels, export the cliff three-dimensional model, compatible with the model format of geological analysis software, including rock texture map, joint attribute table and structural plane topological relationship, ensure that the cliff three-dimensional model has high-fidelity topographic expression and geological semantic information, and can be directly used for stability analysis or visualization application; The specific work involves: analyzing the rock strata texture based on the rock interface and joint linear features extracted from geological data of steep cliff areas, determining its color distribution, bedding direction, and thickness variation patterns; mapping the rock strata texture to a digital elevation model using image registration technology, ensuring that the texture coordinates are aligned with the geometric vertices to avoid stretching or misalignment; simultaneously, constructing an initial geological constraint framework by combining iteratively optimized structural surface constraint parameters, so that the model retains topographic details while the rock strata interface presents continuous distribution characteristics in three-dimensional space, ensuring that the texture-mapped model is both visually realistic and reflects the original geometric morphology of the rock strata; and based on the structural surface recognition results, providing joint... Linear features are assigned geological attributes, including parameters such as spacing, continuity, and roughness, which are then transformed into geometric constraints. A parametric modeling method is used to adjust the triangular mesh topology, densifying nodes at joint intersections to ensure spatial continuity of structural surfaces within the model while conforming to geological statistical laws. A constrained Delaunay triangulation algorithm is employed, forcing joint boundaries as hard constraint edges embedded in the mesh to avoid geometric distortion caused by smoothing processes. This generates a steep cliff 3D model containing high-precision terrain surfaces and semantic labels for joint surfaces. The optimized steep cliff 3D model is exported to a format compatible with geological analysis software (OBJ, LAS, or Gocad TSuf), including rock texture maps, joint attribute tables, and structural surface topological relationships. The model retains the integrity of geometric accuracy and semantic information, ensuring that the geometric representation of rock interfaces, joint surfaces, and their intersections conforms to geological laws. The attribute table records parameters such as the attitude and spacing of joint groups, while the topological relationships are stored in a graph structure, describing spatial associations such as intersection and parallelism between structural surfaces. The steep cliff 3D model can be directly used for stability calculations, fracture network analysis, or 3D visualization applications.

[0022] Example 3, as Figure 1 As shown, based on Examples 1-2, the present invention also provides a 3D modeling system for steep cliffs based on structural surface constraints, used to implement the above-mentioned 3D modeling method for steep cliffs based on structural surface constraints, including: The geological data acquisition and preprocessing module is used to acquire multi-view images and ground laser scanning point cloud data of steep cliff areas, and to build a geological prior knowledge base in combination with geological literature to provide high-precision, unobstructed geological data of steep cliff areas. The structural surface identification and analysis module is used to identify the spatial orientation of structural surfaces of rock strata interfaces and joint surfaces using image recognition technology and point cloud geometric analysis, accurately identify the spatial distribution of structural surfaces, provide hard constraints for modeling, avoid geometric distortion, and ensure that the rock strata attitude and joint surface distribution conform to geological laws through spatial topological relationship analysis. The constrained Delaunay triangulation module is used to construct a control triangular net framework by taking the identified structural plane as a hard constraint condition, to ensure that the occurrence of the rock stratum and the spatial distribution of the joint plane comply with the geological law, to generate a control triangular net framework that complies with the geological structure constraint, to retain the sharp features of the rock stratum and the joint, and to avoid the suspended triangular plane and the forced smoothing distortion in the traditional method, and to improve the geometric precision of the model; The digital elevation model generation optimization module is used to generate a terrain surface by using an adaptive triangulation algorithm, to construct a digital elevation model, to eliminate the geometric distortion problem of the traditional DEM in the cliff area, to improve the elevation precision of the model, and to ensure that the terrain surface complies with the geological law through the structural plane constraint; The model iterative optimization module is used to compare the model error by the measured profile line and the point cloud data, to iteratively optimize the structural plane constraint parameter, to generate a cliff three-dimensional model that has both geometric precision and semantic label, and to ensure the model precision through the iterative optimization, and to reduce the deviation from the measured data.

[0023] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A 3D modeling method for steep cliffs based on structural surface constraints, characterized in that, Includes the following steps: Step 1: Collect geological data of the steep cliff area, including multi-view images of the steep cliff and dense point cloud data of the steep cliff surface, and construct a geological prior knowledge base; Step 2: Use image recognition technology to process multi-view images of steep cliffs, and combine point cloud geometric analysis to identify the spatial orientation of structural surfaces, including rock layer interfaces and joint surfaces. Step 3: Using the identified structural surfaces as hard constraints, construct the control triangular mesh framework; Step 4: Combine point cloud density and structural surface constraints, and use an adaptive triangulation algorithm to generate terrain surfaces and construct a digital elevation model; Step 5: Check the consistency of the triangulation network based on the topological relationship of the structural surface, remove abnormal patches in the digital elevation model, and repair discontinuous areas of the terrain. Step 6: By comparing the model error with the measured profile lines and point cloud data, the structural surface constraint parameters are iteratively optimized. Step 7: Integrate rock strata texture and joint surface properties to output a 3D model of the steep cliff that incorporates geological constraints.

2. The method for three-dimensional modeling of steep cliffs based on structural surface constraints according to claim 1, characterized in that: In step 1, the process of constructing the geological prior knowledge base is as follows: Based on the area of ​​the steep cliff and the required modeling accuracy, a data acquisition plan was formulated. At the same time, a UAV for acquiring multi-view images of the steep cliff and a ground laser scanner for acquiring point cloud data were selected and debugged and calibrated to acquire geological data of the steep cliff area, including multi-view images of the steep cliff and dense point cloud data of the steep cliff surface. According to the predetermined route and stations, drones were used to take oblique photos of the steep cliffs from different perspectives to obtain multi-view image data of the steep cliffs. Ground laser scanners were used to obtain dense point cloud data of the steep cliff surface and record accurate spatial coordinate information. The collected multi-view images of steep cliffs and dense point cloud data on the cliff surface were preprocessed to remove noise and outliers. The preprocessed data were then classified and stored. Combined with geological literature, a geological prior knowledge base was constructed through integrated analysis.

3. The method for three-dimensional modeling of steep cliffs based on structural surface constraints according to claim 1, characterized in that: In step 2, the process of identifying the spatial orientation of structural surfaces, including rock strata interfaces and joint surfaces, is as follows: Semantic segmentation was performed on the multi-view image data of the steep cliff to identify the rock layer interface and joint linear features. An image recognition algorithm with edge detection was used to enhance the geometric contour of the structural surface. The fracture information was enhanced by combining high-contrast images. The SfM sparse point cloud of the image was calculated simultaneously to obtain the three-dimensional spatial distribution of feature points. Based on dense point cloud data of steep cliff surface, rock layer interfaces and joint surfaces are divided by normal vector estimation and clustering algorithm. Principal component analysis is used to calculate the rock layer attitude, i.e. dip or dip angle. Combined with two-dimensional features extracted from images, the linear joint features are mapped to three-dimensional point cloud through spatial registration to verify the geometric consistency of structural surfaces and remove abnormal interpretation results. By integrating the structural surface parameters of geological data in steep cliff areas, the dominant orientation and spacing distribution of joint surfaces are analyzed using extreme density maps. The spatial topological relationship of structural surfaces is established, a controlling geometric network is generated, and a three-dimensional vector model of rock strata attitude and joint distribution is output.

4. The method for three-dimensional modeling of steep cliffs based on structural surface constraints according to claim 1, characterized in that: In step 3, the process of constructing the control triangular mesh skeleton is as follows: The extracted rock strata interfaces and joint surfaces are converted into three-dimensional polygon boundaries, which serve as hard constraints for Delaunay triangulation, forcing the edges of the triangular mesh to align with the structural surfaces, and control points are inserted through the constraint boundaries. The mesh size is dynamically adjusted based on the density of the structural surface and the distribution of the point cloud. The triangular mesh is densified in the dense area of ​​the joint surface. The Ruppert algorithm is used to optimize the quality of the triangles and eliminate narrow triangles with a minimum angle ≥ 25°. The dense area of ​​the joint surface is the region with a spacing < 0.5m. The triangular mesh is densified to make the side length ≤ 0.1m. Verify the continuity of the intersection lines of the structural surfaces, eliminate self-intersecting or overhanging triangular surfaces, identify holes and abnormal patches using the Alpha Shape algorithm, repair discontinuous regions by combining radial basis function interpolation, and output the control triangular mesh skeleton.

5. The method for three-dimensional modeling of steep cliffs based on structural surface constraints according to claim 4, characterized in that: In step 4, the process of constructing the digital elevation model is as follows: Integrate dense point cloud data on steep cliff surfaces, analyze point cloud density distribution characteristics, clarify the density of data in different areas, and extract structural surface information and convert it into constraints. Based on point cloud density and structural surface constraints, an adaptive triangulation algorithm is used to dynamically adjust the mesh size, generate a triangular mesh that fits the terrain and initially eliminates overhanging surfaces, and construct the terrain surface. The forced smoothing distortion problem in the terrain surface is checked and addressed. Through local adjustment and optimization algorithms, a high-fidelity digital elevation model that meets the requirements is output.

6. The method for three-dimensional modeling of steep cliffs based on structural surface constraints according to claim 4, characterized in that: In step 5, the process of removing abnormal patches from the digital elevation model and repairing discontinuous terrain areas is as follows: A spatial topology network is constructed based on the intersection lines of structural surfaces. The topological relationships of structural surfaces are extracted, and the intersecting, parallel, or adjacent relationships between structural surfaces are clarified. The triangular network is then mapped to the topological relationships of structural surfaces, and the structural surfaces associated with each triangular facet are marked. Based on the topological relationship of the structural surfaces, check whether the facets in the triangular mesh conform to the consistency rules, and mark and remove abnormal facets that do not conform to the rules. After removing abnormal patches, the location and characteristics of the terrain discontinuities are analyzed. Based on the topological relationships of the surrounding structural surfaces and triangular network information, interpolation and smoothing methods are used to repair the terrain discontinuities.

7. The method for three-dimensional modeling of steep cliffs based on structural surface constraints according to claim 6, characterized in that: The process of marking and removing abnormal patches that do not conform to the rules is as follows: Load the structural surface topology network data, including structural surface nodes, intersection nodes, and topology graph structure, and load the triangular mesh data. Traverse all the facets in the triangular mesh, check the matching of their geometric properties with geological constraints, verify the deviation of the facet normal vector from the attitude of the associated rock strata, the fit of the boundary edge with the intersection line of the structural surface, and whether there are self-intersecting or suspended facets. By comparing spatial location with normal vectors, it is determined whether the requirements are met. Then, attribute flags are set to mark abnormal surfaces that deviate from the constraints, and the location, associated structural surface information, and specific non-compliance items of the abnormal surfaces are recorded in detail. Using the adjacency matrix data structure, based on attribute identifier bits, all faces marked as abnormal are found and removed from the triangulation. While removing abnormal faces, the topological connectivity between the remaining faces is maintained. After removing abnormal faces, the connectivity of the triangulation is analyzed to find isolated faces or missing regions, identify the resulting local incomplete regions, and perform geometric adjustments and optimizations.

8. The method for three-dimensional modeling of steep cliffs based on structural surface constraints according to claim 7, characterized in that: Step 6 specifically includes: The target steep cliff area was measured in the field to obtain profile data, and point cloud data was collected simultaneously. The two types of data were preprocessed by denoising and filtering, and the coordinate system was unified. The measured profile line is compared point by point with the corresponding elevation value in the digital elevation model. The root mean square error and maximum deviation are calculated. The error distribution pattern is identified through spatial statistical analysis. Areas with insufficient or excessive structural surface constraints are located. An error heat map is generated to guide parameter adjustment. Based on the error analysis results, the structural surface constraint parameters are iteratively optimized in a targeted manner, the model shape is adjusted, and the deviation between the model and the measured data is gradually reduced. The optimized model was compared with the measured data again to verify whether the error was within the allowable range.

9. The method for three-dimensional modeling of steep cliffs based on structural surface constraints according to claim 3, characterized in that: Step 7 specifically includes: Based on the rock layer interface and joint linear features obtained from the geological data of the steep cliff area, the rock layer texture and joint surface properties are analyzed, and the rock layer texture is mapped to the digital elevation model that iteratively optimizes the structural surface constraint parameters to construct an initial geological constraint framework. Based on the structural surface identification results, joint surface attributes, including spacing, continuity and roughness, are assigned to the linear features of joints. The joint surface attributes are then transformed into geometric constraints. The mesh topology is adjusted through parametric modeling to make the structural surfaces continuous in the model and conform to geological laws, thus forming a 3D model of a steep cliff with semantic labels. Export a 3D model of a steep cliff, compatible with the model format of geological analysis software, including rock texture maps, joint attribute tables, and structural surface topological relationships.

10. A 3D modeling system for steep cliffs based on structural surface constraints, used to implement the 3D modeling method for steep cliffs based on structural surface constraints as described in any one of claims 1-9, characterized in that, include: The geological data acquisition and preprocessing module is used to acquire multi-view images and ground laser scanning point cloud data of steep cliff areas, and to build a geological prior knowledge base in combination with geological literature. The structural surface identification and analysis module is used to identify the spatial orientation of structural surfaces of rock strata interfaces and joint surfaces by using image recognition technology and point cloud geometric analysis. The constrained Delaunay triangulation module is used to construct a control triangulation framework by using the identified structural surfaces as hard constraints. The digital elevation model generation and optimization module is used to combine point cloud density and structural surface constraints, and to generate terrain surfaces using an adaptive triangulation algorithm to construct a digital elevation model. The model iteration and optimization module is used to compare the model error with the measured profile lines and point cloud data, iteratively optimize the structural surface constraint parameters, and generate a steep cliff 3D model with both geometric accuracy and semantic labels.

Images