Live-action three-dimensional intelligent analysis system for dangerous areas of hidden danger points of geological disasters

Abstract

The invention relates to the technical field of geographic information systems, and discloses a real-scene three-dimensional intelligent analysis system for a risk area of a hidden danger point of a ground disaster, and the system comprises a data obtaining module which is used for obtaining a real-scene three-dimensional grid model and a digital elevation model of a to-be-monitored area; the topological feature extraction module is used for extracting side length statistical distribution features of the grid units and generating topological strain entropies representing geometric evolution laws; the risk analysis decision module is used for executing gradient projection operation on the topological strain entropy by utilizing the elevation field gradient vector so as to construct an attribute response deformation field, and determining an evaluation instruction according to the space-time cooperation probability between the image optical flow field and the attribute response deformation field; according to the method, the disorder degree evolution trend of the geometric characteristics of the grid units is excavated to capture the topological distortion induced by the displacement of the rock-soil body, and the recognition precision of the system on early creep deformation is improved; non-physical topology jump is stripped through a causal association model of gravity driving logic and image texture change, and the early warning confidence coefficient in a complex environment is enhanced.

Description

Technical Field

[0001] This invention relates to a smart analysis system for geological disaster hazard risk areas based on real-world 3D scenes, belonging to the field of geographic information system technology. Background Technology

[0002] Currently, using oblique photography to generate 3D mesh models to characterize the topographic features of geographic entities has become the mainstream approach. By comparing the geometric changes of mesh models generated at different times, the deformation trend of potential hazard points can be inferred. However, when 3D mesh models are applied in complex geological environments, they face the dilemma of a decoupling between geometric representation accuracy and physical and mechanical reliability.

[0003] Existing technologies treat 3D mesh models as static geometric envelopes, and their recognition logic relies on the comparison of absolute displacements of mesh nodes at different time phases. However, in real-world engineering scenarios, UAV aerial surveys are affected by changes in ambient lighting, shadow shifts, or vegetation swaying, causing non-physical topological jitter in the modeling algorithm during mesh reconstruction. This topological artifact generated by the modeling mechanism has highly overlapping geometric features with early soil and rock creep, making it difficult for the system to distinguish between spontaneous mesh distortion and actual physical displacement. Solving this by adding external sensors or improving full-field scanning accuracy leads to a significant increase in computational costs and insufficient edge processing performance. Currently, the industry is attempting to use filtering algorithms or curvature analysis to remove model noise, but these methods ignore the mechanical properties inherent in the mesh topology. Logically, geometric smoothing can easily lead to the loss of key deformation features and fails to establish a physical correlation between grid evolution and gravity-driven instability. Existing analytical logic is limited to data correlation and probability extrapolation, failing to touch the physical essence of terrain evolution. For example, Chinese invention patent CN118095866B discloses a method and system for assessing the risk of landslides in disaster-bearing bodies based on artificial intelligence. It uses landslide induction factors and shared coefficient indicators for quantitative analysis, treating the three-dimensional grid as an isolated set of data points at the surface level. The statistical regression-based assessment model lacks in-depth mining of the mechanical properties of the grid topology and cannot establish a causal mapping between deformation vectors and gravity field constraints. When faced with endogenous topological noise in the modeling algorithm, it is difficult to accurately remove deformation signals, resulting in a bottleneck in the physical reliability of early warning.

[0004] Therefore, the technical problem to be solved by this invention is how to decouple deformation signals from modeling noise by mining the inherent topological evolution laws of three-dimensional mesh models and combining them with the physical constraints of gravitational fields, thereby improving the confidence of intelligent analysis of risk areas. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A smart analysis system for geological disaster hazard point risk areas based on real-scene 3D, comprising:

[0006] The data acquisition module is used to acquire a real-world 3D grid model and a digital elevation model of the geospatial area to be monitored.

[0007] The topology feature extraction module is used to extract the statistical distribution features of the side lengths of the grid cells in the real-world 3D mesh model and generate the topology strain entropy that reflects the geometric evolution law of the real-world 3D mesh model.

[0008] The risk analysis and decision-making module includes an elevation field coupling solution unit and a signal arbitration unit;

[0009] The elevation field coupled solution unit is used to retrieve the elevation field gradient vector in the digital elevation model and perform gradient projection operation on the topological strain entropy in the direction of the elevation field gradient vector to construct the attribute response deformation field after eliminating spatial modeling noise.

[0010] The signal arbitration unit is used to extract the image texture of the image in the real-world 3D mesh model, establish a causal relationship model between geometric deformation and image texture changes, and output an evaluation command that characterizes the risk status of the geospatial area to be monitored by combining the attribute response deformation field and the quantitative judgment rules of the preset risk threshold. Among them, the signal arbitration unit constructs the image optical flow field according to the displacement vector of the image texture in the spatial coordinate system, and determines the trigger level of the evaluation command according to the spatiotemporal coordination probability between the characteristic jump frequency of the attribute response deformation field and the physical response of the image optical flow field. The topological strain entropy is characterized by calculating the evolution value of the statistical distribution variance of the grid side length in adjacent time phases. The gradient projection operation extracts the three-dimensional components of the elevation field gradient vector in the geographic coordinate system and maps the evolution value to the motion vector direction that satisfies the gravity constraint to eliminate isotropic algorithm artifact noise.

[0011] Preferably, the topology feature extraction module includes a dynamic resampling unit; the dynamic resampling unit adjusts the subdivision depth of the spatial index according to the curvature distribution gradient of the surface of the real-world 3D mesh model; in areas where the curvature change is greater than a preset curvature threshold, the dynamic resampling unit increases the mesh topology connection density; in areas where the curvature change is less than or equal to the preset curvature threshold, the dynamic resampling unit performs topology thinning to merge redundant patches, thereby reducing data redundancy.

[0012] Preferably, when constructing the attribute response deformation field, the elevation field coupled solution unit performs the following anisotropic calibration steps: Step S1, calculate the statistical distribution variance of the mesh edge length between two real-world 3D mesh models at different time phases; Step S2, extract the gravity field vector projection at the corresponding location in the digital elevation model; Step S3, calculate the projection weight coefficient to obtain the quantized components of the attribute response deformation field; the calculation formula is: ,in, ΔH is the projection weighting coefficient, ΔH is the evolution value of the statistical distribution variance of the grid side length within the preset observation period, and θ is the angle between the evolution vector of the statistical distribution variance of the grid side length and the projection of the gravity field vector.

[0013] Preferably, when the attribute response deformation field produces a characteristic jump, the signal arbitration unit simultaneously retrieves the image optical flow field to perform image consistency verification; when the displacement characteristics of the image optical flow field and the evolution characteristics of the attribute response deformation field meet the preset spatiotemporal correlation criteria, the signal arbitration unit confirms it as a geographic entity displacement signal.

[0014] Preferably, the risk analysis and decision-making module also includes a media adaptive compensation unit; the media adaptive compensation unit performs fractal dimension analysis on the real-world 3D mesh model and identifies the surface media type based on the results of the fractal dimension analysis; the media adaptive compensation unit dynamically modulates the preset risk threshold according to the surface media type; for areas identified as loose deposits, the media adaptive compensation unit lowers the preset risk threshold by 10% to 20%.

[0015] Preferably, when calculating the topological strain entropy, the topological feature extraction module identifies the topological connectivity of each grid cell in the real-world 3D grid model and the statistical distribution of grid edge lengths; the topological strain entropy is used to characterize the degree of evolution of the geometric shape of the monitored geographic space area from order to disorder, so as to identify the topological singularity features caused by the displacement of geographic entities.

[0016] Preferably, the data acquisition module further includes a temporal synchronization subunit; the temporal synchronization subunit is used to perform spatial reference alignment and illumination consistency correction on multi-phase real-scene 3D mesh models acquired at different time nodes, and input the corrected mesh data into the topology feature extraction module.

[0017] Preferably, the risk analysis and decision-making module is also used to calculate the deformation acceleration of the attribute response deformation field; when the value of the deformation acceleration is continuously greater than 0 mm / s within a preset observation period. 2 At that time, the risk analysis and decision-making module determined that the geospatial area to be monitored had entered an accelerated deformation phase.

[0018] Preferably, the risk analysis and decision-making module also includes a risk zone delineation unit; the risk zone delineation unit identifies geometrically closed-loop regions affected by deformation based on the spatial connectivity analysis of the attribute response deformation field, and maps the geometrically closed-loop regions into independent deformation target feature boundaries in geographic space.

[0019] Preferably, the risk zone delineation unit is also used to perform potential energy analysis based on the terrain slope and elevation difference characteristics within the deformation target feature boundary, and to calculate the potential impact range after the deformation target feature boundary becomes unstable based on the results of the potential energy analysis; the input of the potential energy analysis is dynamically corrected by the real-time topological displacement provided by the attribute response deformation field.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] 1. In the intelligent analysis of risk areas of potential geological hazards, this invention enables the transition of geographic entities from visual feature reproduction to mechanical response perception. The grid side length distribution entropy operator constructed in this invention transforms the static geometric envelope of the real-world 3D grid model into a topological field with strain response properties. Since the system does not rely on the absolute coordinate offset of nodes, but extracts the disorder evolution trend of the geometric features on the grid surface, the system can capture the topological distortion induced by the micro-displacement of the soil and rock mass. This monitoring logic based on the principle of statistical topology improves the geographic information system's ability to identify early creep of potential hazard points without increasing the load on physical sensors.

[0022] 2. A modeling error suppression mechanism based on gravity-driven logic was constructed. By extracting the anisotropic principal vector of the grid side length evolution in the geographic feature subclusters and performing cosine similarity calculation with the local terrain slope vector, this invention introduces gravitational field consistency constraints at the geometric analysis level. This deep contrarian relationship between physical mechanism and geometric topology enables the system to extract real deformation features that conform to the geological instability law from the chaotic grid evolution signal. For non-physical topological jumps caused by changes in ambient light or vegetation shading, since they do not meet the slope aspect consistency discrimination criterion, the system automatically judges them as modeling algorithm errors and suppresses them, thereby improving the early warning confidence in complex field environments.

[0023] 3. Achieving a nonlinear dynamic balance between analytical accuracy and computational efficiency, this invention introduces a spatial unit dynamic resampling mechanism based on geometric curvature gradients. According to the curvature distribution of the 3D model surface, the subdivision depth of the spatial index is automatically adjusted. In critical deformation areas with high terrain complexity, such as cliffs and ravines, the system enhances the sensitivity to capturing small geometric distortions by encrypting topological connections. In flat areas, topological thinning is performed to merge redundant patches. This constraint-driven computational resource allocation strategy eliminates background noise interference from redundant data in risk identification in large-scale geographical scenes. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall process of the intelligent analysis system for geological disaster hazard points and risk areas based on real-scene 3D based on the present invention;

[0025] Figure 2 This is a schematic diagram of the signal arbitration and decision logic of the present invention. Detailed Implementation

[0026] The technical solution proposed by the present invention will be described in detail below with reference to the accompanying drawings. The following embodiments are intended to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0027] A smart analysis system for geological disaster hazard risk areas based on real-world 3D models comprises a data acquisition module, a topology feature extraction module, and a risk analysis and decision-making module. The data acquisition module acquires a real-world 3D grid model and a digital elevation model (DEM) of the geographic area to be monitored. The topology feature extraction module extracts the statistical distribution characteristics of the side lengths of the grid cells in the real-world 3D grid model and generates a topology strain entropy reflecting the geometric evolution of the real-world 3D grid model. The risk analysis and decision-making module consists of an elevation field coupling solution unit and a signal arbitration unit. The elevation field coupling solution unit retrieves the elevation field gradient vector from the DEM and applies the topology strain entropy along the elevation field... The gradient vector direction performs gradient projection operations to construct the attribute response deformation field. The signal arbitration unit is used to establish a causal relationship model between geometric deformation and image texture changes, and outputs evaluation instructions characterizing the risk status of the geospatial area to be monitored by combining the attribute response deformation field with the quantitative judgment rules of the preset risk threshold. In the process of identifying the geospatial area to be monitored, the topology feature extraction module extracts the statistical distribution features of the side lengths of the grid cells in the real-scene 3D grid model, and calculates the evolution value of the statistical distribution variance of the grid side lengths in adjacent time phases. The topology feature extraction module generates topological strain entropy by calculating the disorder change of the grid side lengths within each geographic feature subcluster. Topological strain entropy The calculation formula is as follows: ,in, For topological strain entropy, For the first The length of the grid edge, The average length of the grid edges within the current geographic feature subcluster, n is the total number of grid edges, and the topological strain entropy is... It is used to characterize the degree of evolution of the geometric shape of the monitored geospatial area from order to disorder, thereby identifying topological distortions caused by the displacement of geographic entities.

[0028] Determine the topological strain entropy When evolving the vector, the grid vertex indices within the geographic feature subclusters under adjacent monitoring time phases are traversed to extract the geometric center coordinates of the geographic feature subclusters in the first time phase. Coordinates of the second phase geometric center Construct a unit displacement direction vector based on the coordinate difference. ,in , topological strain entropy The scalar evolution value acts as the modulus on the direction vector of the unit displacement. Generate the topology evolution vector for performing gradient projection operations. ,in The randomness variation of grid statistical characteristics is transformed into spatially directional motion characteristics, eliminating the directional gap problem of scalar parameters in subsequent gravity field constraint calculations. To address topological jitter noise generated during modeling, the elevation field coupled solution unit retrieves the elevation field gradient vector from the digital elevation model and incorporates the topological strain entropy. Gradient projection is performed along the direction of the elevation field gradient vector. By extracting the three-dimensional components of the elevation field gradient vector in the geographic coordinate system, the topological strain entropy is obtained. The evolution value is mapped to the motion vector direction that satisfies gravity constraints to eliminate isotropic algorithm artifact noise. The quantization components of the property response deformation field are calculated by projection weighting coefficients. The formula for obtaining and calculating this is as follows: ,in, Here, ΔH represents the projection weighting coefficient, and ΔH represents the topological strain entropy. The evolution value within the preset observation period, where θ is the topological strain entropy. The angle between the evolution vector and the projection direction of the gravity field vector.

[0029] When the attribute response deformation field generates a characteristic jump, the signal arbitration unit simultaneously retrieves the image optical flow field to perform image consistency verification. The image optical flow field is constructed based on the displacement vector of the image texture in the spatial coordinate system. The signal arbitration unit determines the trigger level of the evaluation command based on the spatiotemporal coordination probability between the characteristic jump frequency of the attribute response deformation field and the physical response of the image optical flow field. When the displacement characteristics of the image optical flow field and the evolution characteristics of the attribute response deformation field meet the preset spatiotemporal correlation criteria, the signal arbitration unit confirms that the current signal is a geographic entity displacement signal and outputs the evaluation command. The signal arbitration unit calculates the displacement vector in the image optical flow field. With topological evolution vector Cosine similarity determines the collaborative operator Collaborative operators According to the formula Calculation, where For topological evolution vectors, For image texture displacement vector, when the cooperative operator It is in the range of 0.75 to 1.0, and the topological strain entropy is... When the cross-correlation coefficient R between the instantaneous growth rate and the image pixel displacement exceeds 0.82, the physical displacement signal is confirmed and an evaluation command is triggered. If the cooperative operator... If the value is below 0.35, the signal is determined to be a topological artifact generated by the modeling algorithm, establishing a mutual verification mechanism based on pixel-level displacement and grid-level entropy change. To further adapt to complex geological environments, the risk analysis and decision-making module includes a media adaptive compensation unit. This unit performs fractal dimension analysis on the real-world 3D mesh model and identifies the surface medium type based on the analysis results. It identifies the surface medium as bedrock or loose deposits. For areas identified as loose deposits, the preset risk threshold is lowered by 10% to 20%, thus achieving adaptive compensation for the deformation sensitivity of different geological media. The media adaptive compensation unit adjusts the risk threshold based on the fractal dimension... Quantization mapping relationship between surface media types modulates preset risk thresholds During the stable period of the geospatial area to be monitored, multi-scale box counting scans were performed to obtain the background fractal dimension baseline value. When real-time monitoring obtains fractal dimensions satisfy Furthermore, when the local average curvature gradient is greater than 0.18 rad / cm, the current geographic entity is identified as a loose aggregate, and the corresponding preset risk threshold is lowered. The original value is reduced to 85%. Based on the feedback loop of geometric texture complexity parameter, sensitivity compensation is performed according to the difference in topological sensitivity of geological structure, so as to achieve differentiated early warning of strata with different mechanical strengths without increasing the load on physical sensors.

[0030] The topology feature extraction module also includes a dynamic resampling unit. This unit adjusts the spatial index subdivision depth based on the curvature distribution gradient of the real-world 3D mesh model surface. In regions where the curvature change exceeds a preset curvature threshold, the dynamic resampling unit increases the mesh topology connection density. In regions where the curvature change is less than or equal to the preset curvature threshold, the dynamic resampling unit merges redundant patches to reduce data redundancy. Furthermore, the data acquisition module includes a temporal synchronization subunit to perform spatial reference alignment and illumination consistency correction on multi-phase real-world 3D mesh models acquired at different time points, ensuring the spatiotemporal consistency of the data input to the topology feature extraction module. In the data preprocessing procedure, time... The phase sub-unit uses global bundle adjustment to correct the spatial relative position deviation between multiple phases of real-scene 3D mesh models. It selects no less than 15 corresponding feature points in the real-scene 3D mesh model and uses nonlinear least squares method to calculate the correction of camera extrinsic parameters, so that the average reprojection error Δe of the multiple phase models in the geographic coordinate system is kept within 1.5 pixels, where Δe is the average reprojection error. The system executes a lighting consistency correction procedure, and establishes a gray-level linear transformation mapping function by calculating the histogram cross-correlation coefficient between the reference image and the image to be corrected. This maps the illumination intensity under different time phases to a unified gray-level reference plane to eliminate the pseudo-topological distortion signal caused by shadow offset.

[0031] Example 1: In the scenario of monitoring potential hazards on steep slopes covered with trees, the wind in the aerial survey environment causes vegetation to shift at different times. This results in random grid distortion when the modeling software reconstructs the surface of geographic entities. This non-physical displacement-induced grid debris jitter overlaps geometrically with the early creep signals generated by the soil and rock mass, leading to false alarms of inaccurate deformation in conventional monitoring methods based on 3D coordinate point comparisons. To address this application scenario, the technical solution of this invention uses a topological feature extraction module to perform topological subdivision on the multi-phase real-scene 3D grid model of the area, extracting the statistical distribution characteristics of the side lengths of each grid cell and calculating its evolution value, generating a topological strain entropy that reflects the disordered evolution of the geometric morphology. The elevation field gradient vector in the digital elevation model is retrieved by the elevation field coupling solution unit, and the topological strain entropy is then calculated. The evolution vector is projected onto the slope gradient direction determined by the digital elevation model, and the projection weight coefficient is obtained by calculation. Projection weighting coefficient The calculation formula is as follows: ,in, ΔH is the projection weighting coefficient, which measures the consistency between the deformation vector and the direction of gravity driving; ΔH is the topological strain entropy. The evolution value within the monitoring period is used to characterize the change in geometric disorder; θ is the topological strain entropy. The angle between the evolution vector and the direction of the local terrain slope.

[0032] When performing the above projection calculation, the mesh connection distortion caused by vegetation movement has a random spatial distribution, and the angle θ between its evolution vector and the local slope direction exhibits isotropic characteristics. The calculated projection weight coefficients... In the low-level range below 0.15, where the direction of soil and rock creep driven by gravity tends to be the same as the direction of topographic slope, the projected weighting coefficient for this region is... The system stably exceeds a judgment threshold of 0.7. By utilizing the coupling of physical constraints of the gravitational field and geometric topological features, it achieves decoupling judgment of deformation signals and modeling noise without the use of external strain sensors. The signal arbitration unit constructs an image optical flow field based on the image texture of the image in the real-world 3D mesh model, and the projection weight coefficient... When the transition is triggered, the system simultaneously analyzes the displacement characteristics and topological strain entropy of the image optical flow field. The spatiotemporal coordination probability between evolutionary features is verified by verifying that the image pixel-level stretching and the evolution of grid topology meet the preset spatiotemporal correlation criteria, eliminating topological artifacts caused by illumination angle offset, and anchoring the output of evaluation instructions to the physically meaningful geographic entity displacement facts. The present invention reconstructs the static geometric envelope into a dynamic topological strain field, and uses the gravitational field gradient as a constraint operator for topological evolution to establish a geographic information analysis procedure based on mechanical logic, so that the geographic information system has a logical closed loop from local visual reproduction to overall strain response perception.

[0033] Example 2: Monitoring was conducted on a Quaternary loose deposit slope with vegetation cover in an indoor physical simulation platform landslide creep test field. The test platform was equipped with a sensor system with a measurement resolution of at least 0.05 mm and a sampling frequency of 1 Hz. The raw test data was obtained from a high-precision oblique photogrammetry device with a ground resolution better than 1.5 cm. A miniature vegetation model driven by wind was deployed on the slope model surface to introduce non-physical topological connection jump noise generated by the modeling algorithm. The sampling period T was determined by a balance logic between the motion characteristics of geographic entities and the data processing load. As the predicted deformation rate increased, the sampling period T tended towards the lower limit of its range, i.e., 1 hour. The formula for calculating the sampling period T is as follows: Where T is the sampling period, Here, V represents the characteristic coefficients, and V represents the modeling resolution. This represents the maximum deformation rate of a geographic entity.

[0034] The experimental group used topological strain entropy A complete solution for extraction, elevation field gradient projection calibration, and image optical flow field arbitration. Control group 1 removes the elevation field gradient projection calibration step, while control group 2 sets projection weight coefficients. The consistency threshold was 0.95. Each group was subjected to an early creep stage where the actual deformation rate remained between 0.8 mm / h and 1.2 mm / h, and mesh coordinate noise with an amplitude of 2.5 mm was input. The initial topological strain entropy extracted from the experimental groups... Including the 0.42 entropy change caused by coordinate noise and the 0.18 entropy change caused by displacement, after gradient projection calculation by the elevation field coupled solution unit, the projection weight coefficient of the noise component in the slope vector direction is... The mean value is 0.12, which is within the suppression range for non-physical deformation determination, while the projection weighting coefficient of the displacement component... The mean value is 0.74, and the temporal co-probability of displacement features and topological evolution features of the image optical flow field is 0.92, satisfying the spatiotemporal correlation criterion. The judgment results show that the recognition accuracy of the experimental group is 96.4% and the false alarm rate is less than 2.1%. Due to the lack of slope direction constraints, the control group 1 could not distinguish between isotropic modeling artifacts and anisotropic deformations, and its false alarm rate increased to 33.8%. The control group 2, due to the excessively high consistency threshold setting, ignored the lateral displacement caused by the non-uniform stress release of the soil, resulting in an effective capture rate of 14.8%. This confirms that the 0.7 threshold defined in this invention is the working window for capturing gravity-driven deformation. When the deformation rate increases to 150 mm / h, the topological connection stability of the real-scene 3D mesh model decreases, and the topological strain entropy increases. The growth rate tends to saturate, indicating that the mesh reconstruction caused by large displacement makes the calculation of geometric disorder lose its benchmark, thus confirming the technical logic of the system to perform risk zone analysis on early creep.

[0035] Example 3: In a slope monitoring scenario with karst landforms and uneven strata exposure, the area includes exposed bedrock and gravel-soil embankments of varying thicknesses. The real-world 3D mesh model input to the data acquisition module contains 5.0 × 10⁻⁶ Å. 5 Each triangular facet has an average side length of 2.5 cm. To determine the execution procedure for the topology feature extraction process, the topology feature extraction module traverses all vertex indices of the real-world 3D mesh model, identifies the edge sequence connected to each vertex, and divides the geospatial area to be monitored into geographic feature subclusters containing 1000 adjacent triangular facets. The topology feature extraction module then reads the side length of each mesh unit within each geographic feature subcluster. The data is stored in a one-dimensional floating-point array, and the topological strain entropy is generated by calculating the variance fluctuation of this one-dimensional floating-point array between adjacent monitoring periods. This transforms continuous geometric deformations into discrete statistical topological features. Secondly, the medium-adaptive compensation unit performs fractal dimension analysis on the aforementioned geographic feature subclusters, using box counting to perform multi-scale coverage on the surface of the real-world 3D mesh model. The fractal dimension is determined by calculating the logarithmic slope of the number of non-empty meshes covering the geographic feature subcluster surface as a function of the mesh scale. During this process, the system uses a preset quantitative classification standard to perform media type identification, when the fractal dimension When the value is greater than 2.65, the geographic entity to which the current geographic feature subcluster belongs is determined to be a loosely packed body with high geometric complexity. The system automatically lowers the risk threshold in the subsequent risk analysis and decision-making module by 15%. When the value is less than or equal to 2.65, the current geographic entity is determined to be a relatively flat bedrock area, and the system maintains the original risk threshold. This procedure establishes the causal relationship between geological medium properties and fractal geometric parameters. To address the deterministic requirement for sampling density adjustment in the dynamic resampling unit, the system introduces a spatial partitioning method based on curvature gradient. This method calculates the normal vector deflection angle of each patch within the geographic feature subcluster, defining the local average curvature K as the average of adjacent normal vector deflection angles. When the spatial gradient of the local average curvature K is greater than 0.15 rad / cm, the dynamic resampling unit increases the spatial index partitioning depth of the current area by one level, increasing the sampling point cloud density of the area from the initial 1500. Increased to 4000 When the spatial gradient of the local mean curvature K is less than or equal to 0.15 rad / cm, the system executes a patch merging procedure to reduce the sampling point cloud density in regions with gentle curvature to 800. .

[0036] Based on the coordinated operation of the above modules, the gravitational field consistency calibration of the attribute response deformation field is performed through the elevation field coupling solution unit, and the projection weight coefficients of each geographic feature subcluster are extracted. Due to the fractal dimension of the loosely packed region The measured value was 2.78, and the local curvature gradient was 0.22 rad / cm. The system automatically performed encrypted sampling and applied the adjusted risk threshold. When soil creep of 1.5 mm occurred, the increased sampling density led to an increase in topological strain entropy. The recognition sensitivity is improved by 25.0%, and the resulting projection weight coefficient Reaching 0.78 triggers the risk status evaluation command. The present invention transforms the complex geographic information processing logic into a closed procedure with clear input, execution logic and output results by establishing a discretized execution program for topological feature extraction, a quantitative classification criterion for fractal dimensions and an adaptive resampling function for curvature gradient. This enables the perception of deformation response characteristics under different geological structural environments, thereby completing the closed-loop analysis of geological disaster risk without using high-cost on-site strain monitoring equipment.

[0037] Example 4: In the deployment scenario of the geological disaster hazard point analysis system, the system uses a pre-deployment calibration procedure to determine the initial background parameters. Five anchor points with geometric stability indices within a preset threshold are selected within the geographic area to be monitored to establish a spatial benchmark. The rigid body transformation parameters between multiple phases of the real-world 3D mesh model are calculated using the nonlinear least squares method, controlling the registration error to be below 1.0 cm. During the initial observation period, the characteristic values ​​of the statistical distribution variance of the mesh side length are extracted to determine the background noise model parameters of the topological strain entropy. ,in, As the topology noise baseline, the system will use it in subsequent calculations. As a background offset subtraction, it is used to eliminate pseudo-entropy change signals caused by fluctuations in model registration accuracy.

[0038] When the topological strain entropy (Hedge) in the monitoring data shows a trend change, the system uses the signal arbitration unit to execute the calculation procedure of spatiotemporal cooperative probability and extract the evolution vector of the attribute response deformation field. displacement vector of the image optical flow field Through calculation and Cosine similarity between them to obtain the collaborative operator The system executes the following decision rule: if the cooperative operator... Greater than 0.72 and projection weight coefficient The angle θ between the evolution direction and the slope gradient is less than 15.0. ∘ If the geographic entity displacement is confirmed, an early warning evaluation command will be output, whereby... For evolution vectors, It is a displacement vector. For cooperative operators, θ is the projection weight coefficient, and θ is the included angle. This method utilizes the correlation of different features in spatial distribution to establish a filtering mechanism, which eliminates topological noise caused by vegetation shaking while ensuring recognition sensitivity.

[0039] Example 5: In the automated deployment procedure of the geological disaster monitoring system, the system executes the procedure for establishing a topological reference field. This involves collecting at least 48 hours of real-world 3D grid model data streams during the stable monitoring period, using a mean filtering algorithm to remove measurement noise, and determining the statistical distribution benchmark for the edge lengths of geographic feature subclusters. These geographic feature subclusters are divided by spatial indexing units based on the adjacency matrix of grid patches. The system calculates the grid edge lengths... The background noise model parameters for determining the topological strain entropy are determined by the steady-state distribution deviation within the reference period. ,in, The length of the grid edge. The system uses the background noise model parameters. topological strain entropy within the observation period Perform bias correction to eliminate periodic mesh artifacts caused by aerial survey lighting cycles.

[0040] When the system is applied to monitoring areas with different image resolutions, the signal arbitration unit executes an adaptive calibration procedure for multimodal spatiotemporal cooperative probabilities. By comparing standard displacement sample points in the historical deformation database, it establishes a causal mapping function between the image optical flow field and the attribute response deformation field. The system then calculates cooperative operators. The sensitivity curve of the deformation displacement Δd is used to determine the consistency threshold for risk assessment. For the cooperative operator, Δd is the deformation displacement. The consistency threshold is set as follows: the cooperative probability is greater than 0.85 under a displacement of 5.0mm and less than 0.15 under background noise fluctuation. The system writes the calibrated parameter matrix into the memory to perform logical filtering on the geographic entity displacement characteristics and the topological noise generated by modeling, so that the triggering of the evaluation command has cross-scene consistency.

[0041] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A smart analysis system for geological disaster hazard point risk zones based on real-scene 3D, characterized in that, include: The data acquisition module is used to acquire a real-world 3D grid model and a digital elevation model of the geospatial area to be monitored. The topology feature extraction module is used to extract the statistical distribution features of the side lengths of the grid cells in the real-world 3D mesh model and generate the topology strain entropy that reflects the geometric evolution law of the real-world 3D mesh model. The risk analysis and decision-making module includes an elevation field coupling solution unit and a signal arbitration unit; The elevation field coupled solution unit is used to retrieve the elevation field gradient vector in the digital elevation model and perform gradient projection operation on the topological strain entropy in the direction of the elevation field gradient vector to construct the attribute response deformation field after eliminating spatial modeling noise. The signal arbitration unit is used to extract the image texture of the image in the real scene 3D mesh model, establish a causal relationship model between geometric deformation and image texture changes, and combine the attribute response deformation field and the quantitative judgment rule of the preset risk threshold to output the evaluation command characterizing the risk status of the geospatial area to be monitored. Among them, the signal arbitration unit constructs the image optical flow field according to the displacement vector of the image texture in the spatial coordinate system, and determines the trigger level of the evaluation command according to the spatiotemporal coordination probability between the characteristic jump frequency of the attribute response deformation field and the physical response of the image optical flow field. Topological strain entropy is characterized by the evolution of the statistical distribution variance of grid side lengths in adjacent time phases. Gradient projection operation extracts the three-dimensional components of the elevation field gradient vector in the geographic coordinate system and maps the evolution value to the motion vector direction that satisfies gravity constraints, so as to eliminate isotropic algorithm artifact noise.

2. The intelligent analysis system for geological disaster hazard point risk areas based on real-scene 3D as described in claim 1, characterized in that, The topology feature extraction module includes a dynamic resampling unit; the dynamic resampling unit adjusts the subdivision depth of the spatial index according to the curvature distribution gradient of the surface of the real-world 3D mesh model; in areas where the curvature change is greater than a preset curvature threshold, the dynamic resampling unit increases the mesh topology connection density; In regions where the curvature change is less than or equal to a preset curvature threshold, the dynamic resampling unit performs topology thinning to merge redundant patches.

3. The intelligent analysis system for geological disaster hazard point risk areas based on real-scene 3D as described in claim 1, characterized in that, When constructing the attribute response deformation field, the elevation field coupled solution unit performs the following anisotropic calibration steps: Step S1, calculate the statistical distribution variance of the mesh side length between two real-world 3D mesh models at different time phases; Step S2, extract the gravity field vector projection at the corresponding location in the digital elevation model; Step S3, calculate the projection weight coefficient to obtain the quantized components of the attribute response deformation field. The calculation formula is: ,in, ΔH is the projection weighting coefficient, ΔH is the evolution value of the statistical distribution variance of the grid side length within the preset observation period, and θ is the angle between the evolution vector of the statistical distribution variance of the grid side length and the projection of the gravity field vector.

4. The intelligent analysis system for geological disaster hazard point risk areas based on real-scene 3D as described in claim 1, characterized in that, When the attribute response deformation field produces a characteristic jump, the signal arbitration unit simultaneously retrieves the image optical flow field to perform image consistency verification; when the displacement characteristics of the image optical flow field and the evolution characteristics of the attribute response deformation field meet the preset spatiotemporal correlation criteria, the signal arbitration unit confirms it as a geographic entity displacement signal.

5. The intelligent analysis system for geological disaster hazard point risk areas based on real-scene 3D as described in claim 1, characterized in that, The risk analysis and decision-making module also includes a media adaptive compensation unit; the media adaptive compensation unit performs fractal dimension analysis on the real-world 3D mesh model and identifies the surface media type based on the results of the fractal dimension analysis; The adaptive compensation unit dynamically modulates the preset risk threshold according to the type of surface medium; for areas identified as loose deposits, the adaptive compensation unit lowers the preset risk threshold by 10% to 20%.

6. The intelligent analysis system for geological disaster hazard point risk areas based on real-scene 3D as described in claim 1, characterized in that, When calculating the topological strain entropy, the topological feature extraction module identifies the topological connectivity of each grid cell in the real-world 3D mesh model and the statistical distribution of grid edge lengths. The topological strain entropy is used to characterize the degree of evolution of the geometric shape of the monitored geographic space area from order to disorder, so as to identify the topological singularity features caused by the displacement of geographic entities.

7. The intelligent analysis system for geological disaster hazard point risk areas based on real-scene 3D as described in claim 1, characterized in that, The data acquisition module also includes a temporal synchronization subunit; the temporal synchronization subunit is used to perform spatial reference alignment and illumination consistency correction on multi-phase real-scene 3D mesh models acquired at different time points, and input the corrected mesh data into the topology feature extraction module.

8. The intelligent analysis system for geological disaster hazard point risk areas based on real-scene 3D as described in claim 1, characterized in that, The risk analysis and decision-making module is also used to calculate the deformation acceleration of the property response deformation field; when the value of the deformation acceleration is continuously greater than 0 mm / s within the preset observation period. 2 At that time, the risk analysis and decision-making module determined that the geospatial area to be monitored had entered an accelerated deformation phase.

9. The intelligent analysis system for geological disaster hazard point risk areas based on real-scene 3D as described in claim 1, characterized in that, The risk analysis and decision-making module also includes a risk zone delineation unit; the risk zone delineation unit identifies geometrically closed-loop regions affected by deformation based on the spatial connectivity analysis of the attribute response deformation field, and maps the geometrically closed-loop regions into independent deformation target feature boundaries in geographic space.

10. The intelligent analysis system for geological disaster hazard point risk areas based on real-scene 3D as described in claim 9, characterized in that, The risk zone delineation unit is also used to perform potential energy analysis based on the terrain slope and elevation difference characteristics within the deformation target feature boundary, and to calculate the potential impact range after the deformation target feature boundary becomes unstable based on the results of the potential energy analysis; the input of the potential energy analysis is dynamically corrected by the real-time topological displacement provided by the attribute response deformation field.

Images