Geographic information visualization intelligent analysis system based on unmanned aerial vehicle surveying and mapping

Abstract

The invention provides a geographic information visualization intelligent analysis system based on unmanned aerial vehicle surveying and mapping, and belongs to the field of surveying and mapping, and the system comprises an unmanned aerial vehicle aerial survey module which carries out regional aerial survey based on a multi-source sensor carried by an unmanned aerial vehicle, obtains multi-modal geographic data, and carries out the preprocessing of the multi-modal geographic data; the data processing module is used for forming an initial geographic data set under a standard coordinate system; the three-dimensional modeling module is used for obtaining the fused geographic feature data, constructing a three-dimensional geographic information model, optimizing the three-dimensional geographic information model and embedding time dimension information to form a four-dimensional spatio-temporal data set; the analysis and decision module is used for generating an analysis report comprising a terrain evolution trend and disaster risk assessment; and the visual platform module is used for visualizing the analysis report through a preset visual interaction interface and adjusting the rendering precision of the optimized three-dimensional geographic information model in real time according to the change of the visual angle of the user. The system provides an efficient and intelligent solution for geographic surveying and mapping and disaster early warning.

Description

Technical Field

[0001] The present invention relates to the field of surveying and mapping technology, and in particular to a geographic information visualization intelligent analysis system based on unmanned aerial vehicle (UAV) surveying and mapping. Background Art

[0002] Geographic Information Systems (GIS) are widely used in surveying and mapping, urban planning, disaster monitoring, and other fields. Combined with drone technology, they enable efficient data collection and analysis. Existing drone mapping systems utilize visible light cameras, infrared sensors, or lidar to capture imagery and point cloud data, generating and visualizing three-dimensional terrain models. Data processing typically relies on ground control point registration, combined with simple feature extraction and visualization tools, to generate static geographic information reports.

[0003] However, existing technologies have the following defects: insufficient multimodal data fusion, resulting in limited model accuracy; lack of time dimension analysis, unable to reflect the dynamic evolution of terrain; poor visualization interactivity, rendering accuracy cannot be dynamically adjusted according to user needs, affecting analysis efficiency and user experience.

[0004] The present invention provides a geographic information visualization intelligent analysis system based on unmanned aerial vehicle surveying and mapping. Summary of the Invention

[0005] The present invention provides a geographic information visualization intelligent analysis system based on drone mapping, which is used to obtain multimodal geographic data through the drone aerial survey module, combined with the spatial alignment of the data processing module to form a high-precision initial geographic data set. The three-dimensional modeling module constructs an optimized four-dimensional spatiotemporal data set through feature extraction and fusion, and embeds the time dimension to improve the dynamic nature of the data. The analysis and decision-making module generates terrain evolution trends and disaster risk assessment reports based on the four-dimensional data set to enhance decision-making support capabilities. The visualization platform module realizes the dynamic presentation of analysis results through an interactive interface, and optimizes the rendering accuracy in real time according to the user's perspective, significantly improving the accuracy, real-time performance and interactive experience of geographic information analysis, and providing an efficient and intelligent solution for geographic mapping and disaster warning.

[0006] The present invention provides a geographic information visualization intelligent analysis system based on UAV surveying and mapping, comprising:

[0007] UAV aerial survey module: Based on the multi-source sensors carried by UAVs, regional aerial surveys are carried out to obtain multimodal geographic data including visible light images, infrared images, and laser point clouds, and perform pre-processing;

[0008] Data processing module: spatially aligns the pre-processed multimodal geographic data with the ground control points in the preset ground control point database to form an initial geographic dataset in a standard coordinate system;

[0009] 3D modeling module: Extracts and fuses features from the initial geographic dataset to obtain fused geographic feature data, constructs a 3D geographic information model based on the fused geographic feature data, optimizes the 3D geographic information model, and embeds time dimension information into the optimized 3D geographic information model to form a 4D spatiotemporal dataset.

[0010] Analysis and decision-making module: Conducts intelligent analysis based on four-dimensional spatiotemporal data sets to generate analysis reports including terrain evolution trends and disaster risk assessments;

[0011] Visualization platform module: visualize the analysis report through a preset visual interactive interface, and adjust the rendering accuracy of the optimized three-dimensional geographic information model in real time according to changes in user perspective.

[0012] Preferably, the drone aerial survey module includes:

[0013] Route planning unit: obtains boundary information and terrain characteristic parameters of the target area, and plans the optimal flight route according to the preset surveying and mapping accuracy requirements;

[0014] Data acquisition unit: controls the drone equipped with multi-source sensors to fly along the optimal flight route, and simultaneously collects multimodal geographic data of the target area, including visible light images, infrared images, and laser point cloud data;

[0015] Flight Adjustment Unit: During the flight acquisition process, the unit dynamically adjusts the drone's flight altitude, shooting interval, and scanning frequency by comparing preset aerial survey parameters with the characteristics of actual collected data. It also evaluates the integrity of data collected by multi-source sensors and triggers a re-flight mechanism when the data loss rate exceeds a preset threshold.

[0016] Data preprocessing unit: The acquired multimodal geographic data including visible light images, infrared images, and laser point cloud data are matched with the POS data recorded by the UAV positioning and attitude determination system in time and space to obtain the preprocessed multimodal geographic data.

[0017] Preferably, the data processing module includes:

[0018] Control point extraction unit: extracts a set of control points that match the target area from a preset ground control point database, including coordinate information and feature descriptions of plane control points and elevation control points;

[0019] Spatial registration unit: establishes feature matching relationships between pre-processed multimodal geographic data and control points in the control point set and performs spatial registration;

[0020] Registration optimization unit: gradually optimizes the spatial registration accuracy and re-performs spatial registration when the registration deviation exceeds a preset threshold;

[0021] Registration Verification Unit: Build a quality control indicator system to evaluate the integrity, consistency, and accuracy of the registered initial geographic dataset to generate a registration reliability index;

[0022] Result output unit: When the registration reliability index meets the preset result, the registered initial geographic dataset is output based on the spatial registration result.

[0023] Preferably, the three-dimensional modeling module includes:

[0024] Feature extraction unit: performs multi-scale feature analysis on the initial geographic data set to extract structured information including terrain contours, feature boundaries, and texture features;

[0025] Feature analysis unit: performs similarity analysis on structured information, aligns and fuses data from different sensors based on similarity, and constructs a fused geographic feature dataset with geometric accuracy and texture authenticity;

[0026] Model conversion unit: uses triangulation algorithm to convert fused data into three-dimensional geographic information model;

[0027] Model optimization unit: optimize the three-dimensional geographic information model;

[0028] Time embedding unit: establishes a time coding mechanism on the three-dimensional geographic information model, associates historical change data with the three-dimensional geographic information model, and generates a four-dimensional spatiotemporal dataset with time dimension information.

[0029] Preferably, the model optimization unit includes:

[0030] Feature separation subunit: performs multi-dimensional feature separation based on the initial geographic data set, decouples the data features obtained by different sensors, and separates geometric features, spectral features, and temporal features;

[0031] Structural optimization subunit: performs feature reorganization based on the multi-dimensional feature separation results, adjusts the feature fusion weight according to the type of ground feature, and generates an optimized structural model;

[0032] Terrain classification subunit: Classify terrain complexity based on optimized structural model and time dimension information;

[0033] Grid refinement subunit: implements dynamic grid subdivision based on terrain complexity classification results;

[0034] Density optimization subunit: obtains the user's viewing distance and optimizes the mesh density after dynamic mesh segmentation based on the user's viewing distance.

[0035] Preferably, the terrain complexity classification result includes dividing the target area into a flat area, a transition area and a complex area.

[0036] Preferably, the analysis and decision-making module includes:

[0037] Feature extraction unit: This unit divides the four-dimensional spatiotemporal data set into a time-domain sliding window and extracts the geometric deformation, spectral reflectance change rate, and point cloud density fluctuation characteristics of each feature area at different time points based on the preset feature classification rules. The extracted original features are then normalized and outliers are filtered out to output a standardized feature matrix with spatiotemporal continuity.

[0038] Region identification unit: Based on the standardized feature matrix, it sets the classification thresholds of flat areas, transition areas, and complex areas according to the geometric deformation and spectral characteristics, and identifies several abnormal areas in combination with the time dimension information;

[0039] Historical correlation unit: extracts historical meteorological data and geological activity records from a preset historical database for each abnormal area, determines the corresponding abnormal cause, and then outputs the terrain evolution results, including the evolution type, spatial range, time process and a list of correlation factors;

[0040] Map generation unit: performs multi-index risk assessment based on terrain evolution results and a preset disaster rule library to generate a risk zoning map;

[0041] Report generation unit: Integrate terrain evolution results with risk zoning maps, and generate analysis reports including terrain evolution trends and disaster risk assessments according to preset templates.

[0042] Preferably, the visualization platform module includes:

[0043] Parameter conversion unit: Based on the preset visualization rule library, the analysis report is converted into unified visualization parameters;

[0044] Hierarchical division unit: Based on the terrain complexity classification results and time dimension information, the 3D geographic information model is divided into several LOD levels to generate a hierarchical management list;

[0045] Parameter determination unit: Based on the embedded posture sensor and interaction event listener, it captures the three-dimensional coordinates, pitch angle and zoom ratio parameters of the user's perspective in real time and determines the standardized perspective parameters;

[0046] View annotation unit: Synchronizes the standardized viewing angle parameters with the hierarchical management list, and annotates the complexity of the target area within the user's current view;

[0047] Instruction generation unit: calls a preset rendering strategy rule library and the complexity of the target area in the user's current field of view to determine the priority of the target area in the user's current field of view, and generates a rendering instruction queue based on the priority of the target area in the user's current field of view;

[0048] Visualization unit: According to the rendering command queue, model blocks and texture maps of different precisions are loaded frame by frame. Asynchronous streaming technology is used to prioritize loading high-precision data in the center of the field of view, and delayed filling of edge areas. After each frame is rendered, the frame rate and video memory usage are counted through the preset performance monitoring tool. If the frame rate is detected to fall below the preset frame rate threshold, the shadow quality and particle special effect level of the non-focused area in the user's current field of view are reduced, and the final output rendering picture is displayed on the preset visual interactive interface.

[0049] Compared with the prior art, the present invention has the following advantages:

[0050] The drone-based aerial survey module acquires multimodal geographic data, combined with spatial registration in the data processing module to form a high-precision initial geographic dataset. The 3D modeling module extracts and fuses features to construct an optimized 4D spatiotemporal dataset, embedding the time dimension to enhance data dynamics. The analysis and decision-making module generates terrain evolution trends and disaster risk assessment reports based on the 4D dataset, enhancing decision-making support capabilities. The visualization platform module dynamically presents analysis results through an interactive interface and optimizes rendering accuracy in real time based on the user's perspective. This significantly improves the accuracy, real-time nature, and interactive experience of geographic information analysis, providing an efficient and intelligent solution for geographic mapping and disaster early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 It is a structural diagram of a geographic information visualization intelligent analysis system based on drone mapping provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0053] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0054] Example 1:

[0055] The embodiment of the present invention provides a geographic information visualization intelligent analysis system based on drone mapping, such as Figure 1 Shown, including:

[0056] UAV aerial survey module: Based on the multi-source sensors carried by UAVs, regional aerial surveys are carried out to obtain multimodal geographic data including visible light images, infrared images, and laser point clouds, and perform pre-processing;

[0057] Data processing module: spatially aligns the pre-processed multimodal geographic data with the ground control points in the preset ground control point database to form an initial geographic dataset in a standard coordinate system;

[0058] 3D modeling module: Extracts and fuses features from the initial geographic dataset to obtain fused geographic feature data, constructs a 3D geographic information model based on the fused geographic feature data, optimizes the 3D geographic information model, and embeds time dimension information into the optimized 3D geographic information model to form a 4D spatiotemporal dataset.

[0059] Analysis and decision-making module: Conducts intelligent analysis based on four-dimensional spatiotemporal data sets to generate analysis reports including terrain evolution trends and disaster risk assessments;

[0060] Visualization platform module: visualize the analysis report through a preset visual interactive interface, and adjust the rendering accuracy of the optimized three-dimensional geographic information model in real time according to changes in user perspective.

[0061] In this embodiment, a drone equipped with multi-source sensors refers to the integration of detection equipment based on different physical principles, such as optical, infrared, and lidar, in a single aerial survey mission. Multi-modal collaborative observation overcomes the limitations of a single sensor. Visible light images provide high-resolution texture details to support object classification, infrared images identify water seepage areas or underground pipeline leakage points through thermal radiation differences, and laser point clouds directly obtain the three-dimensional geometric structure of the surface. For example, in the subsidence monitoring of a certain mining area, the drone simultaneously obtains RGB images (to identify new cracks), thermal infrared images (to delineate low-temperature anomaly areas caused by water seepage), and LiDAR point clouds (to calculate millimeter-level elevation changes). The three are collected under completely synchronized conditions in time and space to ensure strict alignment of the data during subsequent fusion analysis.

[0062] In this embodiment, the pre-set ground control point database is a long-established benchmark coordinate reference system, containing landmarks with known geodetic coordinates (such as manually placed targets and permanent building corners) and their precise spatial location information. By matching the aerial survey data with these control points, systematic deviations such as drone positioning errors and lens distortion can be eliminated, achieving a unified spatial reference for multiple data periods.

[0063] In this embodiment, the preset visual interactive interface adopts multi-level detail (LOD) technology and ergonomic design principles to achieve dynamic adaptive presentation of analysis results and three-dimensional models. The basic interface includes a timeline control, a risk level legend, and a profile analysis tool. When the user zooms or switches the perspective through gestures, the system automatically adjusts the model rendering accuracy according to the viewing distance - the long-range view displays a simplified terrain outline, and the close-up view loads high-precision textures and deformation vector arrows. For example, in a geological disaster emergency command scenario, when the commander looks at the large screen, the interface displays a risk heat map of the entire area. When he focuses on a landslide and rotates the perspective to a bird's-eye view, the system immediately loads the crack distribution vector layer and displacement animation of the area. At the same time, the associated meteorological data curve pops up in the sidebar to form a multi-dimensional collaborative visualization solution.

[0064] The beneficial effects of the above technical solution are as follows: The drone aerial survey module acquires multimodal geographic data, which, combined with spatial registration in the data processing module, forms a high-precision initial geographic dataset. The 3D modeling module extracts and fuses features to construct an optimized 4D spatiotemporal dataset, embedding the time dimension to enhance data dynamics. The analysis and decision-making module generates terrain evolution trends and disaster risk assessment reports based on the 4D dataset, enhancing decision-making support capabilities. The visualization platform module dynamically presents analysis results through an interactive interface and optimizes rendering accuracy in real time based on the user's perspective, significantly improving the accuracy, real-time nature, and interactive experience of geographic information analysis, providing an efficient and intelligent solution for geographic mapping and disaster early warning.

[0065] Example 2:

[0066] The embodiment of the present invention provides a geographic information visualization intelligent analysis system based on drone mapping, and a drone aerial survey module, including:

[0067] Route planning unit: obtains boundary information and terrain characteristic parameters of the target area, and plans the optimal flight route according to the preset surveying and mapping accuracy requirements;

[0068] Data acquisition unit: controls the drone equipped with multi-source sensors to fly along the optimal flight route, and simultaneously collects multimodal geographic data of the target area, including visible light images, infrared images, and laser point cloud data;

[0069] Flight Adjustment Unit: During the flight acquisition process, the unit dynamically adjusts the drone's flight altitude, shooting interval, and scanning frequency by comparing preset aerial survey parameters with the characteristics of actual collected data. It also evaluates the integrity of data collected by multi-source sensors and triggers a re-flight mechanism when the data loss rate exceeds a preset threshold.

[0070] Data preprocessing unit: The acquired multimodal geographic data including visible light images, infrared images, and laser point cloud data are matched with the POS data recorded by the UAV positioning and attitude determination system in time and space to obtain the preprocessed multimodal geographic data.

[0071] In this embodiment, obtaining the boundary information and terrain characteristic parameters of the target area is a basic prerequisite for route planning. The boundary information not only includes the geographic coordinate range, but also needs to consider constraints such as airspace control or no-fly zones. The terrain characteristic parameters include key indicators such as average elevation, maximum height difference, and slope distribution, which directly affect the flight safety of the drone and the efficiency of sensor coverage.

[0072] In this embodiment, the optimal flight route is planned based on preset mapping accuracy requirements. This involves analyzing the target area boundaries (such as administrative boundaries or project scopes) and terrain characteristic parameters (such as average elevation and slope gradient), and then generating a route using an adaptive flight strip design algorithm based on the resolution and overlap ratio specified in the mapping accuracy requirements. The algorithm automatically adjusts flight strip spacing based on terrain undulations—increasing the spacing in flat areas to improve efficiency and reducing it in steep areas to avoid data leaks. The algorithm also dynamically calculates flight altitude based on the sensor's field of view and terrain elevation differences to ensure that all locations meet accuracy requirements. For example, during the mapping of a hydropower station reservoir, the system identified a steep cliff on the left bank (slope > 45°) and a gentle slope on the right bank (slope < 10°). It then automatically generated differentiated flight strips: a dense parallel flight path with oblique photography was used in the steep cliff area, while a conventional flight path with increased flight altitude was used in the gentle slope area. This ultimately reduced flight time by 40% while maintaining millimeter-level identification of cliff cracks. In this embodiment, the optimal flight route is the global optimal solution that comprehensively considers energy consumption, data quality, and operation time. It must meet three core constraints: first, effective sensor coverage to ensure that image overlap and point cloud density meet standards; second, avoidance of terrain obstacles, using three-dimensional route design to prevent the risk of collision with mountains or loss of control due to updrafts; and third, mission feasibility, including battery life and emergency plans for communication interruptions. For example, in a certain island reef mapping project, the system simulated and calculated real-time sea conditions to design a circular route with headwind takeoff and landing and downwind mapping. This not only utilized sea breezes to increase the drone's hovering time, but also used a tidal compensation algorithm to eliminate the impact of waves on laser ranging. The route also reserves an emergency return point. When the wind speed suddenly changes, the current route can be immediately interrupted and the nearest return can be made, demonstrating the intelligent system's ability to adapt to complex environments.

[0073] In this embodiment, the preset aerial survey parameters constitute the quality control benchmark for data acquisition, covering two major dimensions: geometric accuracy and spectral fidelity. Geometric parameters include heading / side overlap rate, scanning line interval, etc., to ensure that there are no blind spots in three-dimensional reconstruction; spectral parameters involve white balance calibration, infrared radiation calibration, etc., to ensure the reliability of physical interpretation of multi-source data. These parameters are not fixed values, but are dynamically adjusted according to the scene type - for example, urban construction areas require the overlap rate to be increased to 80% to deal with occlusion by high-rise buildings, while vegetation monitoring requires enhanced near-infrared band calibration. Taking the leakage monitoring of a chemical park as an example, the system loads a special parameter template for hazardous chemical facilities: visible light imaging uses high dynamic range (HDR) mode to capture pipeline corrosion details, the infrared sensor is set to a range of -20℃ to 300℃ to cover the possible leakage temperature range, and the lidar uses multi-echo mode to penetrate smoke interference. When the actual data characteristics deviate from the preset, real-time closed-loop control is triggered;

[0074] In this embodiment, the dynamic adjustment of the drone's flight altitude, shooting interval, and scanning frequency includes: comparing the characteristics of the collected data with preset aerial survey parameters in real time, and adjusting the flight parameters through a PID control algorithm. For example, when strong backlight is detected, causing overexposure of the visible light image, the flight altitude is automatically lowered and the exposure time is shortened; when the LiDAR scan encounters dense vegetation, causing multiple surges in the echo rate, the scanning frequency is dynamically increased to obtain a sufficient valid point cloud. The data integrity assessment module uses a sliding window statistical method to calculate the proportion of valid data pixels within the current flight band in real time. If a large area of infrared data is invalidated due to sudden electromagnetic interference, the system immediately marks the coordinates of the abnormal area and plans a re-flight path. In a mining inspection case, the drone encountered a sudden dust storm during flight, causing the clarity of the visible light image to drop sharply. The system immediately activated the LiDAR active re-scan mode and adjusted the route to avoid the dust-dense area. Afterwards, re-flight missions were automatically generated for areas that did not meet the standards to ensure that no exploration data was missed.

[0075] In this embodiment, the multimodal geographic data obtained, including visible light images, infrared images, and laser point cloud data, are temporally and spatially matched with the POS data recorded by the drone positioning and attitude system to obtain pre-processed multimodal geographic data. This includes: using the millisecond-accurate timestamps recorded by the POS data (position and attitude system) and the IMU inertial measurement data to construct a time series function of the sensor's exterior orientation elements, and correcting the spatial position of each frame of image / each laser pulse to a unified reference system through a spatiotemporal interpolation algorithm. To address the problem of heterogeneous sampling of multi-source data (such as the difference between the LiDAR scanning frequency and the camera frame rate), a sliding time window alignment strategy is adopted to ensure that the spatial representation of the same feature in different modal data is consistent. For example, in a highway deformation monitoring project, the drone was affected by crosswinds and produced high-frequency jitter, resulting in misalignment between the visible light image and the LiDAR point cloud. The six-degree-of-freedom motion model of the flight trajectory was reconstructed using the POS data, and the image was compensated by affine transformation. Ultimately, the position deviation of the asphalt pavement cracks in the image and point cloud was controlled to the sub-pixel level, meeting the subsequent millimeter-level deformation analysis requirements.

[0076] The beneficial effects of this technical solution are as follows: The drone aerial survey module optimizes the flight path based on the characteristics of the target area through the route planning unit, ensuring surveying and mapping accuracy. The data acquisition unit simultaneously acquires multimodal geographic data such as visible light, infrared, and laser point clouds, enhancing data diversity. The flight adjustment unit dynamically optimizes flight parameters, assesses data integrity in real time, and triggers a re-flight mechanism to reduce data loss rates. The data preprocessing unit integrates multimodal data and POS information through spatiotemporal matching to form a high-quality preprocessed dataset, providing a precise foundation for subsequent modeling and analysis. The overall solution significantly improves the efficiency, data integrity, and adaptability of drone surveying and mapping, providing reliable support for geographic information visualization and intelligent analysis.

[0077] Example 3:

[0078] The embodiment of the present invention provides a geographic information visualization intelligent analysis system based on drone mapping, and the data processing module includes:

[0079] Control point extraction unit: extracts a set of control points that match the target area from a preset ground control point database, including coordinate information and feature descriptions of plane control points and elevation control points;

[0080] Spatial registration unit: establishes feature matching relationships between pre-processed multimodal geographic data and control points in the control point set and performs spatial registration;

[0081] Registration optimization unit: gradually optimizes the spatial registration accuracy and re-performs spatial registration when the registration deviation exceeds a preset threshold;

[0082] Registration Verification Unit: Build a quality control indicator system to evaluate the integrity, consistency, and accuracy of the registered initial geographic dataset to generate a registration reliability index;

[0083] Result output unit: When the registration reliability index meets the preset result, the registered initial geographic dataset is output based on the spatial registration result.

[0084] In this embodiment, a set of control points that matches the target area is extracted from a preset ground control point database, including coordinate information and feature descriptions of plane control points and elevation control points, including: intelligent screening based on the spatial coverage requirements and feature stability of the target area, giving priority to long-term unchanged obvious features as control points. Plane control points usually include artificial signs such as road marking intersections and building corners, while elevation control points are selected from terrain feature points (such as ridges and riverbed turning points). The system not only matches spatial coordinates, but also verifies the recognition of control points in multi-temporal images. For example, when surveying in mountainous areas, it automatically avoids points that are easily affected by snow cover, and gives priority to using exposed rock formations as markers. For urban renewal areas, the database will integrate historical surveying and mapping data to avoid selecting buildings to be demolished as benchmark points.

[0085] In this embodiment, a feature matching relationship is established between the preprocessed multimodal geographic data and the control point set, and spatial registration is performed. This includes: using the SIFT algorithm to match the texture of the control points in the drone's visible light imagery, extracting building edge features through plane fitting in the laser point cloud, and assisting in verifying the reliability of the match using infrared data. For example, in bridge monitoring, the system aligns the bridge tower image acquired through oblique photography with the coordinates of the control points measured by the total station. The registration parameters are then fine-tuned using the geometric features of the cable point cloud to ensure consistency between the deformation analysis data and the geodetic coordinate system. The registration process must address the conversion issues between different sensor coordinate systems and automatically remove invalid control points that are temporarily obscured or contaminated.

[0086] In this embodiment, the spatial registration accuracy is gradually optimized and the re-registration mechanism is triggered, including: improving the registration accuracy through iterative calculation, using coarse matching to obtain the initial conversion parameters in the first round, and subsequently using algorithms such as ICP to optimize local errors. When the residual in a certain area exceeds the threshold, the backup control points are automatically switched or the registration strategy is adjusted. For example, in mine monitoring, some control points are displaced due to blasting vibrations. After the system detects the abnormal deviation, it immediately calls the surrounding stable bedrock points for recalculation to control the error within the allowable range of the project. The optimization process also introduces terrain adaptive weights to strengthen the elevation accuracy constraints in steep slope areas, while giving priority to plane accuracy in flat areas.

[0087] In this embodiment, a three-dimensional quality assessment is performed on the registered geographic dataset, including generating a reliability index through a triple check of completeness, consistency, and accuracy. Integrity analysis detects data holes (such as areas obscured by clouds), consistency verifies the logical relationship of multi-source data (such as the matching degree between point cloud building height and image shadow), and accuracy is verified through back-substitution of reserved control points. In urban subway monitoring, the system found that the tunnel entrance point cloud was missing (lack of completeness), but the registered section was consistent with the gyroscope measurement results, so it output a partitioned and graded reliability report to guide subsequent re-measurements.

[0088] This embodiment outputs standardized geographic datasets based on reliability indicators. After passing quality verification, the system intelligently generates a results package tailored to the application scenario. For example, in an oil pipeline inspection task, the output data includes a global model in the WGS84 coordinate system, a local precision model in the construction coordinate system, and a lightweight AR inspection model. Each data set is accompanied by metadata documenting the control points used, accuracy range, and applicable scenarios. Key areas are also marked with confidence heat maps. For low-reliability data, a warning watermark is automatically added and a list of recommended retests is generated to ensure risk management for downstream applications.

[0089] The beneficial effects of the above technical solution are as follows: the data processing module accurately obtains the plane and elevation control points of the target area through the control point extraction unit to ensure the accuracy of the basic registration data. The spatial registration unit realizes the precise alignment of multimodal geographic data and control points through feature matching. The registration optimization unit iteratively optimizes the registration accuracy, dynamically corrects deviations, and improves data quality. The registration verification unit builds a comprehensive quality control indicator system to evaluate the integrity, consistency and accuracy of the data set to ensure the reliability of the registration. The result output unit outputs a standardized initial geographic data set based on the high-reliability registration results, providing high-precision and high-consistency data support for subsequent three-dimensional modeling and intelligent analysis, significantly improving the surveying and mapping efficiency and analysis reliability of the geographic information system.

[0090] Example 4:

[0091] The embodiment of the present invention provides a geographic information visualization intelligent analysis system based on drone mapping, and a three-dimensional modeling module, including:

[0092] Feature extraction unit: performs multi-scale feature analysis on the initial geographic data set to extract structured information including terrain contours, feature boundaries, and texture features;

[0093] Feature analysis unit: performs similarity analysis on structured information, aligns and fuses data from different sensors based on similarity, and constructs a fused geographic feature dataset with geometric accuracy and texture authenticity;

[0094] Model conversion unit: uses triangulation algorithm to convert fused data into three-dimensional geographic information model;

[0095] Model optimization unit: optimize the three-dimensional geographic information model;

[0096] Time embedding unit: establishes a time coding mechanism on the three-dimensional geographic information model, associates historical change data with the three-dimensional geographic information model, and generates a four-dimensional spatiotemporal dataset with time dimension information.

[0097] In this embodiment, a multi-scale feature analysis is performed on the initial geographic data set, including: extracting structured features of different levels from the original data through a multi-scale algorithm to adapt to various spatial analysis needs. In terrain contour extraction, the system uses a Gaussian difference pyramid to identify macro-geomorphic features (such as ridge lines and watershed boundaries), and combines local binary patterns to analyze micro-topography undulations; the boundary detection of land objects integrates the Canny operator and the deep learning edge detection network to accurately segment the building outline and road markings; the texture feature analysis quantifies the spatial frequency characteristics of the surface cover through a Gabor filter group. For example, in a wetland protection project, this method successfully separated the reed area (high texture complexity) from the open water surface (low texture response), providing a basis for ecological zoning. The multi-scale strategy can also avoid feature misjudgment at a single scale, such as avoiding mistaking a small area of bare soil for a road gap.

[0098] In this embodiment, multi-source data registration and fusion based on similarity is to achieve spatial alignment of multimodal data such as laser point clouds and optical images through feature similarity calculation. The system constructs a hybrid feature descriptor to uniformly map the geometric accuracy of the point cloud (such as the curvature of the building roof line) and the radiation characteristics of the image (such as the spectral reflectance of the asphalt pavement) to a common metric space. The improved RANSAC algorithm is used for registration, and feature pairs with high consistency across sensors (such as the vertical edge of a street light pole and a point cloud cylinder) are matched first. In a case of ancient building restoration, millimeter-level fusion was achieved by comparing the brick joint texture of the drone image with the concave and convex features of the masonry from the ground laser scan, reproducing the weathered and worn carving details while retaining the original color information.

[0099] In this embodiment, the triangulation algorithm is used to convert the fused data into a three-dimensional geographic information model. The constrained Delaunay triangulation algorithm is used in the conversion process to optimize the model efficiency while maintaining the topological integrity of feature lines (such as road centerlines). The system automatically identifies the feature hierarchy and dynamically adjusts the density of triangles: large-size triangles are used for regular structures such as building facades, while detailed areas such as relief decorations are locally encrypted. A subway station modeling project used this method to streamline 3 billion original point clouds into a lightweight model of 8 million faces, completely retaining the steel structure node information of the station hall dome, while ensuring that the BIM software can be loaded smoothly. The algorithm also supports multi-precision LOD generation to meet different needs from overall planning to construction details.

[0100] In this embodiment, generating a four-dimensional spatiotemporal dataset with time dimension information includes: the time dimension is implemented through a version control tree, and each node is associated with a three-dimensional model of a specific temporal state and change metadata (such as construction stage, natural evolution). The system uses a linear reference system to interpolate and model continuously changing elements (such as river erosion lines), and discrete changes (such as building demolition) are recorded with event snapshots. In urban expansion analysis, this method connects 20 years of satellite imagery, cadastral data and current real-life models in series, which can dynamically simulate the evolution of a development zone from farmland to CBD, and quantify the changes in volume ratio in each period. The timeline also supports reverse tracing, such as locating historical construction records of a leakage point in a pipeline network.

[0101] The beneficial effects of the above technical solution are as follows: the 3D modeling module uses the feature extraction unit to perform multi-scale analysis, accurately extracting terrain contours, feature boundaries, and texture features, ensuring the comprehensiveness of structured information. The feature analysis unit uses similarity analysis to achieve efficient registration and fusion of multi-source sensor data, generating a fused geographic feature dataset with precise geometry and realistic textures. The model conversion unit uses a triangulation algorithm to construct a high-precision 3D geographic information model, and the model optimization unit further improves model quality. The time embedding unit incorporates historical change data into the model through a time encoding mechanism, forming a four-dimensional spatiotemporal dataset, significantly enhancing the ability to analyze terrain dynamics, providing high-precision, dynamic data support for disaster assessment and planning, and improving the analytical efficiency and application value of geographic information systems.

[0102] Example 5:

[0103] The embodiment of the present invention provides a geographic information visualization intelligent analysis system based on drone mapping, and a model optimization unit, including:

[0104] Feature separation subunit: performs multi-dimensional feature separation based on the initial geographic data set, decouples the data features obtained by different sensors, and separates geometric features, spectral features, and temporal features;

[0105] Structural optimization subunit: performs feature reorganization based on the multi-dimensional feature separation results, adjusts the feature fusion weight according to the type of ground feature, and generates an optimized structural model;

[0106] Terrain classification subunit: Classify terrain complexity based on optimized structural model and time dimension information;

[0107] Grid refinement subunit: implements dynamic grid subdivision based on terrain complexity classification results;

[0108] Density optimization subunit: obtains the user's viewing distance and optimizes the mesh density after dynamic mesh segmentation based on the user's viewing distance.

[0109] In this embodiment, multimodal features with physical meaning are separated from the initial geographic data set: geometric (G), spectral (S), and temporal (T). Method: The original data is mapped into three types of feature vectors using feature extraction operators and sensor calibration matrices:

[0110]

[0111] Among them, Φ g (D), Φ s (D), Φ t (D) are geometric, spectral, and temporal feature extraction operators (such as SIFT, NDVI, and temporal difference); M geom 、M spec and M temp The characteristic weight matrix of sensor calibration is a dimensionless weight matrix used to normalize sensor response;

[0112] In this embodiment, the feature combination weight is dynamically adjusted according to the feature type L to generate the optimized structural model value E. The combination method is:

[0113] And ∑w=1

[0114] Among them, G i is the geometric characteristic value (the unit depends on the specific index, such as curvature, volume, S i is the spectral characteristic value (such as vegetation index such as NDVI); T i is the texture eigenvalue (such as the contrast of the gray-level co-occurrence matrix). The above eigenvalues are standardized to eliminate the influence of dimension and unit. Building area: mainly based on geometry, w G,i =0.6, w S,i =0.2, w T,i =0.2; Vegetation area: mainly based on the spectrum, w G,i =0.3, w S,i =0.6, w T,i =0.1;

[0115] In this embodiment, terrain complexity classification is performed based on the optimized structural model and time dimension information, including: by calculating the local curvature (h) and the coefficient of variation of elevation (i), the terrain is divided into a flat area (curvature < 0.1, coefficient of variation < 5%), a transition area (0.1 ≤ curvature < 0.3, 5% ≤ coefficient of variation < 15%), and a complex area (curvature ≥ 0.3, coefficient of variation ≥ 15%); the local curvature C:

[0116]

[0117] Among them, z i is the elevation of the adjacent vertex (unit: meter), z0 is the elevation of the central vertex; di is the horizontal distance from the neighborhood point to the center (unit: meter), n is the number of neighborhood points (usually 6-8);

[0118] Elevation variation coefficient V:

[0119]

[0120] Among them, σ z is the local elevation standard deviation, μ z is the local mean elevation.

[0121] In this embodiment, dynamic mesh subdivision is implemented based on the terrain complexity classification results, including: flat areas: using sparse triangular meshes (side length ≥ 5 meters) to reduce rendering load; transition areas: mixed quadrilateral-triangular meshes (side length 2-5 meters) to balance accuracy and performance; complex areas: high-density adaptive meshes (side length ≤ 1 meter), and using edge sharpening technology to enhance features such as fracture lines and steep slopes.

[0122] In this embodiment, the user viewing distance is obtained, and the density of the mesh after dynamic mesh segmentation is optimized based on the user viewing distance, including: adjusting the final rendering mesh density in real time according to the user viewing distance d:

[0123]

[0124] Among them, Δ f inal is the final rendered mesh density, d ref It is a reference viewing distance (such as 1000m) and indicates the normal browsing distance. For example, when observing a complex area at close range (200m), the grid is encrypted to 0.5m. When browsing at a long distance (>1000m), the grid automatically falls back to the original density.

[0125] The beneficial effects of this technical solution are as follows: the model optimization unit decouples multidimensional features through the feature separation subunit, separating geometric, spectral, and temporal features to improve data processing accuracy. The structure optimization subunit adjusts fusion weights based on feature type to generate a high-quality optimized structural model. The terrain classification subunit combines temporal information to perform complexity classification, providing a basis for detailed modeling. The grid refinement subunit optimizes the grid structure through dynamic subdivision, and the density optimization subunit dynamically adjusts the grid density based on the user's viewing distance, significantly improving the rendering efficiency and visual realism of the 3D model. The overall solution enhances the model's adaptability and dynamic optimization capabilities, providing efficient and accurate support for geographic information visualization.

[0126] Example 6:

[0127] The embodiment of the present invention provides a geographic information visualization intelligent analysis system based on UAV mapping, and the terrain complexity classification result includes: dividing the target area into: flat area, transition area and complex area.

[0128] In this embodiment, a gentle area refers to an area where the surface morphology changes more regularly and has simple geometric features. Such areas usually exhibit low undulations and stable spectral reflectance characteristics, such as plains, farmlands, grasslands or open waters. In the model construction of the gentle area, larger grid units can be used, and the texture detail requirements are lower, because the structure of the ground objects is simple and the deformation pattern is easy to predict. For example, in the drone mapping of a certain agricultural area, a large area of wheat planting was identified as a gentle area - the terrain elevation difference is very small, and multiple images show regular spectral changes during the crop growth cycle (such as the gradual change of near-infrared reflectivity from the green stage to the mature stage). In terms of geometry, only key lines such as the boundaries of the ridges need to be retained. After the grid is simplified, the computational load can be significantly reduced.

[0129] In this embodiment, the transition zone refers to an area between a gentle area and a complex area, with a moderate degree of terrain variability and mixed landform features. Such areas may include gentle slopes, sparse buildings, or foothills with uneven vegetation coverage. Their geometric structure and spectral characteristics have local fluctuations but are not extremely complex. For example, in a mapping scene at the junction of an urban and suburban area, the area where scattered houses and orchards intersect is marked as a transition zone - the roofs of the houses have regular geometric shapes but are loosely distributed, and the height differences of the orchard trees lead to uneven point cloud density. The system assigns medium grid accuracy to such areas and balances the sharpness of the building edges with the natural transition effect of the vegetation when fusion of features, ensuring visual continuity while avoiding over-segmentation.

[0130] In this embodiment, a complex area refers to an area where the surface morphology changes dramatically and the types of land features are highly mixed. Typical scenarios include urban central areas, steep mountains, geological disaster risk points, etc., which are characterized by broken geometric structures (such as the height difference of dense buildings), heterogeneous spectral characteristics (such as the large difference in reflectivity between glass curtain walls and asphalt pavement), and significant time-varying characteristics (such as the continuous displacement of landslide bodies). For example, in the drone mapping of a mountainous city, the collapsed rock belts on both sides of the canyon were identified as complex areas - the development of rock cracks resulted in a large number of voids in the point cloud data. Multi-period monitoring showed that the monthly displacement of local rock blocks varied by tens of centimeters. The system needs to adopt an adaptive subdivision grid: sub-meter-level fine-grained segmentation of the crack edges to capture the rupture trend, and appropriately reduce the resolution of stable rock areas to optimize storage efficiency.

[0131] The beneficial effects of the above technical solution are as follows: the terrain complexity classification results divide the target area into flat areas, transition areas, and complex areas, achieving accurate classification based on the differences in terrain characteristics, and providing a scientific basis for subsequent mesh refinement and model optimization. This classification method fully considers the complexity and spatiotemporal changes of terrain, can effectively guide dynamic segmentation and rendering optimization, and improve the accuracy and efficiency of building three-dimensional geographic information models. The classification results support adaptive resource allocation, reduce the computational load in flat areas, and focus on optimizing the detailed expression of complex areas, significantly improving visualization effects and analysis accuracy, providing high-quality data support for disaster risk assessment and terrain evolution analysis, and enhancing the practicality and adaptability of the system.

[0132] Example 7:

[0133] The embodiment of the present invention provides a geographic information visualization intelligent analysis system based on drone mapping, and the analysis and decision-making module includes:

[0134] Feature extraction unit: This unit divides the four-dimensional spatiotemporal data set into a time-domain sliding window and extracts the geometric deformation, spectral reflectance change rate, and point cloud density fluctuation characteristics of each feature area at different time points based on the preset feature classification rules. The extracted original features are then normalized and outliers are filtered out to output a standardized feature matrix with spatiotemporal continuity.

[0135] Region identification unit: Based on the standardized feature matrix, it sets the classification thresholds of flat areas, transition areas, and complex areas according to the geometric deformation and spectral characteristics, and identifies several abnormal areas in combination with the time dimension information;

[0136] Historical correlation unit: extracts historical meteorological data and geological activity records from a preset historical database for each abnormal area, determines the corresponding abnormal cause, and then outputs the terrain evolution results, including the evolution type, spatial range, time process and a list of correlation factors;

[0137] Map generation unit: performs multi-index risk assessment based on terrain evolution results and a preset disaster rule library to generate a risk zoning map;

[0138] Report generation unit: Integrate terrain evolution results with risk zoning maps, and generate analysis reports including terrain evolution trends and disaster risk assessments according to preset templates.

[0139] In this embodiment, dividing the four-dimensional spatiotemporal data set by a time domain sliding window is a dynamic segmentation analysis method, which divides the three-dimensional geographic information of a continuous time series into several time slices, each slice representing the change in the surface state during a specific period of time. The window size is dynamically adjusted to adapt to the evolution rate of different landforms - for example, a short window (such as once a day) is used in a rapidly changing landslide area, while a long window (such as once a month) is used in a long-term settlement area. For example, when monitoring the ground settlement in a mining area, the system uses a sliding step of 7 days and takes 30 days of point cloud data each time to calculate the cumulative deformation, which is equivalent to forming overlapping "observation windows" on the time axis to ensure that both short-term drastic changes (such as sudden blasting vibrations) and long-term trends (such as slow settlement) are captured, and finally outputs a sequence of deformation heat maps indexed by time.

[0140] In this embodiment, feature extraction in combination with preset feature classification rules includes multimodal data association analysis and feature fusion. The feature classification rules define identification marks of different categories (such as buildings, vegetation, and exposed surfaces) (such as buildings have regular geometric edges and vegetation has high near-infrared reflectivity). Based on this, the target features are separated from each time window, and then their key indicators are calculated: geometric deformation (by comparing elevation changes through point cloud registration), spectral reflectivity change rate (multi-period image band difference), and point cloud density fluctuation (reflecting surface roughness changes). For example, for urban expansion areas, the system identifies newly built roads (regular long strips of low reflectivity) and extracts the trend of their roadbed point cloud density increasing over time. At the same time, it detects whether the surrounding buildings are tilted (geometric deformation exceeds the limit). Finally, the sensor differences are eliminated through Z-score standardization to generate a space-time matrix with consistent dimensions, where each row represents a feature instance and each column is a time-varying feature value.

[0141] In this embodiment, a dynamic adaptive method is used to set the grading threshold based on the standardized feature matrix, taking into account the differences in regional geological background. The deformation threshold in gentle areas (such as plains) is set to a low value (a small change will trigger an alarm), while higher fluctuations are allowed in complex areas (such as fault zones). The spectral feature threshold is adjusted in combination with the phenological laws of vegetation - the increase in reflectivity in the dry season may be a normal phenomenon. For example, the area around a reservoir is identified as a transition zone. The system detects that the deformation of local points on the reservoir bank has exceeded 2 times the weekly average threshold for three consecutive weeks. At the same time, the near-infrared reflectivity has dropped abnormally (possibly due to water seepage causing vegetation to wither). Combined with the time dimension, it is determined that the abnormality is a non-seasonal fluctuation. Finally, an abnormal area of 200 meters × 150 meters is framed and marked with a priority.

[0142] In this embodiment, a causal inference engine is used to correlate historical data and output terrain evolution results, matching similar cases from a geological database. For example, an abnormal deformation zone in a mountainous area is matched to a record of a shallow landslide caused by heavy rain at the same location five years ago. The current deformation pattern (trailing edge tensile cracks + leading edge bulging) is highly consistent with this record. At the same time, the recent cumulative rainfall is retrieved to reach the 90th percentile of the historical record. Based on this, the output evolution result is a "shallow creeping landslide" with a spatial range extending 350 meters along the slope direction. The time process shows an accelerating trend. The list of associated factors includes "heavy rainfall" and "loose accumulation layer". The results are structured and stored as a traceable evolution event chain.

[0143] In this embodiment, a multi-layer fusion architecture is used to integrate data to generate an analysis report. The terrain evolution results (such as landslide boundary vectors) are superimposed on the risk map (red-yellow-green partitions), and conclusions are automatically generated through natural language templates. For example: "Area A (coordinates XXX) has accumulated a 12cm sinking in the past three months, and the deformation rate has increased by 210% compared with the previous period. Referring to the "Geological Hazard Risk Assessment Standards", it is determined to be a high-risk area, and it is recommended to deploy GNSS monitoring stations." The report is synchronously embedded with a timeline controller, and users can interactively view the risk evolution process at each stage.

[0144] The beneficial effects of the above technical solution are as follows: the analysis and decision-making module extracts a standardized feature matrix from the four-dimensional spatiotemporal dataset through the feature extraction unit, ensuring spatiotemporal continuity and data quality. The region identification unit accurately identifies abnormal areas in flat areas, transition areas, and complex areas based on graded thresholds. The historical correlation unit combines historical meteorological and geological data to accurately determine the causes of anomalies and output detailed terrain evolution results. The map generation unit generates risk zoning maps through multi-indicator assessment, visually displaying disaster risks. The report generation unit integrates analysis results and produces structured analysis reports, providing a scientific basis for terrain evolution trends and disaster risk assessment. This solution significantly improves analysis accuracy and efficiency, providing reliable support for disaster prevention and control and planning decisions.

[0145] Example 8:

[0146] The embodiment of the present invention provides a geographic information visualization intelligent analysis system based on drone mapping, and a visualization platform module, including:

[0147] Parameter conversion unit: Based on the preset visualization rule library, the analysis report is converted into unified visualization parameters;

[0148] Hierarchical division unit: Based on the terrain complexity classification results and time dimension information, the 3D geographic information model is divided into several LOD levels to generate a hierarchical management list;

[0149] Parameter determination unit: Based on the embedded posture sensor and interaction event listener, it captures the three-dimensional coordinates, pitch angle and zoom ratio parameters of the user's perspective in real time and determines the standardized perspective parameters;

[0150] View annotation unit: Synchronizes the standardized viewing angle parameters with the hierarchical management list, and annotates the complexity of the target area within the user's current view;

[0151] Instruction generation unit: calls a preset rendering strategy rule library and the complexity of the target area in the user's current field of view to determine the priority of the target area in the user's current field of view, and generates a rendering instruction queue based on the priority of the target area in the user's current field of view;

[0152] Visualization unit: According to the rendering command queue, model blocks and texture maps of different precisions are loaded frame by frame. Asynchronous streaming technology is used to prioritize loading high-precision data in the center of the field of view, and delayed filling of edge areas. After each frame is rendered, the frame rate and video memory usage are counted through the preset performance monitoring tool. If the frame rate is detected to fall below the preset frame rate threshold, the shadow quality and particle special effect level of the non-focused area in the user's current field of view are reduced, and the final output rendering picture is displayed on the preset visual interactive interface.

[0153] In this embodiment, the preset visualization rule base is a multi-dimensional mapping logic library, which is used to define how abstract data in the analysis report (such as terrain evolution rate, risk level) is converted into visual elements (color, transparency, dynamic marking, etc.). The rule base adopts a hierarchical structure, including a risk level-color mapping table (such as red represents an extremely high risk area), a change trend-dynamic arrow direction association (such as an upward arrow indicates surface uplift), and a time dimension-timeline segmentation strategy (such as near-real-time data using high refresh rate animation). For example, when the analysis report identifies that a certain area has an average annual surface subsidence disaster risk of 5 cm, the rule base automatically matches the orange gradient fill corresponding to the "medium risk" level, and superimposes a downward arrow symbol and a translucent shadow in the area, and generates a timeline slider in the map sidebar for users to review the subsidence process over the past three years.

[0154] In this embodiment, converting the analysis report into a unified visualization parameter includes two stages: data standardization and visual encoding. First, the structured indicators (such as deformation and risk value) in the report are extracted, and the dimensional differences are eliminated through normalization. Then, based on the rule base, the numerical interval is mapped to a visual variable. For example, the medium-risk interval of 0.7 is converted to a yellow stripe overlay, and the high-risk interval of 0.7 to 1 uses a red warning flashing border. For example, when a landslide risk assessment report in a mountainous area shows that the risk value of area A is 0.82 and that of area B is 0.45, the system automatically generates a red flashing warning box for area A and a yellow diagonal line fill for area B, and distinguishes the two with highlight outlines of different brightness on the three-dimensional model.

[0155] In this embodiment, LOD levels are divided and a hierarchical management list is generated, including: dividing the three-dimensional model into multiple levels of detail (LOD) levels according to the terrain complexity classification results (flat area / transition area / complex area) and time dimension information (such as the frequency of changes in the past five years). Flat areas (such as farmland) only require 15 levels of high-precision models, and additional timeline keyframe caches are added to areas with frequent time changes. For example, in a flood simulation scenario, the river area (complex area) retains the building outline and texture with a precision of 0.5 meters, and the flood history data is stored as a keyframe every 6 hours; while the surrounding plains (flat areas) only use a 2-meter resolution grid, and the timeline is sampled on a daily basis. The final generated list clearly marks the LOD index and time resolution of each area.

[0156] In this embodiment, the embedded attitude sensor (such as IMU) and the interactive event listener (such as mouse / VR handle) jointly capture the original viewing angle signal. The sensor provides device posture data (such as the XYZ coordinates of the helmet), and the interactive event parses the user's operation intention (such as the viewing angle focus offset when zooming). After eliminating the jitter noise through Kalman filtering, the stable viewing angle parameters are output. For example, when a user wears a VR device to observe a geological disaster site, the system tracks the helmet position (such as 2.3 meters east, 15° depression angle) and the handle zoom action (focal length adjusted to 200 meters) in real time, and converts it into standardized "observation point (120, 50, 10), line of sight vector (0.97, 0, 0.25), viewing distance 200 meters" parameters for subsequent field of view calculation.

[0157] In this embodiment, standardized viewing angle parameters are dynamically associated with the hierarchical management list. The current visible area is calculated through view frustum clipping, and key areas are labeled with terrain complexity tags. For example, when the user's view is facing a hillside (complex area), the view frustum space equation determines that 70% of the hillside is within the center of the field of view and is labeled as "high complexity - mandatory LOD level 4." At the same time, the woods at the edge of the field of view (transition area) are identified as "medium complexity - downgradable to LOD level 2," achieving dynamic load distribution.

[0158] In this embodiment, the rendering strategy rule library combines the target area complexity (e.g., buildings = priority 1) with the user's gaze state (e.g., continuous gaze for more than 2 seconds = increasing priority) to generate layered rendering instructions. For example, when the user focuses on an earthquake fault zone (priority 1), the command queue prioritizes loading a 0.2-meter-precision fracture mesh in the next frame. If farmland (priority 3) is also present on the edge of the field of view, its simplified 1-meter-precision model is loaded later, and GPU task scheduling is managed through the command buffer pool.

[0159] In this embodiment, the visualization unit dispatches resources based on the frame of the instruction queue. The ground fissure in the center of the field of view (high priority) immediately triggers the upload of high-precision textures, while the distant highway (low priority) uses a placeholder model and is gradually refined when the GPU is idle. When performance monitoring finds that the frame rate drops to 25FPS, the shadow sampling rate of the non-gazing area (such as the background mountain) is automatically reduced, and the water surface reflection effect is disabled to ensure that the physical simulation of the main fissure remains at 60FPS. For example, in a volcanic eruption simulation, the lava flow area always maintains particle effects, but when the frame rate fluctuates, the leaf swaying animation of the distant vegetation will be paused to save computing power.

[0160] The beneficial effects of the above technical solution are as follows: the visualization platform module converts the analysis report into unified visualization parameters through the parameter conversion unit to ensure data consistency. The hierarchical division unit generates an LOD hierarchical management list based on the terrain complexity and time dimension to optimize model management. The parameter determination unit captures the user's perspective parameters in real time, and the field of view annotation unit accurately annotates the complexity of the area within the field of view. The instruction generation unit generates an optimized rendering queue based on the rendering strategy and priority. The visualization unit adopts asynchronous streaming, prioritizes loading high-precision data in the center of the field of view, dynamically adjusts the rendering quality of the edge area, and optimizes the frame rate and video memory occupancy through performance monitoring. This solution significantly improves rendering efficiency, interactive fluency and visualization effects, providing users with an efficient and intuitive geographic information interaction experience.

[0161] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. The geographic information visualization intelligent analysis system based on UAV mapping is characterized by: include: UAV aerial survey module: Based on the multi-source sensors carried by UAVs, regional aerial surveys are carried out to obtain multimodal geographic data including visible light images, infrared images, and laser point clouds, and perform pre-processing; Data processing module: spatially aligns the pre-processed multimodal geographic data with the ground control points in the preset ground control point database to form an initial geographic dataset in a standard coordinate system; 3D modeling module: Extracts and fuses features from the initial geographic dataset to obtain fused geographic feature data, constructs a 3D geographic information model based on the fused geographic feature data, optimizes the 3D geographic information model, and embeds time dimension information into the optimized 3D geographic information model to form a 4D spatiotemporal dataset. Analysis and decision-making module: Conducts intelligent analysis based on four-dimensional spatiotemporal data sets to generate analysis reports including terrain evolution trends and disaster risk assessments; Visualization platform module: visualize the analysis report through a preset visual interactive interface, and adjust the rendering accuracy of the optimized three-dimensional geographic information model in real time according to changes in user perspective.

2. The geographic information visualization intelligent analysis system based on drone mapping according to claim 1 is characterized in that: UAV aerial survey module, including: Route planning unit: obtains boundary information and terrain characteristic parameters of the target area, and plans the optimal flight route according to the preset surveying and mapping accuracy requirements; Data acquisition unit: controls the drone equipped with multi-source sensors to fly along the optimal flight route, and simultaneously collects multimodal geographic data of the target area, including visible light images, infrared images, and laser point cloud data; Flight Adjustment Unit: During the flight acquisition process, the unit dynamically adjusts the drone's flight altitude, shooting interval, and scanning frequency by comparing preset aerial survey parameters with the characteristics of actual collected data. It also evaluates the integrity of data collected by multi-source sensors and triggers a re-flight mechanism when the data loss rate exceeds a preset threshold. Data preprocessing unit: The acquired multimodal geographic data including visible light images, infrared images, and laser point cloud data are matched with the POS data recorded by the UAV positioning and attitude determination system in time and space to obtain the preprocessed multimodal geographic data.

3. The geographic information visualization intelligent analysis system based on drone mapping according to claim 1 is characterized in that: Data processing module, including: Control point extraction unit: extracts a set of control points that match the target area from a preset ground control point database, including coordinate information and feature descriptions of plane control points and elevation control points; Spatial registration unit: establishes feature matching relationships between pre-processed multimodal geographic data and control points in the control point set and performs spatial registration; Registration optimization unit: gradually optimizes the spatial registration accuracy and re-performs spatial registration when the registration deviation exceeds a preset threshold; Registration Verification Unit: Build a quality control indicator system to evaluate the integrity, consistency, and accuracy of the registered initial geographic dataset to generate a registration reliability index; Result output unit: When the registration reliability index meets the preset result, the registered initial geographic dataset is output based on the spatial registration result.

4. The geographic information visualization intelligent analysis system based on drone mapping according to claim 1 is characterized in that: 3D modeling module, including: Feature extraction unit: performs multi-scale feature analysis on the initial geographic data set to extract structured information including terrain contours, feature boundaries, and texture features; Feature analysis unit: performs similarity analysis on structured information, aligns and fuses data from different sensors based on similarity, and constructs a fused geographic feature dataset with geometric accuracy and texture authenticity; Model conversion unit: uses triangulation algorithm to convert fused data into three-dimensional geographic information model; Model optimization unit: optimize the three-dimensional geographic information model; Time embedding unit: establishes a time coding mechanism on the three-dimensional geographic information model, associates historical change data with the three-dimensional geographic information model, and generates a four-dimensional spatiotemporal dataset with time dimension information.

5. The geographic information visualization intelligent analysis system based on drone mapping according to claim 4 is characterized in that: Model optimization unit, including: Feature separation subunit: performs multi-dimensional feature separation based on the initial geographic data set, decouples the data features obtained by different sensors, and separates geometric features, spectral features, and temporal features; Structural optimization subunit: performs feature reorganization based on the multi-dimensional feature separation results, adjusts the feature fusion weight according to the type of ground feature, and generates an optimized structural model; Terrain classification subunit: Classify terrain complexity based on optimized structural model and time dimension information; Grid refinement subunit: implements dynamic grid subdivision based on terrain complexity classification results; Density optimization subunit: obtains the user's viewing distance and optimizes the mesh density after dynamic mesh segmentation based on the user's viewing distance.

6. The geographic information visualization intelligent analysis system based on drone mapping according to claim 5 is characterized in that: The terrain complexity classification results include: dividing the target area into: flat area, transition area and complex area.

7. The geographic information visualization intelligent analysis system based on drone mapping according to claim 1 is characterized in that: Analysis and decision-making module, including: Feature extraction unit: This unit divides the four-dimensional spatiotemporal data set into a time-domain sliding window and extracts the geometric deformation, spectral reflectance change rate, and point cloud density fluctuation characteristics of each feature area at different time points based on the preset feature classification rules. The extracted original features are then normalized and outliers are filtered out to output a standardized feature matrix with spatiotemporal continuity. Region identification unit: Based on the standardized feature matrix, it sets the classification thresholds of flat areas, transition areas, and complex areas according to the geometric deformation and spectral characteristics, and identifies several abnormal areas in combination with the time dimension information; Historical correlation unit: extracts historical meteorological data and geological activity records from a preset historical database for each abnormal area, determines the corresponding abnormal cause, and then outputs the terrain evolution results, including the evolution type, spatial range, time process and a list of correlation factors; Map generation unit: performs multi-index risk assessment based on terrain evolution results and a preset disaster rule library to generate a risk zoning map; Report generation unit: Integrate terrain evolution results with risk zoning maps, and generate analysis reports including terrain evolution trends and disaster risk assessments according to preset templates.

8. The geographic information visualization intelligent analysis system based on drone mapping according to claim 5 is characterized in that: Visualization platform modules, including: Parameter conversion unit: Based on the preset visualization rule library, the analysis report is converted into unified visualization parameters; Hierarchical division unit: Based on the terrain complexity classification results and time dimension information, the 3D geographic information model is divided into several LOD levels to generate a hierarchical management list; Parameter determination unit: Based on the embedded posture sensor and interaction event listener, it captures the three-dimensional coordinates, pitch angle and zoom ratio parameters of the user's perspective in real time and determines the standardized perspective parameters; View annotation unit: Synchronizes the standardized viewing angle parameters with the hierarchical management list, and annotates the complexity of the target area within the user's current view; Instruction generation unit: calls a preset rendering strategy rule library and the complexity of the target area in the user's current field of view to determine the priority of the target area in the user's current field of view, and generates a rendering instruction queue based on the priority of the target area in the user's current field of view; Visualization unit: According to the rendering command queue, model blocks and texture maps of different precisions are loaded frame by frame. Asynchronous streaming technology is used to prioritize loading high-precision data in the center of the field of view, and delayed filling of edge areas. After each frame is rendered, the frame rate and video memory usage are counted through the preset performance monitoring tool. If the frame rate is detected to fall below the preset frame rate threshold, the shadow quality and particle special effect level of the non-focused area in the user's current field of view are reduced, and the final output rendering picture is displayed on the preset visual interactive interface.

Images