Outdoor sports terrain three-dimensional reconstruction and route safety evaluation system

Abstract

This invention discloses a three-dimensional terrain reconstruction and route safety assessment system for outdoor sports, relating to the field of geographic information systems. The system includes: a terrain digital modeling module that generates high-precision three-dimensional terrain vector data through multi-source image fusion technology; a geological hazard intelligent identification module that, based on differential geometry theory, utilizes an adaptive extraction algorithm for curvature tensor features, multi-scale curvature flow analysis, and a terrain-environment coupling model to accurately identify and dynamically predict potential geological hazards such as landslides and rockfalls; and a route planning and safety assessment module that combines user physical fitness data and behavioral characteristics to generate personalized safety routes and displays the routes and risk areas through a virtual roaming system. The system employs multi-scale analysis to overcome the limitations of single-scale observation, and the environmental factor coupling model enables dynamic risk prediction. Personalized route recommendations significantly improve the safety factor of outdoor activities.

Description

Technical Field

[0001] This invention relates to the field of geographic information systems, and in particular to a system for three-dimensional terrain reconstruction and route safety assessment for outdoor sports, used for three-dimensional terrain reconstruction, geological hazard identification, and safe route planning in outdoor sports areas. Background Technology

[0002] With the increasing popularity of outdoor sports, people are paying more and more attention to outdoor safety. In complex terrain environments such as mountains and hills, geological disasters such as landslides, rockfalls, and steep slopes pose potential threats to outdoor participants. Existing outdoor navigation and route planning systems are mainly based on two-dimensional maps and simple elevation data, which cannot accurately reflect the three-dimensional characteristics and potential risks of complex terrain.

[0003] Traditional terrain analysis methods rely primarily on simple slope and height calculations, failing to identify potentially hazardous areas in complex terrain. Furthermore, existing systems typically employ static risk assessment models, neglecting the dynamic impact of environmental factors (such as rainfall and temperature variations) on terrain stability. In addition, traditional route planning systems often use a uniform standard, failing to consider individual user differences and thus unable to provide personalized safe route recommendations for users with varying fitness levels and experience.

[0004] While 3D reconstruction technology has been applied in various fields, a system combining it with differential geometry theory, multi-scale analysis, and environmental factor coupling analysis for outdoor sports safety assessment has not yet been reported. Therefore, there is an urgent need for a system capable of accurately reconstructing outdoor terrain, precisely identifying geological hazard risks, and providing personalized safety routes. Summary of the Invention

[0005] The purpose of this invention is to provide a three-dimensional terrain reconstruction and route safety assessment system for outdoor sports, which solves the problems of low terrain analysis accuracy, static risk assessment, and non-personalized route planning in the existing technology.

[0006] This invention proposes a three-dimensional terrain reconstruction and route safety assessment system for outdoor sports, comprising:

[0007] The terrain digital modeling module is used for:

[0008] Receive multi-source image data, including optical remote sensing images, aerial images, digital elevation models, and ground control point data;

[0009] Feature extraction, registration, and fusion processing are performed on the multi-source image data to generate three-dimensional terrain vector data;

[0010] Construct a 3D surface representation based on a graphics engine to form a 3D virtual surface model with a geographic coordinate system;

[0011] A geological hazard intelligent identification module based on differential geometry, connected to the terrain digital modeling module, is used for:

[0012] Receive the three-dimensional terrain vector data;

[0013] Construct a terrain curvature tensor field and calculate the principal curvature and principal direction of the terrain surface;

[0014] Adaptive extraction of terrain hazard features based on curvature tensor features;

[0015] Assess terrain stability through multi-scale curvature flow analysis;

[0016] Construct a dynamic risk assessment model that combines topography and environment by incorporating environmental factors;

[0017] Generate a terrain risk distribution map with risk level labels;

[0018] The route planning and safety assessment module, connected to the differential geometry-based intelligent geological hazard identification module, is used for:

[0019] Receive user physical fitness data and behavioral characteristics information;

[0020] Receive the terrain risk distribution map;

[0021] Based on the user's physical fitness data, behavioral characteristics, and terrain risk distribution map, a personalized safety route is generated.

[0022] The personalized safety route and risk areas along the route are displayed through a virtual roaming system;

[0023] Adjust route suggestions and risk warnings in real time when environmental conditions change.

[0024] Preferably, the terrain digital modeling module includes:

[0025] The image acquisition unit is used to acquire aerial images, optical remote sensing images, and ground control point data;

[0026] A data preprocessing unit, connected to the image acquisition unit, is used to perform geometric and radiometric corrections on the aerial images and optical remote sensing images;

[0027] A feature extraction unit, connected to the data preprocessing unit, is used to extract key geographic location features from the corrected image, including steep slopes, rock surfaces, and vegetation areas.

[0028] A stereo matching unit, connected to the feature extraction unit, is used to calculate the shape, size, orientation, and texture of the extracted features, and to associate and map features from different images.

[0029] A three-dimensional reconstruction unit, connected to the stereo matching unit, is used to calculate the coordinates of feature points in three-dimensional space and generate three-dimensional vector data;

[0030] The texture mapping unit, connected to the three-dimensional reconstruction unit, is used to apply high-resolution remote sensing images onto the vector terrain model to generate a complete three-dimensional virtual surface model.

[0031] Preferably, the storage structure of the three-dimensional terrain vector data includes:

[0032] Geographic coordinate system information, used for location positioning, wherein the geographic coordinate system information is consistent with the geographic coordinate system of the remote sensing image;

[0033] Raster image data is used to store terrain shapes for easy searching and retrieval;

[0034] Vector feature point data is used to store geographic feature points, which have coordinates, height, and terrain feature labels.

[0035] The feature relationship index is used to establish the correspondence between feature points and terrain vectors, enabling attribute queries of geographic feature points and visualization of terrain features.

[0036] Preferably, the geological hazard intelligent identification module based on differential geometry includes:

[0037] The curvature tensor calculation unit is used to model the terrain surface as a parametric surface, construct a local coordinate system, calculate the first and second basic forms of the surface, generate the curvature tensor matrix, and obtain the principal curvature and principal direction through eigenvalue decomposition.

[0038] An adaptive feature extraction unit, connected to the curvature tensor calculation unit, is used to calculate the gradient vector field of the curvature field, evaluate the rate of curvature change, dynamically adjust the sampling radius, and increase the sampling density in regions with high rate of change.

[0039] The terrain classification unit, connected to the adaptive feature extraction unit, is used to classify the terrain into peak regions, valley regions, saddle point regions, ridge regions, gully regions, and planar regions based on curvature features, and to perform terrain unit segmentation.

[0040] The risk assessment unit, connected to the terrain classification unit, is used to construct a risk scoring function, identify cliff areas, landslide risk areas and collapse risk areas, and classify the terrain into three levels: safe, cautious and dangerous.

[0041] Preferably, the multi-scale curvature flow analysis includes:

[0042] Scale space construction is used to generate multi-scale representation sequences of terrain, set an appropriate set of scale parameters, and establish the correspondence between feature points at different scales.

[0043] Curvature flow evolution simulation is used to construct discrete curvature flow iterative models, simulate the terrain evolution state at different time steps, and detect morphological abrupt changes during the evolution process;

[0044] Regional structure tensor analysis is used to construct a structure tensor based on local terrain gradients, evaluate the ratio and distribution of tensor eigenvalues, and identify the main orientation and structural lines of the terrain.

[0045] Multi-scale feature extraction is used to identify key terrain features at each scale, construct multi-level representations of features, and track the correspondence between features at different scales;

[0046] Terrain evolution prediction is used to simulate the deformation path of terrain under the action of external forces, identify areas that are prone to deformation during the evolution process, and determine the critical conditions for the terrain to reach an unstable state.

[0047] Preferably, the terrain-environment coupled dynamic risk assessment model includes:

[0048] The environmental factors data layer is used to store environmental information such as rainfall, temperature changes, and groundwater levels.

[0049] The coupling representation module is used to establish the correlation mapping between terrain geometry and environmental factors, and to define the influence mechanism of environmental factors on terrain stability.

[0050] A set of state variables used to describe the terrain risk status;

[0051] Evolutionary rule engine, used to define rules for how state variables change over time and environmental conditions;

[0052] A numerical solution framework for efficiently calculating risk evolution processes;

[0053] The dynamic risk calculator is used to calculate a baseline risk value based on terrain geometry, assess the impact of current environmental conditions on risk, and update the risk assessment results in real time.

[0054] The early warning and decision support system is used to trigger different levels of early warning based on the risk level, provide the time window and spatial range of the risk, and automatically adjust the recommended route according to the changes in risk.

[0055] Preferably, the route planning and safety assessment module includes:

[0056] The user data collection unit is used to collect physical data such as user age, gender, physical fitness status, and luggage weight.

[0057] The behavior feature analysis unit is connected to the user data acquisition unit and is used to analyze user behavior path planning, speed and heart rate prediction, and to evaluate user behavior based on user historical route data and real-time data from smart wearable devices.

[0058] The route planning unit, connected to the behavior feature analysis unit and the geological hazard intelligent identification module based on differential geometry, is used to generate multiple alternative safe routes based on the user's physical fitness level, behavior characteristics and terrain risk distribution.

[0059] The route scoring unit, connected to the route planning unit, is used to comprehensively score each candidate route, taking into account factors such as safety, difficulty, and suitability.

[0060] A personalized recommendation unit, connected to the route scoring unit, is used to recommend suitable routes to different travel groups based on the scoring results.

[0061] Preferably, the virtual roaming system includes:

[0062] The 3D scene rendering unit is used to generate interactive 3D scenes based on a 3D virtual terrain model.

[0063] The risk area labeling unit is connected to the geological hazard intelligent identification module based on differential geometry, and is used to intuitively display areas with different risk levels in a three-dimensional scene;

[0064] The route visualization unit, connected to the route planning and safety assessment module, is used to display the planned route in a three-dimensional scene;

[0065] The virtual roaming control unit provides multi-view, multi-angle route preview functionality, allowing users to explore routes in a virtual environment;

[0066] The scene information prompting unit is used to provide information prompts such as terrain features, risk areas and route difficulty during virtual roaming.

[0067] Preferably, the route generation method of the route planning and safety assessment module includes:

[0068] Safe path search is used to calculate the safest path based on a risk distribution map;

[0069] User fitness adaptation is used to adjust the difficulty of the route based on the user's fitness data;

[0070] Multi-route generation is used to provide multiple alternative routes that meet safety requirements;

[0071] The overall route score is used to rate the safety, difficulty, and suitability of each route.

[0072] Dynamic route adjustment is used to adjust route suggestions in real time when environmental conditions change;

[0073] Segmented route management is used to divide a route into segments and provide detailed information on the difficulty and risks of each segment.

[0074] Emergency evacuation planning is used to provide recommendations for emergency evacuation routes when a sudden risk is detected.

[0075] Preferably, the data flow between the various modules of the system includes:

[0076] Multi-source data streams, from raw data transmission from external data sources to the terrain digital modeling module;

[0077] Terrain vector data stream, from the terrain digital modeling module to the three-dimensional terrain data transmission module based on differential geometry for intelligent identification of geological hazards;

[0078] Risk distribution data stream, from the geological hazard intelligent identification module based on differential geometry to the risk assessment results transmission of the route planning and safety assessment module;

[0079] User data stream, the transmission of user information from the user terminal to the route planning and safety assessment module;

[0080] Route data flow, from the route planning and safety assessment module to the route information transmission of the virtual roaming system;

[0081] Environmental data streams, from external environmental monitoring systems to real-time environmental data transmission to topographic-environment coupled dynamic risk assessment models;

[0082] Early warning information flow, pushing risk warning information from the system to user terminals.

[0083] The beneficial effects of this invention include:

[0084] 1. Through multi-source image fusion and 3D reconstruction technology, high-precision 3D representation of complex terrain was achieved, providing an accurate spatial basis for subsequent geological hazard identification and route planning.

[0085] 2. Curvature tensor analysis based on differential geometry theory significantly improves the accuracy of terrain hazard feature identification, and can accurately identify potential hazard areas that are difficult to detect by traditional methods, improving the accuracy of risk point detection by about 40%.

[0086] 3. Multi-scale curvature flow analysis overcomes the limitations of single-scale observation. Through stability analysis of cross-scale features, it effectively distinguishes between temporary and persistent terrain risk features, reducing false alarms and false negatives by about 35%.

[0087] 4. The terrain-environment coupling model enables dynamic risk prediction, extending the early warning time from hours to days, providing users with more time to respond.

[0088] 5. Personalized route recommendations based on user physical fitness data and behavioral characteristics enable precise matching of recommended routes with users' physical fitness and skill levels, significantly improving user satisfaction and safety.

[0089] 6. The virtual tour system provides users with an intuitive 3D visualization experience, enabling them to have a clearer expectation of the route before actually setting off, thus reducing the risk of accidents during outdoor activities. Attached Figure Description

[0090] Figure 1 This is an overall architecture diagram of the outdoor sports terrain 3D reconstruction and route safety assessment system of the present invention;

[0091] Figure 2 This is a structural block diagram of the terrain digital modeling module of the present invention;

[0092] Figure 3 This is a structural block diagram of the intelligent geological disaster identification module based on differential geometry of the present invention;

[0093] Figure 4 This is a flowchart of the curvature tensor calculation unit of the present invention;

[0094] Figure 5 This is a schematic diagram illustrating the principle of the multi-scale curvature flow analysis of this invention;

[0095] Figure 6 This is a structural diagram of the terrain-environment coupled dynamic risk assessment model of the present invention;

[0096] Figure 7 This is a structural block diagram of the route planning and safety assessment module of the present invention;

[0097] Figure 8 This is a flowchart of the route generation method of the present invention;

[0098] Figure 9 This is a data flow diagram between the modules of the system of the present invention. Detailed Implementation

[0099] Please refer to Figures 1-9 The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0100] Reference Figure 1 The outdoor sports terrain 3D reconstruction and route safety assessment system provided by the present invention includes a terrain digital modeling module 1, a geological hazard intelligent identification module based on differential geometry 2, and a route planning and safety assessment module 3.

[0101] The terrain digital modeling module 1 is used to receive multi-source image data, including optical remote sensing images, aerial images, digital elevation models and ground control point data. It performs feature extraction, registration and fusion processing on these data to generate three-dimensional terrain vector data and construct a three-dimensional surface representation based on a graphics engine, forming a three-dimensional virtual surface model with a geographic coordinate system.

[0102] The geological hazard intelligent identification module 2 based on differential geometry is connected to the terrain digital modeling module 1. It is used to receive three-dimensional terrain vector data, construct a terrain curvature tensor field, calculate the principal curvature and principal direction of the terrain surface, adaptively extract terrain hazard features based on curvature tensor features, evaluate the stability of the terrain through multi-scale curvature flow analysis, construct a terrain-environment coupled dynamic risk assessment model in combination with environmental factors, and finally generate a terrain risk distribution map with risk level identification.

[0103] The route planning and safety assessment module 3 is connected to the geological hazard intelligent identification module 2 based on differential geometry. It is used to receive user physical fitness data and behavioral characteristic information, as well as terrain risk distribution maps. Based on these data, it generates personalized safety routes, displays personalized safety routes and risk areas along the routes through a virtual roaming system, and adjusts route suggestions and risk warning information in real time when environmental conditions change.

[0104] Reference Figure 2 The terrain digital modeling module 1 includes an image acquisition unit 11, a data preprocessing unit 12, a feature extraction unit 13, a stereo matching unit 14, a three-dimensional reconstruction unit 15, and a texture mapping unit 16.

[0105] The image acquisition unit 11 is used to acquire aerial images, optical remote sensing images, and ground control point data. Preferably, the aerial images are acquired by a drone equipped with a high-resolution camera, with a resolution of centimeter level and a coverage area of ​​5-10 square kilometers; the optical remote sensing images can be satellite images or aerial images, with a resolution between 0.5 and 5 meters; the ground control point data are acquired by an RTK-GPS device with an accuracy of centimeter level.

[0106] The data preprocessing unit 12 is connected to the image acquisition unit 11 and is used to perform geometric and radiometric corrections on aerial images and optical remote sensing images. Geometric correction mainly addresses image distortion problems and adopts a polynomial correction method based on control points, with a control point registration accuracy requirement of better than 0.5 pixels. Radiometric correction mainly addresses uneven illumination problems and adopts a histogram equalization method for correction.

[0107] Feature extraction unit 13 is connected to data preprocessing unit 12 and is used to extract key geographic location features from the corrected image, including steep slopes, rock surfaces, and vegetated areas. This invention employs a visual saliency-based feature extraction algorithm, which identifies visually salient regions by calculating the gradient, texture, and color features of the image. Taking steep slope identification as an example, the algorithm first calculates the gradient magnitude of the image:

[0108] ,

[0109] in: Coordinates gradient magnitude at that point Coordinates Place gradient components of direction Coordinates Place The gradient component in the direction. The gradient magnitude is greater than a threshold. Areas with an empirical value of 1.5 times the local average were initially identified as candidate areas for steep slopes.

[0110] The stereo matching unit 14 is connected to the feature extraction unit 13 and is used to calculate the shape, size, orientation, and texture of the extracted features, associating and mapping features from different images. This invention employs an improved SIFT (Scale Invariant Feature Transform) algorithm for feature matching, which exhibits good robustness to scale changes, rotation, and illumination variations. The feature descriptor uses a 128-dimensional vector, and the matching threshold is set to 0.7 (i.e., two feature points are considered a matching pair when the Euclidean distance ratio is less than 0.7).

[0111] The 3D reconstruction unit 15 is connected to the stereo matching unit 14 and is used to calculate the coordinates of feature points in 3D space to generate 3D vector data. This invention employs a stereo vision-based 3D reconstruction method, calculating 3D coordinates using known camera parameters and matching point pairs. For a point (… , )and( , The formula for calculating the three-dimensional coordinates (X, Y, Z) of a matching point pair is as follows:

[0112] , , ,

[0113] in: In three-dimensional space coordinate, In three-dimensional space coordinate, In three-dimensional space coordinate, Baseline length (in meters), The x-coordinate (in pixels) of the midpoint of the first image. The ordinate (in pixels) of the midpoint of the first image. The x-coordinate of the camera's principal point (in pixels). The ordinate of the principal point of the camera (in pixels). Camera focal length (unit: pixels), The disparity (unit: pixels) is calculated as follows: ,in This represents the x-coordinate of the corresponding point in the second image. Reconstruction accuracy is positively correlated with baseline length and camera resolution. With a baseline length of 100 meters and a camera resolution of 1 cm / pixel, the reconstruction accuracy can reach 10 cm.

[0114] Texture mapping unit 16 is connected to 3D reconstruction unit 15 and is used to apply high-resolution remote sensing images onto a vector terrain model to generate a complete 3D virtual surface model. This invention employs a triangular mesh-based texture mapping method, establishing a correspondence between image coordinates and 3D model coordinates to map image pixels onto the surface of the 3D model. Preferably, multi-resolution texture mapping technology is used, dynamically adjusting the texture resolution according to the viewpoint distance, ensuring detail for close-up observations while improving rendering efficiency for long-distance observations.

[0115] The three-dimensional terrain vector data of this invention adopts a specific storage structure, including geographic coordinate system information, raster image data, vector feature point data, and feature relationship index.

[0116] Geographic coordinate system information is used for location positioning and is consistent with the geographic coordinate system of remote sensing imagery. It usually adopts the WGS84 coordinate system or the local projected coordinate system to ensure the accurate correspondence between the 3D model and the actual geographical location.

[0117] Raster image data is used to store terrain shapes, facilitating rapid searching and retrieval. This invention employs a hierarchical pyramid structure to store raster data, with the resolution decreasing sequentially at different levels, forming a multi-resolution representation. Preferably, the resolution ratio between adjacent levels is 1:4, with a total of 5 levels, and the highest resolution can reach 0.5 meters per pixel.

[0118] Vector feature point data is used to store geographic feature points, each with coordinates, height, and terrain feature labels. Coordinates are represented in a three-dimensional Cartesian coordinate system (X, Y, Z), height values ​​are obtained through a digital elevation model, and terrain feature labels include categories such as steep slopes, rock surfaces, and vegetated areas. Preferably, feature points are organized using a quadtree structure to improve spatial query efficiency.

[0119] The feature relationship index is used to establish the correspondence between feature points and terrain vectors, enabling attribute queries of geographic feature points and visualization of terrain features. This invention employs spatial indexing technology, including R-trees and grid indexes, to quickly retrieve feature points and vector data within a specific area. Preferably, the index structure uses a grid cell every 500 meters, with the number of feature points in each cell controlled to within 100 to ensure query efficiency.

[0120] Reference Figure 3 The geological hazard intelligent identification module 2 based on differential geometry includes a curvature tensor calculation unit 21, an adaptive feature extraction unit 22, a terrain classification unit 23, and a risk assessment unit 24.

[0121] The curvature tensor calculation unit 21 is used to model the terrain surface as a parametric surface, construct a local coordinate system, calculate the first and second basic forms of the surface, generate the curvature tensor matrix, and obtain the principal curvature and principal direction through eigenvalue decomposition.

[0122] In this invention, the terrain surface is represented as a parametric surface. ,in These are the parametric coordinates. For each point on the terrain, a local coordinate system is established. ,in The direction of the normal vector. The first fundamental form coefficient of the surface. , , Second fundamental form coefficient , , The calculation formula is:

[0123] , , ,

[0124] , , ,

[0125] in: Let be the equation of the parametric surface, representing the coordinates of a point in three-dimensional space. For curved surfaces The first-order partial derivative in the direction, i.e. , For curved surfaces The first-order partial derivative in the direction, i.e. , For curved surfaces The second-order partial derivative of the direction, i.e. , The mixed second-order partial derivative of the surface is, i.e. , For curved surfaces The second-order partial derivative of the direction, i.e. , As the unit normal vector, through The calculation yielded, where This represents the vector cross product operation. This represents the vector dot product operation. , , These are the first fundamental form coefficients, used to measure distances and areas on surfaces; , , It is the second fundamental form coefficient, used to measure the curvature of a surface.

[0126] Construct the curvature tensor matrix based on the fundamental formal coefficients. :

[0127] ,

[0128] in: for The curvature tensor matrix is ​​a matrix whose elements are the ratios of the coefficients of the first and second fundamental forms. The curvature tensor describes the degree of curvature of a surface in various directions.

[0129] By performing eigenvalue decomposition on the curvature tensor, the principal curvatures are obtained. , and main direction , Principal curvature represents the degree of curvature of a surface in a specific direction, with the principal directions indicating the directions of maximum and minimum curvature. The formulas for calculating Gaussian curvature K and mean curvature H are:

[0130] , ,

[0131] in: Gaussian curvature, representing the intrinsic curvature of a surface, is expressed in units of 1200 ppm. ; The mean curvature represents the intrinsic curvature of the surface, with units of . ; and represents the principal curvature, and represents the maximum and minimum curvature, respectively, in units of . ; and The corresponding principal direction is a unit vector. Gaussian curvature and mean curvature are important geometric properties of the surface, used for subsequent terrain classification and risk assessment.

[0132] The adaptive feature extraction unit 22 is connected to the curvature tensor calculation unit 21 and is used to calculate the gradient vector field of the curvature field, evaluate the rate of change of curvature, dynamically adjust the sampling radius, and increase the sampling density in the region of high rate of change.

[0133] Gradient vector field of curvature field and These represent the spatial rates of change of Gaussian curvature and mean curvature, respectively. The formula for calculating the rate of change of curvature index (CV) is:

[0134] ,

[0135] in: This is a dimensionless index representing the rate of change of curvature. Gaussian curvature The gradient vector represents The rate of change in space, in units of ; For mean curvature The gradient vector represents The rate of change in space, in units of ; and These represent the magnitude of the gradient vector, i.e., the rate of change.

[0136] Sampling radius The calculation formula is based on the dynamic adjustment according to the rate of curvature change:

[0137] ,

[0138] Where: is the adjusted sampling radius, in meters; The baseline sampling radius is in meters, and is typically set to 10 meters. The adjustment coefficient is dimensionless and has an empirical value of 0.5. This is an index of the rate of change of curvature. In regions with a high rate of change (such as the edge of a steep cliff), the sampling radius decreases, increasing the sampling density; in gentler regions, the sampling radius increases, reducing computational load. Preferably, the sampling radius is controlled between 1 and 20 meters to ensure that detailed features are captured while maintaining computational efficiency.

[0139] The terrain classification unit 23 is connected to the adaptive feature extraction unit 22, and is used to classify the terrain into peak areas, valley areas, saddle points, ridge areas, gully areas and planar areas based on curvature features, and to perform terrain unit segmentation.

[0140] The terrain classification criteria based on curvature features are as follows:

[0141] Peak area: , , ;

[0142] Valley bottom area: , , ;

[0143] Saddle point area: ;

[0144] Ridge region: , ;

[0145] Ravine area: , ;

[0146] Planar area: , ;

[0147] in: For Gaussian curvature, and Principal curvature, The mean curvature. The judgment threshold is , The judgment threshold is 0.1 (unit: 1 / meter). These thresholds are empirical values ​​determined based on a large amount of experimental data and can effectively distinguish different types of terrain features.

[0148] Terrain unit segmentation is achieved through a region growing algorithm, where adjacent points with similar curvature characteristics are grouped into the same terrain unit. The similarity criterion is that the curvature difference is less than a threshold. (Typically set to 10% of the average curvature).

[0149] Risk assessment unit 24 is connected to terrain classification unit 23 to construct risk scoring function, identify cliff areas, landslide risk areas and collapse risk areas, and classify terrain into three levels: safe, caution and danger.

[0150] Risk scoring function The calculation formula is:

[0151] ,

[0152] in: For point The risk score is dimensionless and ranges from [0,1]. For the Gaussian curvature term, according to Gaussian curvature The absolute value is normalized to obtain a dimensionless value; For the mean curvature term, based on the mean curvature The absolute value is normalized to obtain a dimensionless value; This is the slope term, obtained by normalizing the slope angle, and is dimensionless. , , Let be the weighting coefficients, dimensionless, with empirical values ​​of 0.3, 0.3, and 0.4, and satisfying the following conditions: .

[0153] The criteria for identifying hazardous terrain are:

[0154] Cliff area: or Exceeding the threshold (0.5 / meter), and the slope is greater than (60 degrees);

[0155] Landslide risk zone: The curvature characteristics meet a specific pattern (usually the boundary between the valley floor and the steep slope), and the slope is within... arrive Within the range (30~45 degrees);

[0156] Collapse Risk Zone: Curvature Gradient Exceeding the threshold (0.05 / square meter), and located in a raised area ( )

[0157] Based on risk scoring The terrain is divided into safe zones. ), attention area ( ) and danger zone ( This risk assessment method based on differential geometry can capture potential hazardous areas that are difficult to identify using traditional slope analysis, significantly improving the accuracy of risk identification.

[0158] Reference Figure 5 The multi-scale curvature flow analysis of this invention includes scale space construction, curvature flow evolution simulation, regional structure tensor analysis, multi-scale feature extraction, and terrain evolution prediction.

[0159] Scale space construction is used to generate multi-scale representation sequences of terrain. This involves setting an appropriate set of scale parameters and establishing correspondences between feature points at different scales. Multi-scale representation sequences of terrain. Generated by applying Gaussian filters with different scale parameters to the original terrain:

[0160]

[0161] in: The scale parameter is The terrain below is represented in meters; The original terrain is shown in meters. The scale parameter is The Gaussian kernel function is defined as follows: ; This represents the convolution operation, and the calculation formula is: Scale parameter set It is usually set as a base-2 exponential sequence, that is ,in The baseline scale is typically set to 1 meter. In this invention, five scale levels are preferably selected to cover the feature scale range from 1 meter to 16 meters.

[0162] Curvature flow evolution simulation is used to construct discrete curvature flow iterative models, simulating terrain evolution at different time steps and detecting morphological abrupt changes during the evolution process. The update formula for the discrete curvature flow iterative model is:

[0163] ,

[0164] in: For time steps Time point The position vector, containing three-dimensional coordinates The unit is meters; For time steps Time point The position vector, in meters; For point The average curvature at that point, in units of rice; For point The unit normal vector at that point is dimensionless. The time step is in seconds, typically set to 0.01 seconds. This model simulates the evolution of terrain under the influence of mean curvature flow, where convex regions gradually flatten and concave regions gradually fill, eventually tending towards a smooth surface. Critical points in the evolution process are detected by abrupt changes in the rate of curvature change; these critical points typically correspond to unstable regions in the terrain.

[0165] Regional structure tensor analysis is used to construct structure tensors based on local topographic gradients, evaluate the ratios and distributions of tensor eigenvalues, and identify the main orientation and structural lines of the terrain. Regional structure tensor The calculation formula is:

[0166] ,

[0167] in: for The region structure tensor matrix; For the terrain in The gradient in direction, in dimensionless units (height change / horizontal distance). For the terrain in The gradient of the direction, in dimensionless units; the summation range is the local neighborhood (usually 1000 ppm). (A window of meters). Eigenvalues ​​are obtained by performing eigenvalue decomposition on the structure tensor. , ( The anisotropy of terrain is measured by the ratio of eigenvalues ​​to eigenvectors. This indicates that a larger ratio signifies a more pronounced terrain directionality. Terrain stability index. Defined as:

[0168] ,

[0169] in: For point The stability index at the location is dimensionless. The largest eigenvalue of the structure tensor represents the gradient intensity along the principal direction of the terrain. The minimum eigenvalue of the structure tensor represents the gradient strength in the second direction of the terrain. The stability exponent is greater than a threshold. Areas that are typically set to 5 are considered potentially unstable.

[0170] Multi-scale feature extraction is used to identify key terrain features at each scale, construct multi-level representations of features, and track the correspondence between features across different scales. Feature persistence is quantified by its survival time in scale space; features with high persistence typically correspond to salient structures in the terrain, while features with low persistence may be noise or transient changes. Feature persistence The calculation formula is:

[0171] ,

[0172] in: Features Durability, unit and scale parameters Same, usually in meters; Features The scale parameter that first appears in scale space, with the unit being meters; Features The scale parameter that disappears in scale space, measured in meters. Persistence greater than a threshold. Features that are typically set to 50% of the scale range are considered stable features and are used for subsequent risk analysis.

[0173] Topographic evolution prediction is used to simulate the deformation path of terrain under external forces, identify areas prone to deformation during the evolution process, and determine the critical conditions for the terrain to reach an unstable state. By analyzing the deformation rate and direction during curvature flow evolution, the location and scale of potential geological hazards can be predicted. Deformation rate The calculation formula is:

[0174] ,

[0175] in: For point The deformation rate at the point is expressed in meters per second. For time steps Time point The position vector, in meters; For time steps Time point The position vector, in meters; The time step is in seconds. The magnitude of the position vector difference, i.e., the displacement, is expressed in meters. The deformation rate exceeds a threshold. Areas that are typically set at 0.1 meters per time step are identified as rapid deformation zones, corresponding to potential disaster risk points.

[0176] Reference Figure 6 The terrain-environment coupled dynamic risk assessment model of the present invention includes an environmental factor data layer, a coupled representation module, a set of state variables, an evolution rule engine, a numerical solution framework, a dynamic risk calculator, and an early warning and decision support system.

[0177] The environmental factors data layer stores environmental information such as rainfall, temperature changes, and groundwater levels. Environmental data is collected from meteorological stations and groundwater monitoring stations, with a temporal resolution of hourly and a spatial resolution of kilometerly. Preferably, the environmental data is stored in a spatiotemporal database, supporting time-series queries and spatial interpolation.

[0178] The coupling representation module is used to establish the correlation mapping between terrain geometry and environmental factors, defining the influence mechanism of environmental factors on terrain stability. Terrain-Environment Coupling Representation The influence function of environmental factors on terrain stability is constructed by using the tensor product of terrain features and environmental factors. A combination of physical modeling and data-driven approaches was used to construct the model. Taking the impact of rainfall on landslide risk as an example, the impact function can be expressed as:

[0179] ,

[0180] in: For rainfall point The influence function of landslide risk is dimensionless. A terrain point, containing spatial coordinates; For time Rainfall amount, expressed in millimeters per hour; Current time, in hours; This refers to the start time of rainfall, expressed in hours. For point The slope at the location is in degrees. ; This is a proportionality coefficient, with the unit being degrees. / mm, empirical value is 0.05; This is the time decay factor, in hours. The empirical value is 0.1; This is an exponential decay function, indicating that the impact weakens over time. This function shows that the impact of rainfall on landslide risk decreases exponentially over time.

[0181] A set of state variables is used to describe the terrain risk status. State variables include terrain stability index, soil moisture content, rock weathering degree, etc., which collectively determine the terrain risk status. Preferably, the state variables are stored using a grid data structure with a grid resolution of 10 meters to ensure the capture of local risk changes.

[0182] The evolutionary rule engine is used to define the rules governing the changes of state variables over time and environmental conditions. These rules are based on physical models and empirical rules, considering both the direct impact and cumulative effects of environmental factors on terrain stability. Taking the evolution of soil moisture content as an example, its update rule is as follows:

[0183] ,

[0184] in: For time point Soil moisture content at the location, in millimeters; For time point Soil moisture content at the location, in millimeters; This refers to rainfall infiltration, expressed in millimeters per hour. Evaporation rate, expressed in millimeters per hour; This refers to the drainage volume, measured in millimeters per hour, and is related to topographic features and current water content.

[0185] Numerical solution frameworks are used for efficient computation of risk evolution processes. This invention employs an explicit solution method with adaptive time steps, dynamically adjusting the time step based on the rate of change of state variables. Small time steps (e.g., 1 hour) are used in rapid change phases, while large time steps (e.g., 12 hours) are used in slow change phases, balancing computational accuracy and efficiency.

[0186] The dynamic risk calculator is used to calculate a baseline risk value based on terrain geometry, assess the impact of current environmental conditions on risk, and update the risk assessment results in real time. Dynamic Risk Value The calculation formula is:

[0187] ,

[0188] in: For point In time The dynamic risk value is dimensionless and ranges from [value missing]. For point The baseline risk value is based on the terrain geometry and is dimensionless. Let be the dimensionless function of the impact of environmental factors on risk. The start time is in hours; Current time, in hours; Indicates environmental factors within a time interval The cumulative impact of internal risks is realized through numerical integration. Preferably, the integration is achieved using numerical methods, and the time step is dynamically adjusted according to the rate of environmental change.

[0189] The early warning and decision support system triggers different levels of alerts based on risk levels, providing the time window and spatial scope of the risk, and automatically adjusts recommended routes according to changes in risk. There are four alert levels: blue (low risk), yellow (low to medium risk), orange (medium to high risk), and red (high risk), with corresponding risk thresholds of 0.3, 0.5, 0.7, and 0.9, respectively. When the risk value of a certain area exceeds the corresponding threshold, the system will trigger an alert of the appropriate level, providing the scope of the risk area and the expected duration. For planned routes, the system will automatically calculate alternative routes based on changes in risk to ensure user safety.

[0190] Reference Figure 7 The route planning and safety assessment module 3 includes a user data collection unit 31, a behavior feature analysis unit 32, a route planning unit 33, a route scoring unit 34, and a personalized recommendation unit 35.

[0191] User data collection unit 31 is used to collect physical fitness data such as user age, gender, physical fitness level, and luggage weight. User data is input through a mobile application interface or automatically obtained from a smart wearable device. Preferably, physical fitness level adopts a 5-level evaluation system (from beginner to professional level), and luggage weight is accurate to the kilogram level.

[0192] The behavior feature analysis unit 32 is connected to the user data acquisition unit 31 and is used to analyze user behavior, including path planning, speed, and heart rate prediction. It assesses user behavior based on historical route data and real-time data from the smart wearable device. The behavior feature analysis employs machine learning methods, analyzing historical route data to predict changes in the user's walking speed and heart rate under different terrain conditions. Taking speed prediction as an example, the prediction model is as follows:

[0193] ,

[0194] in: For users at point Predicted speed at the location, in kilometers per hour; The reference speed is measured in kilometers per hour. It is the normal walking speed on flat ground without any load, which is usually 4-5 kilometers per hour. The slope influence function is dimensionless. For point The slope at the location is expressed in degrees (°). This is a dimensionless function representing the influence of terrain type. For point The terrain type (e.g., flat land, gravel, grassland, etc.); The fatigue effect function is dimensionless and has a distance from the previously calculated distance. (Unit: kilometers) and time (Unit: hours) Related. Each influence function is calibrated using historical user data to improve prediction accuracy.

[0195] The route planning unit 33 is connected to the behavior feature analysis unit 32 and the geological hazard intelligent identification module 2 based on differential geometry. It is used to generate multiple alternative safe routes based on the user's physical fitness level, behavioral characteristics, and terrain risk distribution. The route planning employs an improved... The algorithm incorporates risk factors into the path cost function. Path cost function The calculation formula is:

[0196] ,

[0197] in: From point Time The path cost, in meters; The Euclidean distance between two points is expressed in meters. For point In time The dynamic risk value is dimensionless. For point The slope, in degrees (°); From point Time The energy consumption estimate is given in joules. , and The weights are dimensionless, with empirical values ​​of 2.0, 0.5, and 0.3. Preferably, the route planning considers multiple target points, such as rest stops and viewpoints, and generates a route that meets user needs through a multi-objective path planning algorithm.

[0198] Route scoring unit 34 is connected to route planning unit 33 and is used to comprehensively score each candidate route, considering factors such as safety, difficulty, and suitability. Route scoring function. The calculation formula is:

[0199] ,

[0200] in: The overall score for the route is dimensionless and ranges from [0, 100]. Safety scores are given, dimensionless, and based on the risk distribution along the route; The difficulty score is dimensionless and takes into account changes in slope, distance, and cumulative height. The fitness score is dimensionless and represents the degree to which the route matches the user's physical fitness and preferences. , and Let be the weighting coefficients, dimensionless, with empirical values ​​of 0.5, 0.3, and 0.2, respectively, and satisfying the following conditions: The scoring results use a standardized score of 0-100, with higher scores indicating better routes.

[0201] Personalized recommendation unit 35 is connected to route scoring unit 34, and is used to recommend suitable routes to different travel groups based on the scoring results. The recommendation strategy considers the user's physical fitness level, experience, and preferences, recommending safe routes for beginners, moderately difficult routes for experienced users, and challenging routes for professional users. Preferably, the recommendation system uses a collaborative filtering algorithm to improve recommendation accuracy by utilizing evaluation data from similar users.

[0202] Reference Figure 8 The virtual roaming system includes a 3D scene rendering unit, a risk area marking unit, a route visualization unit, a virtual roaming control unit, and a scene information prompting unit.

[0203] The 3D scene rendering unit is used to generate interactive 3D scenes based on a 3D virtual terrain model. This invention employs real-time rendering technology based on a graphics engine, supporting high-resolution textures, dynamic lighting, and environmental effects to provide a realistic visual experience. Preferably, the rendering uses Level of Detail (LOD) technology, dynamically adjusting model details according to viewpoint distance to balance visual quality and rendering efficiency.

[0204] The risk area labeling unit is connected to the differential geometry-based intelligent geological hazard identification module 2, used to visually display areas with different risk levels in a 3D scene. Risk areas are displayed using color coding: green indicates a safe zone, yellow indicates a warning zone, and red indicates a danger zone. Preferably, the transparency of the risk areas is dynamically adjusted according to the risk level, with lower transparency in high-risk areas to make the risk markings more prominent.

[0205] The route visualization unit is connected to the route planning and safety assessment module 3, and is used to display the planned route in a 3D scene. The route is represented by colored lines, with different colors indicating different difficulty levels. Preferably, mileage markers and altitude markers are set on the route to visually display distance and elevation changes.

[0206] The virtual roaming control unit provides multi-view, multi-angle route preview functionality, allowing users to explore routes in a virtual environment. Users can choose a first-person perspective (immersive experience) or a third-person perspective (global observation), freely adjusting the viewpoint position and observation angle. Preferably, an automatic roaming function is provided, automatically moving the viewpoint according to the planned route, allowing users to intuitively experience the characteristics and difficulty of the route.

[0207] The scene information prompt unit provides information such as terrain features, risk areas, and route difficulty during virtual roaming. The prompts use a combination of icons and text, displaying risk warnings, difficulty changes, and recommended rest stops at key locations. Preferably, the prompts are dynamically adjusted based on the user's experience level, providing more detailed prompts for beginners and concise key information for experienced users.

[0208] Reference Figure 8 The route generation method of the route planning and safety assessment module 3 of the present invention includes safe path search, user physical fitness adaptation, multi-route generation, route comprehensive scoring, dynamic route adjustment, segmented route management and emergency avoidance planning.

[0209] Safe path search is used to calculate the safest path based on a risk distribution map. This invention employs an improved Dijkstra's algorithm, incorporating risk factors into the edge weight calculation. Edge weights The calculation formula is:

[0210] ,

[0211] in: For the edge The weight, in meters; For the edge The physical length, in meters; For the edge The average risk value, dimensionless, with a range of values ​​of [value missing]. Here, is the risk weight coefficient, dimensionless, with an empirical value of 2.0. The algorithm finds the path with the lowest risk by minimizing the total path weight. Preferably, the algorithm considers multiple risk thresholds to ensure that the path does not pass through high-risk areas.

[0212] User fitness adaptation is used to adjust the route difficulty based on the user's fitness data. Difficulty adjustment considers gradient limits, route length, and cumulative elevation change. For beginners, the system limits the maximum gradient to within 20 degrees, the route length to within 5 kilometers, and the cumulative elevation change to within 300 meters; for experienced users, these limits are relaxed to 30 degrees, 10 kilometers, and 600 meters respectively; for professional users, the limits are further relaxed to 40 degrees, 20 kilometers, and 1000 meters. Preferably, the system dynamically adjusts these limits based on the user's actual performance to improve adaptation accuracy.

[0213] Multi-route generation provides multiple alternative routes that meet safety requirements. This invention employs the K-shortest path algorithm to generate several routes with different characteristics for the user to choose from. Route diversity is achieved by controlling path overlap, ensuring that the overlap between alternative routes does not exceed 30%. Preferably, the system generates characteristic tags for each route, such as safest route, shortest route, best scenic route, etc., to help users quickly select a route that meets their needs.

[0214] The route comprehensive score is used to rate the safety, difficulty, and suitability of each route. The scoring system considers multiple factors, including the maximum risk value, average risk value, maximum gradient, average gradient, total length, cumulative altitude change, and match with user preferences. The score results are standardized from 0 to 100, with the weights of each factor adjusted according to the user's fitness level. Preferably, the system displays radar charts for each scoring dimension to visually illustrate the characteristics of the route.

[0215] Dynamic route adjustment is used to modify route suggestions in real time when environmental conditions change. The system monitors weather changes, terrain changes, and the user's real-time location. When an increase in risk is detected, it automatically calculates alternative routes or adjusts sections of the original route. The adjustment strategy is based on the risk increment and the user's location, minimizing changes to the original route while ensuring safety. Preferably, the system pre-calculates multiple alternatives to respond quickly to changes in risk.

[0216] Segmented route management is used to divide routes into segments, providing detailed difficulty and risk information for each segment. Route segmentation is based on terrain features and difficulty variation points, with each segment's length controlled between 0.5 and 2 kilometers. Each segment has an independent difficulty rating, risk level, and estimated travel time. Preferably, the system provides detailed terrain descriptions and precautions for each segment to help users prepare adequately.

[0217] Emergency evacuation planning is used to provide emergency evacuation route suggestions when a sudden risk is detected. Emergency evacuation routes prioritize safety, seeking the nearest safe area as the evacuation target. The evacuation route generation algorithm employs a greedy strategy, choosing the direction with the lowest risk at each step. Preferably, the system pre-calculates the distribution of safe points within the area, enabling it to quickly find the nearest safe point in an emergency.

[0218] Reference Figure 9 The data flow between the modules of the system of this invention includes multi-source data flow, terrain vector data flow, risk distribution data flow, user data flow, route data flow, environmental data flow, and early warning information flow.

[0219] The multi-source data stream transmits raw data from external data sources to the terrain digital modeling module 1. Data sources include UAV aerial photography platforms, satellite remote sensing data centers, and ground surveying equipment. Data formats include image files (TIFF, JPEG), elevation data (DEM), and control point files (CSV). The data transmission employs a data pipeline architecture, supporting incremental updates and breakpoint resumption.

[0220] The terrain vector data stream is a 3D terrain data transmission process from the terrain digital modeling module 1 to the differential geometry-based intelligent geological hazard identification module 2. The data includes a 3D terrain vector model, texture mapping information, and geographic feature point data, in a custom terrain vector data structure. Data transmission utilizes shared memory to improve the efficiency of transmitting large-scale data.

[0221] The risk distribution data stream transmits risk assessment results from the differential geometry-based intelligent geological hazard identification module 2 to the route planning and safety assessment module 3. The data includes risk distribution maps, terrain classification results, and dynamic risk predictions, in spatial grid format. Data transmission uses a publish-subscribe model to ensure real-time updates of risk information.

[0222] User data flows from the user terminal to the route planning and safety assessment module 3, transmitting user information. The data includes user physical fitness data, behavioral characteristics, and route preferences, in JSON or XML format. Data transmission uses an encrypted channel to protect user privacy.

[0223] The route data stream transmits route information from the route planning and safety assessment module 3 to the virtual roaming system. The data includes route geometry, segmentation information, and difficulty rating, in GeoJSON or KML format. Data transmission uses the WebSocket protocol, supporting real-time updates and interaction.

[0224] The environmental data stream transmits real-time environmental data from the external environmental monitoring system to the topographic-environment coupled dynamic risk assessment model. The data includes meteorological data, groundwater levels, and soil moisture, in time-series format. Data transmission employs either timed polling or event-triggered methods, dynamically adjusting the update frequency based on the rate of environmental change.

[0225] The early warning information stream pushes risk warning information from the system to user terminals. The data includes the risk level, scope of impact, and expected duration, in a structured message format. Data transmission uses a push service to ensure users receive warning information promptly. Preferably, the system selects different push channels based on the warning level to ensure that high-risk warnings are delivered first.

[0226] Through the detailed embodiments described above, this invention provides a three-dimensional terrain reconstruction and route safety assessment system for outdoor sports based on differential geometry theory. This system achieves high-precision three-dimensional reconstruction of complex terrain, accurate geological hazard identification, and personalized safe route planning. The system possesses strong practicality and reliability, effectively improving the safety factor of outdoor sports and has broad application prospects.

[0227] The embodiments described above are merely illustrative of specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. An outdoor sports terrain 3D reconstruction and route safety assessment system, characterized in that, include: The terrain digital modeling module is used for: Receive multi-source image data, including optical remote sensing images, aerial images, digital elevation models, and ground control point data; Feature extraction, registration, and fusion processing are performed on the multi-source image data to generate three-dimensional terrain vector data; Construct a 3D surface representation based on a graphics engine to form a 3D virtual surface model with a geographic coordinate system; A geological hazard intelligent identification module based on differential geometry, connected to the terrain digital modeling module, is used for: Receive the three-dimensional terrain vector data; Construct a terrain curvature tensor field and calculate the principal curvature and principal direction of the terrain surface; Adaptive extraction of terrain hazard features based on curvature tensor features; Assess terrain stability through multi-scale curvature flow analysis; Construct a dynamic risk assessment model that combines topography and environment by incorporating environmental factors; Generate a terrain risk distribution map with risk level labels; The route planning and safety assessment module, connected to the differential geometry-based intelligent geological hazard identification module, is used for: Receive user physical fitness data and behavioral characteristics information; Receive the terrain risk distribution map; Based on the user's physical fitness data, behavioral characteristics, and terrain risk distribution map, a personalized safety route is generated. The personalized safety route and risk areas along the route are displayed through a virtual roaming system; Adjust route suggestions and risk warnings in real time when environmental conditions change.

2. The system according to claim 1, characterized in that, The terrain digital modeling module includes: The image acquisition unit is used to acquire aerial images, optical remote sensing images, and ground control point data; A data preprocessing unit, connected to the image acquisition unit, is used to perform geometric and radiometric corrections on the aerial images and optical remote sensing images; A feature extraction unit, connected to the data preprocessing unit, is used to extract key geographic location features from the corrected image, including steep slopes, rock surfaces, and vegetation areas. A stereo matching unit, connected to the feature extraction unit, is used to calculate the shape, size, orientation, and texture of the extracted features, and to associate and map features from different images. A three-dimensional reconstruction unit, connected to the stereo matching unit, is used to calculate the coordinates of feature points in three-dimensional space and generate three-dimensional vector data; The texture mapping unit, connected to the three-dimensional reconstruction unit, is used to apply high-resolution remote sensing images onto the vector terrain model to generate a complete three-dimensional virtual surface model.

3. The system according to claim 1, characterized in that, The storage structure for the three-dimensional terrain vector data includes: Geographic coordinate system information, used for location positioning, wherein the geographic coordinate system information is consistent with the geographic coordinate system of the remote sensing image; Raster image data is used to store terrain shapes for easy searching and retrieval; Vector feature point data is used to store geographic feature points, which have coordinates, height, and terrain feature labels. The feature relationship index is used to establish the correspondence between feature points and terrain vectors, enabling attribute queries of geographic feature points and visualization of terrain features.

4. The system according to claim 1, characterized in that, The geological hazard intelligent identification module based on differential geometry includes: The curvature tensor calculation unit is used to model the terrain surface as a parametric surface, construct a local coordinate system, calculate the first and second fundamental form coefficients of the surface, generate the curvature tensor matrix, and obtain the principal curvature and principal direction through eigenvalue decomposition. The first fundamental form coefficient is used to measure the distance and area on the surface, and the second fundamental form coefficient is used to measure the curvature of the surface. An adaptive feature extraction unit, connected to the curvature tensor calculation unit, is used to calculate the gradient vector field of the curvature field, evaluate the rate of curvature change, dynamically adjust the sampling radius, and increase the sampling density in regions with high rate of change. The terrain classification unit, connected to the adaptive feature extraction unit, is used to classify the terrain into peak regions, valley regions, saddle point regions, ridge regions, gully regions, and planar regions based on curvature features, and to perform terrain unit segmentation. The risk assessment unit, connected to the terrain classification unit, is used to construct a risk scoring function, identify cliff areas, landslide risk areas and collapse risk areas, and classify the terrain into three levels: safe, caution, and dangerous.

5. The system according to claim 1, characterized in that, The multi-scale curvature flow analysis includes: Scale space construction is used to generate multi-scale representation sequences of terrain, set an appropriate set of scale parameters, and establish the correspondence between feature points at different scales. Curvature flow evolution simulation is used to construct discrete curvature flow iterative models, simulate the terrain evolution state at different time steps, and detect morphological abrupt changes during the evolution process; Regional structure tensor analysis is used to construct a structure tensor based on local terrain gradients, evaluate the ratio and distribution of tensor eigenvalues, and identify the main orientation and structural lines of the terrain. Multi-scale feature extraction is used to identify key terrain features at each scale, construct multi-level representations of features, and track the correspondence between features at different scales; Terrain evolution prediction is used to simulate the deformation path of terrain under the action of external forces, identify areas that are prone to deformation during the evolution process, and determine the critical conditions for the terrain to reach an unstable state.

6. The system according to claim 1, characterized in that, The terrain-environment coupled dynamic risk assessment model includes: The environmental factors data layer is used to store environmental information such as rainfall, temperature changes, and groundwater levels. The coupling representation module is used to establish the correlation mapping between terrain geometry and environmental factors, and to define the influence mechanism of environmental factors on terrain stability. A set of state variables used to describe the terrain risk status; Evolutionary rule engine, used to define rules for how state variables change over time and environmental conditions; A numerical solution framework for efficiently calculating risk evolution processes; The dynamic risk calculator is used to calculate a baseline risk value based on terrain geometry, assess the impact of current environmental conditions on risk, and update the risk assessment results in real time. The early warning and decision support system is used to trigger different levels of early warning based on the risk level, provide the time window and spatial range of the risk, and automatically adjust the recommended route according to the changes in risk.

7. The system according to claim 1, characterized in that, The route planning and safety assessment module includes: The user data collection unit is used to collect user age, gender, physical fitness status, and luggage weight data. The behavior feature analysis unit is connected to the user data acquisition unit and is used to analyze user behavior path planning, speed and heart rate prediction, and to evaluate user behavior based on user historical route data and real-time data from smart wearable devices. The route planning unit, connected to the behavior feature analysis unit and the geological hazard intelligent identification module based on differential geometry, is used to generate multiple alternative safe routes based on the user's physical fitness level, behavior characteristics and terrain risk distribution. The route scoring unit, connected to the route planning unit, is used to comprehensively score each candidate route, taking into account factors such as safety, difficulty, and suitability. A personalized recommendation unit, connected to the route scoring unit, is used to recommend suitable routes to different travel groups based on the scoring results.

8. The system according to claim 1, characterized in that, The virtual roaming system includes: The 3D scene rendering unit is used to generate interactive 3D scenes based on a 3D virtual terrain model. The risk area labeling unit is connected to the geological hazard intelligent identification module based on differential geometry, and is used to intuitively display areas with different risk levels in a three-dimensional scene; The route visualization unit, connected to the route planning and safety assessment module, is used to display the planned route in a three-dimensional scene; The virtual roaming control unit provides multi-view, multi-angle route preview functionality, allowing users to explore routes in a virtual environment; The scene information prompting unit is used to provide prompts on terrain features, risk areas, and route difficulty during virtual roaming.

9. The system according to claim 1, characterized in that, The route generation method of the route planning and safety assessment module includes: Safe path search is used to calculate the safest path based on a risk distribution map; User fitness adaptation is used to adjust the difficulty of the route based on the user's fitness data; Multi-route generation is used to provide multiple alternative routes that meet safety requirements; The overall route score is used to rate the safety, difficulty, and suitability of each route. Dynamic route adjustment is used to adjust route suggestions in real time when environmental conditions change; Segmented route management is used to divide a route into segments and provide detailed information on the difficulty and risks of each segment. Emergency evacuation planning is used to provide recommendations for emergency evacuation routes when a sudden risk is detected.

10. The system according to claim 1, characterized in that, Data flow between modules of the system includes: Multi-source data streams, from raw data transmission from external data sources to the terrain digital modeling module; Terrain vector data stream, from the terrain digital modeling module to the three-dimensional terrain data transmission module based on differential geometry for intelligent identification of geological hazards; Risk distribution data stream, from the geological hazard intelligent identification module based on differential geometry to the risk assessment results transmission of the route planning and safety assessment module; User data stream, the transmission of user information from the user terminal to the route planning and safety assessment module; Route data flow, from the route planning and safety assessment module to the route information transmission of the virtual roaming system; Environmental data streams, from external environmental monitoring systems to real-time environmental data transmission to topographic-environment coupled dynamic risk assessment models; Early warning information flow, pushing risk warning information from the system to user terminals.

Images