A topographic mapping method and system based on artificial intelligence

Abstract

The application discloses a topographic surveying and mapping method based on artificial intelligence, which comprises the following steps: (1) multi-source data fusion collection; (2) intelligent extraction of topographic features; (3) three-dimensional reconstruction optimization; and (4) dynamic topographic evolution modeling. The application also discloses a system for realizing the topographic surveying and mapping method based on artificial intelligence, which comprises a data collection and preprocessing module, an intelligent extraction of topographic features module, a three-dimensional reconstruction optimization module, a dynamic topographic evolution modeling module and an interactive visualization and decision-making platform. The data collection and preprocessing module, the intelligent extraction of topographic features module, the three-dimensional reconstruction optimization module, the dynamic topographic evolution modeling module and the interactive visualization and decision-making platform are sequentially connected, the modules communicate through a gRPC interface and support independent upgrading.

Description

Technical Field

[0001] The present invention relates to the technical field of terrain surveying and mapping, and more particularly, to a terrain surveying and mapping method and system based on artificial intelligence. Background Art

[0002] With the advent of photography technology, photogrammetry began to be used in topographic mapping. Through photos taken from the air or on the ground, three-dimensional models were created and topographic information was extracted. In the process of carrying out engineering construction, topographic mapping is the preliminary preparation work. The measurement and collection of relevant data of geographic space can make the relationship between buildings and nature more harmonious, and make the information of artificial buildings clearer in the process of standardizing the geographic space location. The spatial location is marked with numbers, names and distances, and the information of the construction project is clearly marked. The distance, place names and various attributes are divided so that the geographic space location can be clearly displayed on the topographic map, which is convenient for the implementation of engineering construction. At present, the traditional topographic mapping method has the following technical bottlenecks:

[0003] Low data fusion efficiency: The temporal and spatial benchmarks of multi-source sensor data are inconsistent, and traditional ICP registration is time-consuming and inaccurate.

[0004] Coarse feature extraction granularity: Networks such as U-Net have difficulty identifying sub-meter terrain features (such as ditches and cracks) due to the loss of feature map resolution (1 / 32);

[0005] Lack of dynamic prediction: Existing 3D reconstruction methods (such as oblique photography) lack physical constraints and cannot simulate the geomechanical evolution process;

[0006] Decision support lags: Traditional numerical simulation calculations are time-consuming and cannot meet real-time warning needs. Summary of the Invention

[0007] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a terrain mapping method and system based on artificial intelligence.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A terrain surveying and mapping method based on artificial intelligence comprises the following steps:

[0010] (1) Multi-source data fusion acquisition: Use multiple devices to collect data sources of the terrain to be mapped, and use artificial intelligence to perform multimodal data registration on the collected data sources to suppress noise and enhance data;

[0011] (2) Intelligent extraction of terrain features: The terrain features of the collected and fused data sources are extracted. The backbone network uses the HRNet+ architecture to maintain high-resolution features. In the task branch, semantic segmentation uses DeepLabv3+ to extract information such as water systems, vegetation, and artificial buildings. Instance segmentation uses Mask R-CNN to identify independent terrain units. Key point detection uses SuperPoint to locate feature markers. Slope calculation and curvature analysis are also performed on the data source.

[0012] (3) 3D reconstruction optimization: The data source after feature extraction is added to the neural radiation field model to obtain a 3D reconstruction model based on spatial image features, and the error of the formed 3D reconstruction model is corrected;

[0013] (4) Dynamic terrain evolution modeling: Construct a multi-mechanism coupled intelligent evolution model that integrates time and space, integrates physical constraint neural networks, multi-factor graph coupling systems, and adversarial generation prediction frameworks, and realizes multi-scale dynamic simulation and interactive inversion of terrain evolution through online learning and mixed reality deduction platforms to meet the needs of geological disaster warning and long-term strategic planning.

[0014] Furthermore, the specific method of using multiple devices to collect data sources for the terrain to be surveyed in step (1) includes deploying a stereo observation network, integrating satellite multispectral images, UAV LiDAR point clouds, oblique photography data from ground mobile mapping systems and geological radar profiles, and constructing a multi-scale data set from centimeters to kilometers.

[0015] Furthermore, the multimodal data registration in step (1) includes adopting a Transformer-based cross-modal registration network to solve the spatiotemporal reference difference through a deformable attention mechanism to achieve high-precision spatial matching of optical images and point cloud data.

[0016] Furthermore, the specific method for suppressing noise and enhancing data in step (1) includes constructing a noise-aware generative adversarial network, separating terrain signals from sensor noise through a spectral normalization discriminator, using the StyleGAN2-ADA algorithm to perform texture enhancement on low-quality areas, and improving the resolution of weak feature areas by more than 4 times. At the same time, a lightweight edge computing module is developed to fuse multi-source data streams in real time and output a standardized terrain feature tensor to support end-to-end feature transfer for subsequent terrain extraction tasks.

[0017] Furthermore, the HRNet+ high-resolution feature preservation in step (2) adopts an improved high-resolution network to maintain spatial accuracy through multi-resolution parallel processing and enhanced feature fusion.

[0018] Furthermore, the multi-resolution parallel processing constructs a resolution sequence of four branches for the input image size H×W, and the formula Compute the feature exchange performed at each stage, where It is a cross-resolution feature transformation operator, including 1×1 convolution and bilinear interpolation, i=1,2,3,4.

[0019] Furthermore, the enhanced feature fusion introduces the coordinate attention module (CA) to optimize feature selection. The processing result of the feature map by the coordinate attention module is expressed as follows: Calculated, where GAP is global average pooling, GMP is global maximum pooling, and σ is Sigmoid activation.

[0020] Furthermore, the specific method of 3D model reconstruction and optimization in step (3) is to adopt feature-enhanced NeRF modeling, fuse feature encoding with spatial coordinates, predict density and color through MLP, generate predicted images using differentiable rendering, jointly optimize reconstruction and feature consistency, and optimize the image quality based on geometric constraints and multi-viewing. Figure 1 Consistency, iteratively optimize model details, at the same time, prune low-density areas, reduce floating noise, convert the point cloud output by NeRF into an explicit mesh, fill in missing patches, force the optical flow of adjacent perspective rendering results to be consistent with the real offset, and introduce a discriminator network to distinguish high-frequency details between real and rendered images.

[0021] Furthermore, the dynamic terrain evolution modeling in step (4) specifically includes using a spatiotemporal tensor data structure and a Transformer architecture to process historical and real-time data, capturing terrain changes at different time scales, combining numerical simulation and a data-driven dual-branch network, ensuring that the prediction complies with physical laws through a differentiable physical layer, utilizing multi-agent simulation and graph neural networks to process the interaction of factors such as erosion and sedimentation, using GAN for multi-resolution prediction, and updating the model through online learning, and developing a visualization platform to support multiple simulations and real-time adjustments.

[0022] Furthermore, an artificial intelligence-based terrain mapping system is provided for implementing an artificial intelligence-based terrain mapping method, the system comprising:

[0023] Data acquisition and preprocessing module, including a multi-sensor control unit, a cross-modal registration engine, a noise suppression enhancer, and an edge computing node, for multi-source heterogeneous data acquisition and intelligent fusion;

[0024] The terrain feature intelligent extraction module includes a high-resolution feature backbone, a semantic segmentation unit, an instance segmentation unit, a key point detection unit, and a terrain parameter calculator for multi-dimensional terrain feature analysis;

[0025] 3D reconstruction optimization module, including neural radiation field core, error correction unit and LOD generator, for high-precision 3D modeling and correction;

[0026] Dynamic terrain evolution modeling module, including a physical-data dual-driven engine, a multi-factor coupling system, an adversarial prediction framework, and an online learning interface, for temporally and spatially coupled terrain evolution prediction;

[0027] Interactive visualization and decision-making platform, including a mixed reality engine, dynamic simulation sandbox, and intelligent decision-making assistance, for multi-dimensional data fusion display and simulation;

[0028] The data acquisition and preprocessing module, terrain feature intelligent extraction module, three-dimensional reconstruction optimization module, dynamic terrain evolution modeling module and interactive visualization and decision-making platform are connected in sequence. Each module communicates through the gRPC interface and supports independent upgrades.

[0029] In summary, this application includes at least one of the following beneficial technical effects:

[0030] This method breaks through the limitations of traditional single data sources through cross-modal intelligent registration and noise suppression mechanisms. At the same time, the spatial-spectral joint registration framework based on deep features effectively overcomes the temporal and spatial reference differences and sensor noise interference of multi-source data, significantly improving data consistency in complex terrain. The dynamic adaptive enhancement of this method uses spectral domain filtering and texture reconstruction technology of generative adversarial networks to achieve autonomous optimization of data quality. It is particularly suitable for feature recovery in harsh acquisition environments such as cloud obscuration and multiple echoes. The system achieves holographic analysis of terrain elements through a multi-task collaborative deep learning architecture. The advanced HRNet+ backbone network retains sub-meter terrain details (such as water tributaries and engineering structure joints) to the greatest extent through multi-scale parallel convolution and feature reconstruction mechanisms, overcoming the problem of small target information loss caused by downsampling in traditional networks.

[0031] This system implements a physics-data-driven approach, embedding terrain parameter calculations (slope / curvature) into the feature extraction process. This system enhances the model's geoscientific interpretability through physical prior constraints, improving the reliability of identifying steeply changing regions and complex geometric structures. Through the gradient interaction between the numerical simulation branch and the data-driven branch, a deep fusion of physical laws and observational data is achieved, ensuring that predictions are both consistent with geomechanical principles and adaptable to complex environmental evolution. Furthermore, the model update architecture, based on incremental learning, dynamically integrates new sensor data and field monitoring feedback, continuously improving the system's generalization performance in unknown terrain scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A flowchart of a terrain surveying and mapping method based on artificial intelligence in the present invention;

[0033] Figure 2 This is a system block diagram of an artificial intelligence-based terrain surveying and mapping system in the present invention. DETAILED DESCRIPTION

[0034] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0035] It should be noted that, unless otherwise specified, all technical and scientific terms used in this application have the same meaning as commonly understood by ordinary technicians in the technical field to which this application belongs.

[0036] According to one embodiment of the present invention, Figure 1 As shown, a terrain mapping method based on artificial intelligence includes the following steps:

[0037] (1) Multi-source data fusion acquisition: Use multiple devices to collect data sources of the terrain to be mapped, and use artificial intelligence to perform multimodal data registration on the collected data sources to suppress noise and enhance data;

[0038] (2) Intelligent extraction of terrain features: The terrain features of the collected and fused data sources are extracted. The backbone network uses the HRNet+ architecture to maintain high-resolution features. In the task branch, semantic segmentation uses DeepLabv3+ to extract information such as water systems, vegetation, and artificial buildings. Instance segmentation uses Mask R-CNN to identify independent terrain units. Key point detection uses SuperPoint to locate feature markers. Slope calculation and curvature analysis are also performed on the data source.

[0039] (3) 3D reconstruction optimization: The data source after feature extraction is added to the neural radiation field model to obtain a 3D reconstruction model based on spatial image features, and the error of the formed 3D reconstruction model is corrected;

[0040] (4) Dynamic terrain evolution modeling: Construct a multi-mechanism coupled intelligent evolution model that integrates time and space, integrates physical constraint neural networks, multi-factor graph coupling systems, and adversarial generation prediction frameworks, and realizes multi-scale dynamic simulation and interactive inversion of terrain evolution through online learning and mixed reality deduction platforms to meet the needs of geological disaster warning and long-term strategic planning.

[0041] The specific method of using multiple devices to collect data sources for the terrain to be surveyed in step (1) includes deploying a stereo observation network, integrating satellite multispectral images, UAV LiDAR point clouds, oblique photography data of ground mobile mapping systems and geological radar profiles, and constructing a multi-scale data set from centimeters to kilometers; multimodal data alignment includes adopting a Transformer-based cross-modal alignment network, solving the temporal and spatial reference differences through a deformable attention mechanism, and achieving high-precision spatial matching of optical images and point cloud data; the specific method of suppressing noise and enhancing data includes constructing a noise-aware generative adversarial network, separating terrain signals from sensor noise through a spectral normalization discriminator, and using the StyleGAN2-ADA algorithm to enhance the texture of low-quality areas, thereby improving the resolution of weak feature areas by more than 4 times. At the same time, a lightweight edge computing module is developed to fuse multi-source data streams in real time and output standardized terrain feature tensors to support end-to-end feature transfer for subsequent terrain extraction tasks.

[0042] The HRNet+ high-resolution feature preservation in step (2) adopts an improved high-resolution network to maintain spatial accuracy through multi-resolution parallel processing and enhanced feature fusion; the multi-resolution parallel processing is to construct a resolution sequence of four branches for the input image size H×W, and the formula is used. Compute the feature exchange performed at each stage, where is a cross-resolution feature transformation operator, including 1×1 convolution and bilinear interpolation, i=1,2,3,4; the enhanced feature fusion introduces the coordinate attention module (CA) to optimize feature selection, and the processing result of the feature map by the coordinate attention module is calculated by the formula Calculated, where GAP is global average pooling, GMP is global maximum pooling, and σ is Sigmoid activation.

[0043] The specific method of 3D model reconstruction and optimization in step (3) is to adopt feature-enhanced NeRF modeling, fuse feature encoding with spatial coordinates, predict density and color through MLP, generate predicted images using differentiable rendering, jointly optimize reconstruction and feature consistency, and optimize the image quality based on geometric constraints and multi-viewing. Figure 1 Consistency, iteratively optimize model details, at the same time, prune low-density areas, reduce floating noise, convert the NeRF output point cloud into an explicit mesh, fill in missing patches, force the optical flow of adjacent view rendering results to be consistent with the real offset, and introduce a discriminator network to distinguish high-frequency details of real and rendered images.

[0044] The dynamic terrain evolution modeling in step (4) specifically includes using spatiotemporal tensor data structures and Transformer architecture to process historical and real-time data, capturing terrain changes at different time scales, combining numerical simulation and data-driven dual-branch networks, ensuring that predictions conform to physical laws through a differentiable physical layer, using multi-agent simulation and graph neural networks to handle the interactions of factors such as erosion and sedimentation, using GAN for multi-resolution predictions, and updating models through online learning, and developing a visualization platform to support multiple simulations and real-time adjustments.

[0045] According to another embodiment of the present invention, Figure 2 As shown, an artificial intelligence-based terrain mapping system is used to implement an artificial intelligence-based terrain mapping method. The system includes a data acquisition and preprocessing module 1 for multi-source heterogeneous data acquisition and intelligent fusion, a terrain feature intelligent extraction module 2 for multi-dimensional terrain element analysis, a three-dimensional reconstruction optimization module 3 for high-precision three-dimensional modeling and correction, a dynamic terrain evolution modeling module 4 for spatiotemporal coupled terrain evolution prediction, and an interactive visualization and decision-making platform 5 for multi-dimensional data fusion display and deduction. The modules are connected in sequence and communicate through the gRPC interface to support independent upgrades.

[0046] The data acquisition and preprocessing module 1 includes:

[0047] The multi-sensor control unit 11 includes integrated satellites, UAV LiDAR, ground MMS, geological radar and other equipment, and realizes hardware synchronization through the ROS robot operating system, with a timestamp alignment accuracy of <1ms.

[0048] The cross-modal registration engine 12 is based on the Transformer CMR-Net network and supports optical image, LiDAR and geological profile data registration, with a registration error RMSE of less than 0.3 pixels.

[0049] The noise suppression enhancer 13 uses the NA-GAN adversarial network to denoise the LiDAR point cloud and remove the clouding from the image.

[0050] The edge computing node 14 deploys a real-time fusion pipeline accelerated by TensorRT through the NVIDIA Jetson AGX Xavier to output a standardized terrain tensor.

[0051] The terrain feature intelligent extraction module 2 includes:

[0052] The high-resolution feature backbone 21 is based on the improved HRNet+ architecture, outputs feature maps with 1 / 4 the original image resolution, and supports a 4-level feature pyramid.

[0053] The semantic segmentation unit 22 is based on the improved DeepLabv3+, integrates the terrain prior knowledge base, and can identify water system boundaries with an accuracy of 0.5m.

[0054] The instance segmentation unit 23 is based on the terrain-adapted version of Mask R-CNN and supports instance segmentation of landslide boundaries.

[0055] The key point detection unit 24 is based on the optimized version of SuperPoint and integrates the slope constraint.

[0056] Terrain Parameter Calculator 25 is based on CUDA-accelerated parallel gradient calculation, supporting 10km 2 The slope / curvature of the DEM is generated in seconds.

[0057] The three-dimensional reconstruction optimization module 3 includes:

[0058] Neural Radiation Field Core 31 is based on the improved NeRF architecture and integrates HRNet+ features as position encoding input, improving training efficiency by 40%.

[0059] The error correction unit 32 is based on ICP (Iterative Closest Point) and GAN joint optimization:

[0060]

[0061] Among them, CD is Chamfer Distance, λ_photo=1.0, λ_geo=0.6.

[0062] LOD Generator 33 automatically generates LOD1~LOD4 multi-level detail models with a compression ratio of 20:1. Among them, the LOD4 model only requires 5MB / km².

[0063] The dynamic terrain evolution modeling module 4 includes:

[0064] The physics-data dual-drive engine 41, the numerical simulation branch and the 3D ConvLSTM branch are coupled through differentiable hierarchical coupling to achieve gradient backpropagation.

[0065] Multi-factor coupled system 42, Graph Attention Network (GAT) models erosion-sedimentation-tectonic movement interaction, node update formula:

[0066]

[0067] where α_ij is calculated by the attention mechanism.

[0068] The adversarial prediction framework43 implements multi-resolution prediction based on the CycleGAN architecture, with a 1m resolution model used for landslide warning and a 10m model used for urban planning.

[0069] The online learning interface 44 supports real-time updates of RT-DInSAR data streams, with a model iteration cycle of less than 2 hours.

[0070] The interactive visualization and decision-making platform 5 includes:

[0071] Mixed Reality Engine 51 is based on Unity3D + Holographic Remoting and supports real-time loading of TB-level terrain data on Hololens 2.

[0072] The dynamic simulation sandbox 52 integrates the Houdini physics engine, which can simulate the influence of parameters such as rainfall and earthquake intensity.

[0073] Intelligent decision assistance 53 generates risk heat maps based on the SVM classifier and uses the PSO algorithm to invert the optimal engineering parameters, with a response delay of less than 200ms.

[0074] In summary, this invention builds a self-evolving intelligent terrain mapping system through a closed-loop technology loop involving multimodal data fusion, multi-granular feature analysis, and multi-mechanism coupled modeling. This method and system not only represents a breakthrough in a single technology but also establishes a collaboratively enhanced "data-model-decision" paradigm, providing a scalable technical foundation for emerging fields such as digital twin geology and smart cities.

[0075] With the above-described preferred embodiments of the present invention as a guide, and with reference to the above description, relevant personnel are fully capable of making various changes and modifications without departing from the technical scope of this invention. The technical scope of this invention is not limited to the contents of the specification and must be determined according to the scope of the claims.

Claims

1. A terrain surveying and mapping method based on artificial intelligence, characterized by: The following steps are involved: (1) Multi-source data fusion acquisition: Use multiple devices to collect data sources of the terrain to be mapped, and use artificial intelligence to perform multimodal data registration on the collected data sources to suppress noise and enhance data; (2) Intelligent extraction of terrain features: The terrain features of the collected and fused data sources are extracted. The backbone network uses the HRNet+ architecture to maintain high-resolution features. In the task branch, semantic segmentation uses DeepLabv3+ to extract water system, vegetation, and artificial building information. Instance segmentation uses Mask R-CNN to identify independent terrain units. Key point detection uses SuperPoint to locate feature markers. Slope calculation and curvature analysis are also performed on the data source. (3) 3D reconstruction optimization: The data source after feature extraction is added to the neural radiation field model to obtain a 3D reconstruction model based on spatial image features, and the error of the formed 3D reconstruction model is corrected; (4) Dynamic terrain evolution modeling: Construct a spatiotemporal fusion multi-mechanism coupled intelligent evolution model, integrating physical constraint neural networks, multi-factor graph coupling systems, and adversarial generation prediction frameworks. Through online learning and mixed reality deduction platforms, multi-scale dynamic simulation and interactive inversion of terrain evolution are achieved to meet the needs of geological disaster early warning and long-term strategic planning. The specific method for suppressing noise and enhancing data in step (1) includes constructing a noise-aware generative adversarial network, separating terrain signals from sensor noise through a spectral normalization discriminator, and using the StyleGAN2-ADA algorithm to perform texture enhancement on low-quality areas, thereby increasing the resolution of weak feature areas by more than 4 times. At the same time, a lightweight edge computing module is developed to fuse multi-source data streams in real time and output a standardized terrain feature tensor to support end-to-end feature transfer for subsequent terrain extraction tasks. The HRNet+ high-resolution feature preservation in step (2) adopts an improved high-resolution network to maintain spatial accuracy through multi-resolution parallel processing and enhanced feature fusion; The multi-resolution parallel processing is to construct a resolution sequence of four branches for the input image size H×W, and then use the formula Compute the feature exchange performed at each stage, where It is a cross-resolution feature transformation operator, including 1×1 convolution and bilinear interpolation; The enhanced feature fusion introduces the coordinate attention module CA to optimize feature selection. The processing result of the feature map by the coordinate attention module is expressed as follows: Calculated; CA is coordinate attention, which is used to embed the coordinate information of the spatial position into the channel attention; GAP is global average pooling, which is used to extract the global average context information of the feature map F; GMP is global maximum pooling, which is used to extract the most significant texture features in the feature map F; σ is the Sigmoid activation function, which converts the calculated score into an attention weight between 0 and 1; Conv is a 1x1 convolution operation, which transforms and fuses the spliced ​​features; It is an element-wise multiplication used to apply the attention weights back to the input features to achieve feature recalibration.

2. The method for terrain surveying and mapping based on artificial intelligence according to claim 1, characterized in that: The specific method of using multiple devices to collect data sources for the terrain to be surveyed in step (1) includes deploying a stereo observation network, integrating satellite multispectral images, UAV LiDAR point clouds, oblique photography data of ground mobile mapping systems and geological radar profiles, and constructing a multi-scale data set from centimeters to kilometers.

3. The method for terrain surveying and mapping based on artificial intelligence according to claim 1, characterized in that: The multimodal data registration in step (1) includes adopting a Transformer-based cross-modal registration network to solve the spatiotemporal reference difference through a deformable attention mechanism to achieve high-precision spatial matching of optical images and point cloud data.

4. The method for terrain surveying and mapping based on artificial intelligence according to claim 2, characterized in that: The specific method for reconstructing and optimizing the three-dimensional model in step (3) is to adopt feature-enhanced NeRF modeling, fuse feature encoding with spatial coordinates, predict density and color through MLP, generate predicted images using differentiable rendering, jointly optimize reconstruction and feature consistency, iteratively optimize model details based on geometric constraints and multi-view consistency, and at the same time, prune low-density areas to reduce floating noise, convert the point cloud output by NeRF into an explicit grid, fill in missing patches, force the optical flow of adjacent view rendering results to be consistent with the real offset, and introduce a discriminator network to distinguish high-frequency details of real and rendered images.

5. The method for terrain surveying and mapping based on artificial intelligence according to claim 4, characterized in that: The dynamic terrain evolution modeling in step (4) specifically includes using a spatiotemporal tensor data structure and a Transformer architecture to process historical and real-time data, capturing terrain changes at different time scales, combining numerical simulation with a data-driven dual-branch network, ensuring that predictions conform to physical laws through a differentiable physical layer, utilizing multi-agent simulation and graph neural networks to handle the interaction of erosion and sedimentation factors, using GANs for multi-resolution predictions, and updating models through online learning, and developing a visualization platform to support multiple simulations and real-time adjustments.

6. An artificial intelligence-based terrain mapping system, used to implement the artificial intelligence-based terrain mapping method according to any one of claims 1 to 5, characterized in that: The system includes: Data acquisition and preprocessing module, including a multi-sensor control unit, a cross-modal registration engine, a noise suppression enhancer, and an edge computing node, for multi-source heterogeneous data acquisition and intelligent fusion; The terrain feature intelligent extraction module includes a high-resolution feature backbone, a semantic segmentation unit, an instance segmentation unit, a key point detection unit, and a terrain parameter calculator for multi-dimensional terrain feature analysis; 3D reconstruction optimization module, including neural radiation field core, error correction unit and LOD generator, for high-precision 3D modeling and correction; Dynamic terrain evolution modeling module, including a physical-data dual-driven engine, a multi-factor coupling system, an adversarial prediction framework, and an online learning interface, for temporally and spatially coupled terrain evolution prediction; Interactive visualization and decision-making platform, including a mixed reality engine, dynamic simulation sandbox, and intelligent decision-making assistance, for multi-dimensional data fusion display and simulation; The data acquisition and preprocessing module, terrain feature intelligent extraction module, three-dimensional reconstruction optimization module, dynamic terrain evolution modeling module and interactive visualization and decision-making platform are connected in sequence. Each module communicates through the gRPC interface and supports independent upgrades.

Images