A high-precision modeling and tracking method and system for characteristic terrain features

By collecting multimodal topographic data through an integrated air-space-ground collaborative observation network, and performing feature-level deep fusion and semantic modeling, the problem of insufficient fusion capability of multi-source observation data was solved. This enabled dynamic tracking and adaptive updating of high-precision topographic models, improving the model's construction accuracy and applicability.

CN123118938APending Publication Date: 2026-11-17INNER MONGOLIA WANZHENG GENERAL AVIATION CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611301007.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-26
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing topographic modeling methods struggle to effectively integrate multi-source, multi-scale topographic information and lack the ability to continuously and dynamically track changes in topography and adapt the model to its own changes, resulting in insufficient model accuracy and reduced applicability.

Method used

An integrated air-space-ground collaborative observation network was established. Through a multi-scale collaborative acquisition system integrating high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles, multimodal terrain and geomorphological data were collected. Feature-level deep fusion and semantic modeling were performed, a dynamic tracking mechanism was established, and a high-precision terrain and geomorphological semantic model was generated.

Benefits of technology

It achieves deep fusion of multi-scale terrain information, improves the accuracy of the terrain and geomorphology semantic model, and realizes the monitoring of the full-cycle changes of terrain and geomorphology and the continuous optimization and updating of the model through a dynamic tracking mechanism.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118938A_ABST
    Figure CN123118938A_ABST
Patent Text Reader

Abstract

This invention discloses a high-precision modeling and tracking method and system for characteristic terrain features, belonging to the field of terrain modeling technology. The method includes: establishing an integrated air-space-ground collaborative observation network; collecting multimodal terrain data of the target characteristic terrain features, including wide-area surface cover and periodic deformation data, terrain and vegetation parameters, and surface micro-deformation; performing feature-level deep fusion and semantic modeling on the multimodal terrain data to generate a high-precision terrain semantic model, while establishing a dynamic tracking mechanism; and performing linked tracking comparison and adaptive updating on the high-precision terrain semantic model to generate a full-cycle dynamic terrain model. This invention solves the technical problems of insufficient multi-source data collaborative fusion capability and low dynamic change tracking accuracy in the existing terrain modeling process, achieving the technical effect of improving the modeling accuracy of characteristic terrain features and realizing continuous tracking and adaptive updating of dynamic changes in terrain features.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of terrain and landform modeling technology, specifically to a high-precision modeling and tracking method and system for characteristic terrain and landforms. Background Technology

[0002] With the development of remote sensing mapping, UAV surveying, and high-precision ground monitoring technologies, topographic modeling technology is increasingly being applied to fields such as natural resource surveys, ecological environment monitoring, geological disaster early warning, and engineering construction assessment. Existing topographic modeling methods typically use one or more of the following as data sources: satellite remote sensing imagery, aerial photogrammetry data, or ground survey data. Regional topographic models are constructed through image interpretation, 3D reconstruction, and spatial interpolation. However, due to differences in spatial resolution, coverage, data types, and sampling periods among different observation methods, existing methods struggle to effectively integrate wide-area topographic structure information, detailed local geomorphic features, and microscale deformation information. This results in topographic models with insufficient spatial representation accuracy and missing local details. Furthermore, existing modeling processes are typically based on static data processing, lacking continuous tracking and dynamic update mechanisms for ongoing changes such as surface subsidence, vegetation change, and geomorphic evolution. This makes it difficult for models to reflect actual topographic changes in a timely manner, reducing their applicability in long-term monitoring and dynamic analysis scenarios. Therefore, how to achieve the collaborative fusion of multi-source, multi-scale topographic and geomorphological observation data, and establish a high-precision topographic and geomorphological model with continuous tracking and adaptive updating capabilities, has become an urgent problem to be solved in the field of topographic and geomorphological modeling. Summary of the Invention

[0003] This application provides a high-precision modeling and tracking method and system for characteristic terrain features, which solves the technical problems in the existing technology of insufficient multi-source observation data fusion capability, low correlation accuracy of terrain information at different scales, and difficulty in continuously and dynamically tracking and adaptively updating the model during terrain feature modeling.

[0004] A first aspect of this application provides a high-precision modeling and tracking method for characteristic terrain features, the method comprising: An integrated air-space-ground collaborative observation network is established, which is a multi-scale collaborative acquisition system integrating high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles. This network collects multimodal topographic and geomorphic data, including wide-area surface cover and periodic deformation data, topographic and vegetation parameters, and surface micro-deformation data. Feature-level deep fusion and semantic modeling are performed on the multimodal topographic and geomorphic data to generate a high-precision topographic and geomorphic semantic model, while a dynamic tracking mechanism is established. Based on this dynamic tracking mechanism, the high-precision topographic and geomorphic semantic model is linked for tracking comparison and adaptive updating to generate a full-cycle dynamic topographic and geomorphic model.

[0005] A second aspect of this application provides a high-precision modeling and tracking system for characteristic terrain features, the system comprising: The observation network construction module establishes an integrated air-space-ground collaborative observation network, which is a multi-scale collaborative acquisition system integrating high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles. The data acquisition module collects multimodal topographic and geomorphic data, including wide-area surface cover and periodic deformation data, topographic and vegetation parameters, and surface micro-deformation data, through the integrated air-space-ground collaborative observation network. The modeling module performs feature-level deep fusion and semantic modeling on the multimodal topographic and geomorphic data to generate a high-precision topographic and geomorphic semantic model, while establishing a dynamic tracking mechanism. The update module performs linked tracking comparison and adaptive updates on the high-precision topographic and geomorphic semantic model based on the dynamic tracking mechanism, generating a full-cycle dynamic topographic and geomorphic model.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, an integrated air-space-ground collaborative observation network was established, which is a multi-scale collaborative acquisition system integrating high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles. Next, multi-modal topographic and geomorphic data, including wide-area land cover and periodic deformation data, topographic and vegetation parameters, and micro-deformation of the target terrain features, were collected through this integrated network. Then, feature-level deep fusion and semantic modeling were performed on the multi-modal topographic and geomorphic data to generate a high-precision topographic and geomorphic semantic model, while a dynamic tracking mechanism was established. Finally, based on the dynamic tracking mechanism, the high-precision topographic and geomorphic semantic model was linked for tracking comparison and adaptive updating, generating a full-cycle dynamic topographic and geomorphic model. This solves the technical problems of insufficient multi-source observation data fusion capability, low correlation accuracy of topographic information at different scales, and difficulty in continuously and dynamically tracking and adaptively updating the topographic and geomorphic change process in existing technologies. It achieves the technical effects of deep fusion of multi-scale collaborative observation data from air, space, and ground, improving the accuracy of topographic and geomorphic semantic model construction, and realizing full-cycle monitoring of topographic and geomorphic changes and continuous optimization and updating of the model through a dynamic tracking mechanism. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 A schematic diagram of a high-precision modeling and tracking method for characteristic terrain features provided in this application embodiment; Figure 2 This is a schematic diagram of the terrain three-dimensional spatial grid feature index provided in the embodiments of this application; Figure 3 This is a schematic diagram of a high-precision modeling and tracking system for characteristic terrain features provided in an embodiment of this application.

[0009] Figure labeling: Observation network construction module 11, data acquisition module 12, modeling module 13, update module 14. Detailed Implementation

[0010] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0011] Example 1, as Figure 1As shown, this application provides a high-precision modeling and tracking method for characteristic terrain features, wherein the method includes: An integrated air-space-ground collaborative observation network is established, which is a multi-scale collaborative acquisition system constructed by integrating high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles.

[0012] Within the target terrain and landform area, a spatial deployment scheme for an integrated air-space-ground collaborative observation network is planned based on the regional spatial scope, landform type distribution, and surface change monitoring needs. Specifically, high-resolution remote sensing satellite observation nodes are set up in the space layer to periodically acquire multi-temporal remote sensing image data of the target area using high-resolution remote sensing satellites. A wide-area land cover observation benchmark is established using satellite orbit parameters, imaging time information, and geographic reference information to characterize the large-scale topographic structure, land cover type, and long-term change trend of the target area. In the low-altitude layer, a multispectral lidar UAV cluster is set up. Based on the topographic complexity of the target area and the distribution of key monitoring areas, multiple UAV observation units are configured at different flight paths and operating altitudes. The multispectral sensors on board collect surface spectral response information, and the lidar acquires three-dimensional point cloud data of the surface to obtain high-precision topographic relief, vegetation structure, and micro-topographic features of the local area. In the ground layer, Beidou high-precision displacement stakes are set up. Based on the geological stability analysis requirements and the distribution of historical deformation risk areas, multiple monitoring points are selected and fixedly deployed in the target area. The spatial coordinate change information of each monitoring point is continuously collected through the Beidou high-precision positioning module to form time-series monitoring data of surface micro-deformation. Furthermore, data collected by high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles are integrated into a unified data collaborative management platform. Spatial coordinate benchmarks are unified, time synchronization is calibrated, and data communication is configured for each observation node. This enables observation data of different spatial scales, sampling periods, and data types to be transmitted in association according to a unified spatiotemporal benchmark. As a result, a multi-scale collaborative acquisition system covering wide-area scales, regional scales, and local micro-scales is constructed, enabling continuous and multi-dimensional observation of target feature terrain and geomorphological regions.

[0013] The integrated air-space-ground collaborative observation network collects multimodal topographic data of the target features, including wide-area surface cover and periodic deformation data, topographic and vegetation parameters, and surface micro-deformation data.

[0014] The target terrain and landform area is collaboratively observed and acquired through the integrated air-space-ground observation network. Specifically, high-resolution remote sensing satellite observation nodes are invoked to acquire multi-temporal remote sensing image data of the target area. Change information is extracted based on remote sensing images corresponding to different time periods to obtain wide-area land cover and periodic deformation data characterizing changes in land cover type, land structure evolution, and periodic deformation trends within the area. The wide-area land cover data includes land cover category, surface texture features, landform zoning features, and vegetation cover range information. The periodic deformation data includes deformation difference information at corresponding locations in different temporal images. A multispectral lidar UAV cluster is invoked to perform low-altitude fine scanning of the target area according to a preset flight path. Multispectral sensors collect surface spectral response data in different bands, and lidar acquires three-dimensional point cloud data of the target area. Based on the multispectral data and three-dimensional point cloud data... Topographic and vegetation parameters are extracted from point cloud data. The topographic parameters include surface elevation, slope, aspect, topographic relief, and local geomorphic boundary features. The vegetation parameters include vegetation cover, canopy height, canopy structure density, and spectral response characteristics. Spatial positioning data of key monitoring locations in the target area are continuously collected using BeiDou high-precision displacement stakes. Time-series analysis of the continuously collected positioning data yields surface micro-deformation data reflecting slow surface subsidence, local displacement changes, and abnormal displacement trends. Furthermore, various data collected by high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement stakes are linked and organized according to a unified spatial coordinate benchmark and time identifier to establish a multimodal topographic and geomorphic dataset containing wide-area surface cover and periodic deformation data, topographic and vegetation parameters, and surface micro-deformation data. This provides a data foundation for subsequent multi-source feature fusion and high-precision topographic and geomorphic semantic modeling.

[0015] The multimodal terrain and geomorphology data are subjected to feature-level deep fusion and semantic modeling to generate a high-precision terrain and geomorphology semantic model, while a dynamic tracking mechanism is established.

[0016] Furthermore, feature-level deep fusion and semantic modeling are performed on the multimodal terrain and geomorphology data to generate a high-precision terrain and geomorphology semantic model, including: The multimodal terrain and geomorphology data are subjected to spatiotemporal alignment processing and feature orientation extraction to obtain a terrain and geomorphology feature association index table. A geomorphology prior knowledge base is introduced to construct multimodal feature fusion rules, which are used to transform geomorphological domain knowledge into executable fusion weight rules. Feature-level deep fusion is performed on the terrain and geomorphology feature association index table using the multimodal feature fusion rules to obtain a terrain and geomorphology fusion feature body. Based on the terrain and geomorphology fusion feature body, hierarchical semantic modeling is performed to generate a high-precision terrain and geomorphology semantic model.

[0017] First, spatiotemporal alignment and feature extraction are performed on the multimodal terrain data. Multi-temporal remote sensing raster data acquired by high-resolution remote sensing satellites, 3D point cloud data and multispectral data acquired by multispectral lidar UAVs, and time-series deformation data acquired by BeiDou high-precision displacement piles are converted to a unified geospatial reference coordinate system. Temporal correlations are established based on the acquisition time of each data source to ensure consistency in spatial location and temporal scale. Further, according to the 3D spatial grid divided according to the target feature terrain region, the spatiotemporally aligned multimodal data are mapped to the corresponding spatial grid cells, and feature extraction is performed for different modalities. Corresponding to topographic and geomorphic features, land cover category, geomorphic boundaries, texture changes, and periodic change features are extracted from remote sensing raster data; elevation changes, topographic relief, slope aspect, and vegetation canopy structure features are extracted from 3D point cloud data; and displacement change trends, deformation rates, and anomalous change point features are extracted from time-series deformation data. By indexing and binding the multi-source features associated with each spatial grid unit, a topographic and geomorphic feature association index table representing the correspondence between spatial location, data source, and feature attributes is formed. Secondly, a pre-constructed geomorphic prior knowledge base is introduced, which includes topographic structural features corresponding to different geomorphic types, Spatial distribution patterns, vegetation association characteristics, and historical evolution characteristics are analyzed by regularizing the domain knowledge in the geomorphological prior knowledge base. This transforms the association between geomorphological types and different modal features into executable multimodal feature fusion rules. These rules determine the corresponding fusion weights based on the importance of various features in different geomorphological scenarios. Then, the multimodal feature fusion rules are used to assign weights to each spatial grid feature in the topographic feature association index table. Based on the fusion weights corresponding to different modal features, feature-level joint calculations are performed on remote sensing features, point cloud features, multispectral features, and deformation features. Different modal features are mapped to a unified feature representation space to obtain a topographic and geomorphic fusion feature body containing topographic structure information, geomorphic attribute information, and dynamic change information. Finally, hierarchical semantic modeling is performed based on the topographic and geomorphic fusion feature body. The spatial structure features, geomorphic category features, and dynamic change features in the fusion feature body are mapped to the topographic skeleton layer, semantic attribute layer, and dynamic deformation layer, respectively. Through three-dimensional spatial reconstruction, semantic label association, and dynamic information fusion processing, the correspondence between spatial location and geomorphic semantics is established, and a high-precision topographic and geomorphic semantic model containing topographic morphology, geomorphic category, vegetation attributes, and change trend information is generated.

[0018] Furthermore, spatiotemporal alignment and feature-oriented extraction are performed on the multimodal terrain and geomorphology data to obtain a terrain and geomorphology feature association index table, including: A reference geodetic coordinate system is constructed, and the multimodal terrain and landform data are mapped to the reference geodetic coordinate system for spatial alignment to obtain first multimodal terrain and landform data. The first multimodal terrain and landform data is then timestamped and aligned to obtain second multimodal terrain and landform data. The entire domain space of the target feature terrain and landform is divided into three-dimensional grids according to a preset grid size to generate a global terrain three-dimensional spatial grid. Based on the global terrain three-dimensional spatial grid, feature-oriented extraction is performed on the second multimodal terrain and landform data to obtain a terrain and landform feature association index table.

[0019] First, based on the geographical range, topographic relief characteristics, and spatial distribution of observation data of the target topographic region, a benchmark geodetic coordinate system is established. This benchmark geodetic coordinate system includes a unified geodetic datum, spatial projection parameters, and an elevation datum, serving as a spatial reference benchmark for topographic data from different sources. For remote sensing raster data acquired by high-resolution remote sensing satellites, 3D point cloud data and multispectral data acquired by multispectral lidar UAVs, and displacement monitoring data acquired by BeiDou high-precision displacement stakes, the original coordinate information corresponding to each data point is read, and the data from different coordinate systems are transformed based on coordinate transformation parameters. The data is converted to the reference geodetic coordinate system, and the converted spatial position is corrected for errors based on ground control points or high-precision positioning reference points, so that different modal data correspond to the same spatial reference frame, resulting in the first multimodal terrain and geomorphology data after spatial alignment. Next, the acquisition time information corresponding to each data acquisition node in the first multimodal terrain and geomorphology data is read, and a unified timestamp is added to each spatial data unit. A time calibration relationship is established based on the sampling period, data transmission delay, and observation time deviation of different observation devices, converting data acquired at different time scales to a unified time reference. Time matching is performed on observation data at different times corresponding to the same spatial location to obtain the second multimodal terrain and geomorphology data with a unified time identifier. Then, according to the spatial range and modeling accuracy requirements of the target feature terrain and geomorphology region, a preset grid size is set, and the three-dimensional spatial range of the target region is regularly partitioned, dividing the continuous geographic space into multiple three-dimensional grid units with unique spatial numbers. Spatial boundary coordinates, elevation range, and adjacency information are configured for each three-dimensional grid unit, generating a global terrain three-dimensional spatial grid covering the entire target region. Finally, the second multimodal terrain and geomorphology data is... Mapped to the corresponding 3D grid cells, targeted feature extraction is performed based on the data type of different modal data. Specifically, land cover category, texture change, and regional morphological features are extracted from remote sensing raster data; elevation change, topographic relief, slope aspect, and vegetation canopy structure features are extracted from 3D point cloud data; and displacement trend, deformation rate, and anomalous change point features are extracted from displacement monitoring data. The extracted features are then associated and stored according to the corresponding spatial grid number, data source, and time identifier to establish the correspondence between spatial location, temporal state, and multimodal features, forming a topographic feature association index table.

[0020] Furthermore, based on the global terrain 3D spatial grid, feature-oriented extraction is performed on the second multimodal terrain data to obtain a terrain feature association index table, including: Feature-oriented extraction is performed on each submodal data in the second multimodal topographic and geomorphological data to sequentially obtain a topographic raster feature set containing global land cover classification features, macro-topographic boundary contour features, and large-scale surface erosion texture features; a three-dimensional point cloud feature set containing topographic elevation change line features, micro-topographic undulation morphology features, and vegetation canopy structure parameter features; and a temporal deformation feature set containing single-pile temporal displacement trend features and anomalous displacement change point features. The topographic raster feature set, the three-dimensional point cloud feature set, and the temporal deformation feature set are then normalized sequentially to obtain a standard submodal topographic and geomorphological feature set. The standard submodal topographic and geomorphological feature set is then associated and bound to an index for each spatial grid in the global topographic three-dimensional spatial grid to construct a topographic and geomorphological feature association index table.

[0021] First, based on the spatial extent and numbering information of each spatial grid unit in the comprehensive terrain 3D spatial grid, spatial partitioning and mapping are performed on each submodal data in the second multimodal terrain and geomorphology data, associating data from different sources with corresponding spatial grid units. For remote sensing raster data acquired by high-resolution remote sensing satellites, feature extraction is performed based on pixel spatial distribution relationships and multi-temporal change information. Different spectral response features are identified using a land cover classification algorithm to obtain comprehensive land cover classification features characterizing changes in regional land cover types. Based on the continuity of geomorphic textures and edge change information in remote sensing images, spatial boundaries between different geomorphic regions are extracted to obtain spatial extents characterizing mountains, slopes, valleys, and other geomorphic units. The study focuses on the macroscopic geomorphic boundary contours; by analyzing texture changes, color differences, and surface structure changes in multi-temporal remote sensing images, it extracts large-scale surface erosion texture features reflecting water erosion, wind erosion, and surface morphology changes, forming a terrain raster feature set containing multi-scale surface information; for 3D point cloud data acquired by multispectral lidar UAVs, it performs 3D spatial analysis based on point cloud spatial coordinates and elevation information, identifies areas of rapid terrain elevation changes by calculating elevation gradients between adjacent points, and extracts terrain elevation change line features; based on the local neighborhood spatial distribution relationship of point clouds, it analyzes the degree of surface undulation, curvature changes, and local morphological structures, extracting features reflecting local morphological changes such as gullies, steps, protrusions, and depressions. Micro-topographic undulation morphology characteristics; combining multispectral data and point cloud height information, spatial segmentation of vegetation areas is performed, and vegetation height distribution, canopy density, and vertical structure variation parameters are calculated to obtain vegetation canopy structure parameter characteristics, forming a three-dimensional point cloud feature set; for continuous displacement monitoring data collected by Beidou high-precision displacement piles, a time series is established according to the spatial grid position corresponding to each displacement pile, and the displacement change within the continuous sampling period is analyzed to extract the single-pile temporal displacement trend characteristics reflecting the slow movement trend of the ground surface. By detecting the displacement change amplitude and rate of change within adjacent time windows, displacement anomalies exceeding the normal change range are identified, and the characteristics of abnormal displacement abrupt change points are extracted to form a temporal deformation feature set; furthermore, to eliminate different modes Based on the differences in dimensions, value ranges, and data distribution of the dynamic features, normalization processing is performed on the terrain raster feature set, 3D point cloud feature set, and temporal deformation feature set, respectively, to convert various feature parameters to a unified numerical range, thereby obtaining a standard submodal terrain and geomorphological feature set. Finally, according to the spatial number corresponding to each spatial grid unit in the global terrain 3D spatial grid, various features in the standard submodal terrain and geomorphological feature set are associated and bound according to spatial location, feature type, and acquisition time, establishing a one-to-one correspondence between each spatial grid unit and the terrain raster feature, 3D point cloud feature, and temporal deformation feature, and constructing a terrain and geomorphological feature association index table containing spatial index, modal type, and feature parameters.

[0022] After completing the spatiotemporal alignment processing of the second multimodal terrain data, the target feature terrain region is divided into a three-dimensional spatial grid according to a preset spatial resolution, dividing the continuous terrain space into multiple grid cells with independent spatial location identifiers. Each grid cell is used to carry multimodal terrain feature information within its corresponding spatial range.

[0023] like Figure 2 As shown, by constructing a three-dimensional spatial grid of the entire terrain, the terrain information at different spatial locations within the target area is discretized. Using grid cells as the basis for feature association, terrain elevation features, land cover features, vegetation structure features, and surface deformation features extracted within the corresponding grid area are spatially bound. This establishes a correlation between terrain and geomorphological data from different sources and at different scales under a unified spatial location, forming a terrain and geomorphological feature association index. Specifically, for each spatial grid cell, a corresponding index identifier is determined based on its spatial coordinates, and the multimodal feature data within that spatial area is associated and stored. This allows for the rapid retrieval of corresponding terrain and geomorphological features based on the spatial grid cell for feature fusion processing and semantic model construction.

[0024] Furthermore, the multimodal feature fusion rules are used to perform feature-level deep fusion on the terrain and landform feature association index table to obtain a terrain and landform fused feature body, including: The multimodal feature fusion rules are used to assign weights to the geomorphic feature labels of each spatial grid in the topographic feature association index table to determine the single-grid multimodal feature weight coefficient set; depth-weighted fusion is performed on the features of each spatial grid in the topographic feature association index table according to the single-grid multimodal feature weight coefficient set to obtain the single-grid topographic feature fusion feature; neighborhood linkage verification and consistency optimization are performed on the single-grid topographic feature fusion feature to obtain the topographic feature fusion body.

[0025] First, the geomorphic feature label information corresponding to each spatial grid in the topographic feature association index table is read. This geomorphic feature label information includes geomorphic type labels, terrain structure labels, vegetation cover labels, and deformation state labels. Based on the multimodal feature fusion rules, the importance of features from different sources within each spatial grid is analyzed. Combining the geomorphic type, terrain complexity, and dynamic change state of the current spatial grid, the fusion weights corresponding to the terrain raster features, 3D point cloud features, and temporal deformation features are determined. Specifically, based on the mapping relationship between different geomorphic types and the contribution of multimodal features in the geomorphic prior knowledge base, the weights of each modality feature are adjusted to give higher weights to key features reflecting the current geomorphic state, while reducing the influence of features affected by environmental interference or with weak expressive power. This forms a single-grid multimodal feature weight coefficient set for each spatial grid. Second, based on the single-grid multimodal feature weight coefficient set, the standard submodal topographic features within the corresponding spatial grid in the topographic feature association index table are weighted and fused. This results in the fusion of the terrain raster feature vector, 3D point cloud feature vector, and temporal deformation feature. The variable feature vectors are jointly calculated according to their corresponding weights, enabling different modal features to complement each other within a unified feature space. This yields single-grid topographic and geomorphic fusion features that characterize the current spatial grid's terrain structure, geomorphic attributes, and changing trends. Furthermore, to reduce discontinuities in single spatial grid features caused by local noise, acquisition errors, or anomalous data, neighborhood linkage verification is performed on the single-grid topographic and geomorphic fusion features corresponding to each spatial grid based on the spatial adjacency relationship between adjacent grids in the global topographic 3D spatial grid. This analyzes the feature continuity, geomorphic category consistency, and deformation trend consistency between adjacent spatial grids. When abrupt feature changes occur between adjacent grids and do not conform to the spatial evolution law of geomorphology, the abnormal fusion results are corrected based on the feature distribution of multiple associated grids within the neighborhood. Spatial consistency optimization is used to adjust the fusion feature parameters, ensuring that the optimized spatial grid fusion features match the surrounding geomorphic structure. Finally, the spatial grid fusion features, after neighborhood linkage verification and consistency optimization, are organized according to spatial numbers to form a topographic and geomorphic fusion feature body covering the target feature terrain and geomorphic region.

[0026] Furthermore, based on the aforementioned terrain and landform fusion feature body, hierarchical semantic modeling is performed to generate a high-precision terrain and landform semantic model, including: The terrain and landform fusion feature body is hierarchically labeled to obtain terrain skeleton layer feature data, semantic attribute layer feature data, and dynamic deformation layer feature data; 3D reconstruction is performed based on the terrain skeleton layer feature data to generate a global terrain basic model; the semantic attribute layer feature data is used to semantically label the global terrain basic model to obtain a terrain and landform semantic model; the dynamic deformation layer feature data is embedded into the terrain and landform semantic model for fusion to generate a high-precision terrain and landform semantic model.

[0027] First, the topographic and geomorphic fusion feature body is analyzed for feature dimensions. Based on the geomorphic meaning represented by different types of features, it is layered and labeled. Features reflecting the spatial geometry of the terrain, elevation changes, and geomorphic boundary relationships are assigned to the topographic skeleton layer; features reflecting surface cover type, vegetation distribution, geomorphic category, and spatial attribute relationships are assigned to the semantic attribute layer; and features reflecting surface displacement changes, deformation trends, and anomalous changes are assigned to the dynamic deformation layer. This yields topographic skeleton layer feature data, semantic attribute layer feature data, and dynamic deformation layer feature data containing spatial structure information, geomorphic semantic information, and temporal change information. Second, three-dimensional spatial reconstruction is performed based on the topographic skeleton layer feature data. According to the elevation information, topographic relief features, and spatial adjacency relationships corresponding to each spatial grid, a three-dimensional spatial grid topology is constructed. Through grid connection, surface fitting, and elevation interpolation, the surface spatial shape of the target area is restored. The process involves several steps: First, generating a global topographic model representing the basic topographic structure of the target area. Second, semantically labeling the global topographic model using the semantic attribute layer feature data. This involves associating and binding the landform category, land cover type, vegetation attributes, and other surface semantic information corresponding to each spatial grid unit with the corresponding spatial location in the model. This assigns corresponding semantic identifiers to different topographic regions, forming a topographic and geomorphic semantic model that includes spatial geometric structure and geomorphic semantic information. Finally, mapping the dynamic deformation layer feature data to the topographic and geomorphic semantic model according to spatial location and temporal state. Dynamic state description information is added to spatial regions with displacement changes, subsidence trends, or local morphological changes. The association between dynamic deformation information and corresponding semantic spatial units is established, enabling the model to simultaneously express static topographic structure, geomorphic semantic attributes, and dynamic evolution state, ultimately generating a high-precision topographic and geomorphic semantic model.

[0028] Furthermore, the semantic attribute layer feature data is used to semantically label the global terrain base model to obtain a terrain and landform semantic model, including: The semantic attribute layer feature data is labeled and anchored to the corresponding grid cells of the global terrain base model to obtain global model cell semantic labels; a neighborhood semantic voting mechanism is used to perform topology optimization on the global model cell semantic labels to obtain global model optimized semantic labels; the global model optimized semantic labels are mapped to the global terrain base model for binding and encapsulation to obtain a terrain and landform semantic model.

[0029] First, the landform category information, land cover information, vegetation attribute information, and other surface semantic features corresponding to each spatial grid in the semantic attribute layer feature data are read. Based on the spatial number, spatial coordinate range, and topological connection relationship of each three-dimensional grid unit in the global terrain basic model, a spatial correspondence between the semantic attribute layer feature data and the three-dimensional model grid units is established. For each three-dimensional grid unit, the semantic attribute features at the corresponding location are parsed, and the semantic category corresponding to the grid unit is determined according to the preset landform semantic classification rules. The semantic category, semantic confidence, and associated attribute information are written into the corresponding three-dimensional grid unit to form a global model unit semantic label containing a mapping relationship between spatial location and landform semantic information. Furthermore, to reduce the semantic label anomaly caused by data acquisition errors, spatial resolution differences, or local boundary transitions, a neighboring grid unit is constructed based on the topological connection relationship between adjacent grid units in the global terrain basic model. The domain semantic association set acquires semantic label information of multiple neighboring grid cells within a preset neighborhood for each target grid cell, and statistically analyzes the distribution of different semantic categories within the neighborhood. Based on the proportion of semantic categories within the neighborhood, spatial distance relationships, and semantic consistency, the optimized semantic category of the target grid cell is determined. When there is a significant deviation between the initial semantic label of the target grid cell and the main semantic category of the neighborhood, the initial semantic label is corrected using neighborhood semantic voting results to obtain a global model optimized semantic label consistent with the spatial distribution pattern of the surrounding terrain. Finally, the global model optimized semantic label is mapped to the three-dimensional grid cells in the global terrain base model according to the corresponding spatial number, and the semantic category, attribute parameters, and label status information are bound and encapsulated with the corresponding model cells, so that each spatial cell in the three-dimensional model has the corresponding terrain semantic expression capability, resulting in a terrain and geomorphological semantic model containing spatial structure information and semantic attribute information.

[0030] Based on the dynamic tracking mechanism, the high-precision terrain and geomorphology semantic model is linked for tracking, comparison and adaptive updating to generate a full-cycle dynamic terrain and geomorphology model.

[0031] Furthermore, based on the aforementioned dynamic tracking mechanism, the high-precision terrain and geomorphology semantic model is subjected to linked tracking, comparison, and adaptive updating to generate a full-cycle dynamic terrain and geomorphology model, including: Based on the dynamic tracking mechanism, multimodal terrain and geomorphology incremental data are collected periodically; the multimodal terrain and geomorphology incremental data are linked and tracked with the high-precision terrain and geomorphology semantic model to obtain terrain and geomorphology change area data; the terrain and geomorphology change area data are used to adaptively update the high-precision terrain and geomorphology semantic model to generate a full-cycle terrain and geomorphology dynamic model.

[0032] First, based on the pre-established spatial grid tracking index, historical observation records, and data acquisition cycle configuration in the dynamic tracking mechanism, the integrated air-space-ground collaborative observation network is invoked at preset time intervals to repeatedly observe the target feature terrain and landform area, acquiring multimodal terrain and landform incremental data for the corresponding time period. This multimodal terrain and landform incremental data includes newly added temporal surface cover and periodic change data acquired by high-resolution remote sensing satellites, newly added terrain and vegetation parameter data acquired by multispectral lidar UAVs, and newly added surface micro-deformation data acquired by BeiDou high-precision displacement piles. Further, the multimodal terrain and landform incremental data undergoes spatial coordinate transformation, time reference calibration, and spatial grid mapping processing consistent with the high-precision terrain and landform semantic model, establishing a correlation between the newly observed data and the corresponding spatial grid units in the high-precision terrain and landform semantic model. Then, the processed multimodal terrain and landform incremental data is linked and compared with the high-precision terrain and landform semantic model. Difference analysis is performed based on the historical semantic features, basic terrain structure parameters, and newly observed feature parameters corresponding to the same spatial grid unit, and terrain features are calculated accordingly. The model identifies differences in structural changes, geomorphic attribute changes, and dynamic deformation changes. Based on preset change judgment rules, it identifies spatial grid regions exceeding change thresholds and determines continuous or associated spatial grids that meet the change conditions as geomorphic change regions. Simultaneously, it generates geomorphic change region data containing the spatial range, type, and degree of change. Further, the high-precision geomorphic semantic model is adaptively updated using this geomorphic change region data. For each change region, incremental data of the corresponding modality is called to re-perform feature extraction and fusion processing, obtaining updated local geomorphic feature expressions. Based on these local geomorphic feature expressions, the 3D structural information, semantic attribute labels, and dynamic deformation parameters in the original model are locally adjusted. Specifically, the 3D model structure is updated for regions where the topography has changed, the semantic labels are updated for regions where land cover or vegetation attributes have changed, and the dynamic state information is updated for regions undergoing continuous deformation. Simultaneously, the original model data for unchanged regions is retained, and the updated local model data is fused and correlated with historical model data to form a full-cycle geomorphic dynamic model containing continuous temporal evolution information.

[0033] Furthermore, the high-precision topographic and geomorphic semantic model is adaptively updated using the topographic and geomorphic change area data to generate a full-cycle dynamic topographic and geomorphic model, including: Based on the data of the terrain and landform change area, local features are re-fused to obtain a local terrain and landform fusion feature body; the local terrain and landform fusion feature body is used to adaptively update the high-precision terrain and landform semantic model to generate a full-cycle terrain and landform dynamic model.

[0034] First, based on the information regarding the spatial range, type, and degree of change contained in the topographic change area data, the corresponding three-dimensional spatial grid cells are located in the high-precision topographic semantic model, and historical topographic structure features, semantic attribute features, and dynamic deformation features corresponding to the change area are extracted. Further, multimodal topographic incremental data for the corresponding time period is reacquired for the change area, and a local updated data set is established based on the spatial range of the change area. The remote sensing raster features, three-dimensional point cloud features, and temporal deformation features in the local updated data set are re-associated. The fusion weights corresponding to each type of feature are re-determined according to the multimodal feature fusion rules, and feature-level weighted fusion is performed according to the spatial grid cells to obtain local topographic fusion features reflecting the current change state. Further, spatial consistency verification is performed on the local topographic fusion features, checking the topographic structure changes and semantic attribute changes of adjacent spatial grids within the change area. The model employs a linked analysis of changes in topography and deformation trends, performing neighborhood optimization on fusion features exhibiting spatial discontinuities or abrupt changes to obtain a local topographic fusion feature body that accurately represents the current state of the changed area. Then, the high-precision topographic semantic model is adaptively updated using this local topographic fusion feature body, mapping the topographic structure information in the feature body to corresponding 3D model mesh cells, and adjusting the spatial morphology of areas experiencing geomorphic changes. The updated semantic attribute features are mapped to corresponding model cells, and land cover categories, vegetation attributes, and geomorphic category labels are updated synchronously. Dynamic deformation features are embedded into corresponding spatial locations to update the displacement trends and evolutionary states of continuously changing areas. Simultaneously, the original model state of unchanged areas remains unchanged, and the updated local model data is associated and stored with historical model data in chronological order, forming a full-cycle dynamic topographic model that reflects the long-term evolution of topography and geomorphology.

[0035] In summary, the embodiments of this application have at least the following technical effects: First, an integrated air-space-ground collaborative observation network was established, which is a multi-scale collaborative acquisition system integrating high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles. Next, multi-modal topographic and geomorphic data, including wide-area land cover and periodic deformation data, topographic and vegetation parameters, and micro-deformation of the target terrain features, were collected through this integrated network. Then, feature-level deep fusion and semantic modeling were performed on the multi-modal topographic and geomorphic data to generate a high-precision topographic and geomorphic semantic model, while a dynamic tracking mechanism was established. Finally, based on the dynamic tracking mechanism, the high-precision topographic and geomorphic semantic model was linked for tracking comparison and adaptive updating, generating a full-cycle dynamic topographic and geomorphic model. This solves the technical problems of insufficient multi-source observation data fusion capability, low correlation accuracy of topographic information at different scales, and difficulty in continuously and dynamically tracking and adaptively updating the topographic and geomorphic change process in existing technologies. It achieves the technical effects of deep fusion of multi-scale collaborative observation data from air, space, and ground, improving the accuracy of topographic and geomorphic semantic model construction, and realizing full-cycle monitoring of topographic and geomorphic changes and continuous optimization and updating of the model through a dynamic tracking mechanism.

[0036] Example 2, based on the same inventive concept as the high-precision modeling and tracking method for characteristic terrain features in the foregoing examples, such as... Figure 3 As shown, this application provides a high-precision modeling and tracking system for characteristic terrain features, wherein the system includes: Observation Network Construction Module 11: Constructs an integrated air-space-ground collaborative observation network, which is a multi-scale collaborative acquisition system integrating high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles; Data Acquisition Module 12: Acquires multimodal topographic and geomorphic data of target features, including wide-area surface cover and periodic deformation data, topographic and vegetation parameters, and surface micro-deformation, through the integrated air-space-ground collaborative observation network; Modeling Module 13: Performs feature-level deep fusion and semantic modeling on the multimodal topographic and geomorphic data to generate a high-precision topographic and geomorphic semantic model, and establishes a dynamic tracking mechanism; Update Module 14: Based on the dynamic tracking mechanism, performs linkage tracking comparison and adaptive updating on the high-precision topographic and geomorphic semantic model to generate a full-cycle dynamic topographic and geomorphic model.

[0037] Furthermore, the modeling module 13 is used to perform the following methods: The multimodal terrain and geomorphology data are subjected to spatiotemporal alignment processing and feature orientation extraction to obtain a terrain and geomorphology feature association index table. A geomorphology prior knowledge base is introduced to construct multimodal feature fusion rules, which are used to transform geomorphological domain knowledge into executable fusion weight rules. Feature-level deep fusion is performed on the terrain and geomorphology feature association index table using the multimodal feature fusion rules to obtain a terrain and geomorphology fusion feature body. Based on the terrain and geomorphology fusion feature body, hierarchical semantic modeling is performed to generate a high-precision terrain and geomorphology semantic model.

[0038] Furthermore, the modeling module 13 is used to perform the following methods: A reference geodetic coordinate system is constructed, and the multimodal terrain and landform data are mapped to the reference geodetic coordinate system for spatial alignment to obtain first multimodal terrain and landform data. The first multimodal terrain and landform data is then timestamped and aligned to obtain second multimodal terrain and landform data. The entire domain space of the target feature terrain and landform is divided into three-dimensional grids according to a preset grid size to generate a global terrain three-dimensional spatial grid. Based on the global terrain three-dimensional spatial grid, feature-oriented extraction is performed on the second multimodal terrain and landform data to obtain a terrain and landform feature association index table.

[0039] Furthermore, the modeling module 13 is used to perform the following methods: Feature-oriented extraction is performed on each submodal data in the second multimodal topographic and geomorphological data to sequentially obtain a topographic raster feature set containing global land cover classification features, macro-topographic boundary contour features, and large-scale surface erosion texture features; a three-dimensional point cloud feature set containing topographic elevation change line features, micro-topographic undulation morphology features, and vegetation canopy structure parameter features; and a temporal deformation feature set containing single-pile temporal displacement trend features and anomalous displacement change point features. The topographic raster feature set, the three-dimensional point cloud feature set, and the temporal deformation feature set are then normalized sequentially to obtain a standard submodal topographic and geomorphological feature set. The standard submodal topographic and geomorphological feature set is then associated and bound to an index for each spatial grid in the global topographic three-dimensional spatial grid to construct a topographic and geomorphological feature association index table.

[0040] Furthermore, the modeling module 13 is used to perform the following methods: The multimodal feature fusion rules are used to assign weights to the geomorphic feature labels of each spatial grid in the topographic feature association index table to determine the single-grid multimodal feature weight coefficient set; depth-weighted fusion is performed on the features of each spatial grid in the topographic feature association index table according to the single-grid multimodal feature weight coefficient set to obtain the single-grid topographic feature fusion feature; neighborhood linkage verification and consistency optimization are performed on the single-grid topographic feature fusion feature to obtain the topographic feature fusion body.

[0041] Furthermore, the modeling module 13 is used to perform the following methods: The terrain and landform fusion feature body is hierarchically labeled to obtain terrain skeleton layer feature data, semantic attribute layer feature data, and dynamic deformation layer feature data; 3D reconstruction is performed based on the terrain skeleton layer feature data to generate a global terrain basic model; the semantic attribute layer feature data is used to semantically label the global terrain basic model to obtain a terrain and landform semantic model; the dynamic deformation layer feature data is embedded into the terrain and landform semantic model for fusion to generate a high-precision terrain and landform semantic model.

[0042] Furthermore, the modeling module 13 is used to perform the following methods: The semantic attribute layer feature data is labeled and anchored to the corresponding grid cells of the global terrain base model to obtain global model cell semantic labels; a neighborhood semantic voting mechanism is used to perform topology optimization on the global model cell semantic labels to obtain global model optimized semantic labels; the global model optimized semantic labels are mapped to the global terrain base model for binding and encapsulation to obtain a terrain and landform semantic model.

[0043] Furthermore, the update module 14 is used to perform the following method: Based on the dynamic tracking mechanism, multimodal terrain and geomorphology incremental data are collected periodically; the multimodal terrain and geomorphology incremental data are linked and tracked with the high-precision terrain and geomorphology semantic model to obtain terrain and geomorphology change area data; the terrain and geomorphology change area data are used to adaptively update the high-precision terrain and geomorphology semantic model to generate a full-cycle terrain and geomorphology dynamic model.

[0044] Furthermore, the update module 14 is used to perform the following method: Based on the data of the terrain and landform change area, local features are re-fused to obtain a local terrain and landform fusion feature body; the local terrain and landform fusion feature body is used to adaptively update the high-precision terrain and landform semantic model to generate a full-cycle terrain and landform dynamic model.

[0045] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A high-precision modeling and tracking method for characteristic terrain features, characterized in that, The method includes: A space-air-ground integrated collaborative observation network is established, which is a multi-scale collaborative acquisition system integrating high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles. The integrated air-space-ground collaborative observation network collects multimodal topographic and geomorphic data of the target features, including wide-area surface cover and periodic deformation data, topographic and vegetation parameters, and surface micro-deformation data. The multimodal terrain and geomorphology data are subjected to feature-level deep fusion and semantic modeling to generate a high-precision terrain and geomorphology semantic model, while a dynamic tracking mechanism is established. Based on the dynamic tracking mechanism, the high-precision terrain and geomorphology semantic model is linked for tracking, comparison and adaptive updating to generate a full-cycle dynamic terrain and geomorphology model.

2. The high-precision modeling and tracking method for characteristic terrain features as described in claim 1, characterized in that, The multimodal terrain and geomorphology data are subjected to feature-level deep fusion and semantic modeling to generate a high-precision terrain and geomorphology semantic model, including: Spatiotemporal alignment and feature orientation extraction are performed on the multimodal terrain and geomorphology data to obtain a terrain and geomorphology feature association index table; A prior knowledge base of landforms is introduced to construct multimodal feature fusion rules, which are used to transform landform domain knowledge into executable fusion weight rules. The multimodal feature fusion rules are used to perform feature-level deep fusion on the terrain and landform feature association index table to obtain a terrain and landform fusion feature body; Based on the aforementioned terrain and landform fusion feature body, hierarchical semantic modeling is performed to generate a high-precision terrain and landform semantic model.

3. The high-precision modeling and tracking method for characteristic terrain features as described in claim 2, characterized in that, The multimodal terrain and geomorphological data are subjected to spatiotemporal alignment processing and feature orientation extraction to obtain a terrain and geomorphological feature association index table, including: A reference geodetic spatial coordinate system is constructed, and the multimodal topographic data is mapped to the reference geodetic spatial coordinate system for spatial alignment to obtain the first multimodal topographic data. The first multimodal terrain and geomorphology data is timestamped and aligned to obtain the second multimodal terrain and geomorphology data. The entire space of the target feature terrain and landform is divided into three-dimensional meshes according to the preset mesh size to generate a three-dimensional spatial mesh of the entire terrain. Based on the global terrain 3D spatial grid, feature orientation extraction is performed on the second multimodal terrain and geomorphology data to obtain a terrain and geomorphology feature association index table.

4. The high-precision modeling and tracking method for characteristic terrain features as described in claim 3, characterized in that, Based on the global terrain 3D spatial grid, feature orientation extraction is performed on the second multimodal terrain and geomorphology data to obtain a terrain and geomorphology feature association index table, including: Feature orientation extraction is performed on each submodal data in the second multimodal topographic and geomorphological data to obtain a topographic raster feature set containing global land cover classification features, macro-topographic boundary contour features, and large-scale land erosion texture features; a three-dimensional point cloud feature set containing topographic elevation change line features, micro-topographic undulation morphology features, and vegetation canopy structure parameter features; and a temporal deformation feature set containing single pile temporal displacement trend features and anomalous displacement change point features. The terrain raster feature set, the three-dimensional point cloud feature set, and the temporal deformation feature set are normalized sequentially to obtain a standard submodal terrain and landform feature set. For each spatial grid in the global terrain 3D spatial grid, the standard submodal terrain and landform feature set is associated and bound with an index to construct a terrain and landform feature association index table.

5. The high-precision modeling and tracking method for characteristic terrain features as described in claim 2, characterized in that, The multimodal feature fusion rules are used to perform feature-level deep fusion on the terrain and landform feature association index table to obtain a fused terrain and landform feature body, including: The multimodal feature fusion rule is used to assign weights to the geomorphic feature labels of each spatial grid in the topographic feature association index table to determine the set of multimodal feature weight coefficients for a single grid. According to the single-grid multimodal feature weight coefficient set, the spatial grid features in the terrain and landform feature association index table are subjected to depth-weighted fusion to obtain single-grid terrain and landform fusion features; The single-grid terrain and landform fusion features are subjected to neighborhood linkage verification and consistency optimization to obtain the terrain and landform fusion feature body.

6. The high-precision modeling and tracking method for characteristic terrain features as described in claim 2, characterized in that, Based on the aforementioned terrain and landform fusion feature body, hierarchical semantic modeling is performed to generate a high-precision terrain and landform semantic model, including: The terrain and landform fusion feature body is layered and labeled to obtain terrain skeleton layer feature data, semantic attribute layer feature data and dynamic deformation layer feature data; Three-dimensional reconstruction is performed based on the terrain skeleton layer feature data to generate a global terrain basic model. The semantic attribute layer feature data is used to semantically label the global terrain base model to obtain a terrain and landform semantic model. The dynamic deformation layer feature data is embedded into the terrain and geomorphology semantic model for fusion to generate a high-precision terrain and geomorphology semantic model.

7. The high-precision modeling and tracking method for characteristic terrain features as described in claim 6, characterized in that, The semantic attribute layer feature data is used to semantically label the global terrain base model to obtain a terrain and landform semantic model, including: The semantic attribute layer feature data is labeled and anchored to the corresponding grid cells of the global terrain basic model to obtain semantic labels for the global model cells. A neighborhood semantic voting mechanism is used to perform topology optimization on the semantic labels of the global model units to obtain optimized semantic labels for the global model. The optimized semantic labels of the global model are mapped to the global terrain base model and bound and encapsulated to obtain the terrain and landform semantic model.

8. The high-precision modeling and tracking method for characteristic terrain features as described in claim 1, characterized in that, Based on the aforementioned dynamic tracking mechanism, the high-precision terrain and geomorphology semantic model is linked for tracking, comparison, and adaptive updating to generate a full-cycle dynamic terrain and geomorphology model, including: Based on the aforementioned dynamic tracking mechanism, incremental multimodal terrain and landform data are collected periodically; The multimodal incremental topographic data is linked and tracked with the high-precision topographic semantic model to obtain data on the area of ​​topographic change. The high-precision topographic and geomorphological semantic model is adaptively updated using the topographic and geomorphological change area data to generate a full-cycle topographic and geomorphological dynamic model.

9. The high-precision modeling and tracking method for characteristic terrain features as described in claim 8, characterized in that, The high-precision topographic and geomorphological semantic model is adaptively updated using the topographic and geomorphological change area data to generate a full-cycle dynamic topographic and geomorphological model, including: Based on the data of the terrain and landform change area, local feature re-fusion is performed to obtain a local terrain and landform fusion feature body; The high-precision terrain and geomorphology semantic model is adaptively updated using the local terrain and geomorphology fusion feature body to generate a full-cycle dynamic terrain and geomorphology model.

10. A high-precision modeling and tracking system for characteristic terrain features, characterized in that, The system is used to implement a high-precision modeling and tracking method for characteristic terrain features as described in any one of claims 1-9, the system comprising: Observation network construction module: Construct an integrated air-space-ground collaborative observation network, which is a multi-scale collaborative acquisition system integrating high-resolution remote sensing satellites, multispectral lidar UAV clusters, and BeiDou high-precision displacement piles; Data acquisition module: Collects multimodal topographic and geomorphic data of target features, including wide-area surface cover and periodic deformation data, topographic and vegetation parameters, and surface micro-deformation data, through the integrated air-space-ground observation network. Modeling module: Performs feature-level deep fusion and semantic modeling on the multimodal terrain and geomorphology data to generate a high-precision terrain and geomorphology semantic model, and establishes a dynamic tracking mechanism at the same time; Update module: Based on the dynamic tracking mechanism, the high-precision terrain and geomorphology semantic model is linked for tracking, comparison and adaptive update to generate a full-cycle dynamic terrain and geomorphology model.