Surveying and mapping geographic system based on remote sensing technology

Abstract

The invention relates to the technical field of geographic information engineering, in particular to a surveying and mapping geographic system based on a remote sensing technology, which is characterized in that a multi-modal remote sensing data acquisition module synchronously acquires satellite remote sensing images, aviation LiDAR point cloud, unmanned aerial vehicle oblique photography and ground Internet of Things sensing data; the data fusion collaboration module carries out geographic coordinate calibration and position coding, and generates a surveying and mapping data set with unified precision in combination with an improved Mamba fusion model; the dynamic terrain high-precision inversion module performs three-dimensional terrain reconstruction, and completes terrain deformation monitoring in combination with an InSAR interference measurement technology and a time sequence analysis algorithm; the real-time error correction module constructs a BP neural network optimized by an ISSA algorithm, and dynamically corrects a terrain three-dimensional model and a terrain deformation result; the intelligent visual application module constructs a digital twinborn simulation model and automatically generates a standardized surveying and mapping result report. Therefore, the problems of weak multi-source data fusion, low terrain inversion precision and the like in the prior art are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geographic information engineering, in particular to a surveying and mapping geographic system based on remote sensing technology. BACKGROUND

[0002] The surveying and mapping geographic system is the core technical support for land space planning, natural disaster monitoring, urban construction operation and maintenance, and ecological environment assessment. The data acquisition accuracy, terrain inversion efficiency and result application capability of the surveying and mapping geographic system directly determine the reliability and value of geographic information services. The current surveying and mapping scene is increasingly complex, from large-scale terrain surveying and mapping in mountainous areas, to fine three-dimensional modeling in urban built-up areas, to dynamic deformation monitoring in geological disaster areas, which puts higher requirements on the multi-source data adaptation, high-precision inversion and real-time application of the system.

[0003] However, the traditional surveying and mapping geographic system has obvious technical limitations: the multi-source data fusion capability is weak, satellite remote sensing images, aerial LiDAR point clouds, unmanned aerial vehicle oblique photography and ground sensing data are simply spliced or linearly fused, the spatial correlation and feature difference of geographic data are not fully considered, the fusion model is easy to lose spatial information or difficult to handle large-scale data, resulting in inconsistent data set accuracy; the dynamic terrain inversion timeliness and precision are insufficient, it mainly relies on a single data source, lacks long-term deformation monitoring capability, deformation rate calculation error is large, terrain update response lag; the error correction mechanism is rigid, only linear error can be processed, BP neural network correction is easy to fall into local optimum, and it is difficult to meet the requirement of millimeter level; the visualization and application intelligence are low, the output is static result, AR superposition is easy to misposition, query interaction is poor, and report generation period is long. The existing system faces multiple challenges, and an integrated system capable of realizing multi-modal data deep fusion, high-precision dynamic inversion, real-time error correction and intelligent visualization is needed to meet the modern surveying and mapping requirements of "high precision, real-time and intelligence". SUMMARY

[0004] The present application provides a surveying and mapping geographic system based on remote sensing technology to solve the problems of weak multi-source data fusion and low terrain inversion precision in the prior art.

[0005] The first aspect of this application provides a mapping and geographic system based on remote sensing technology, including: a multimodal remote sensing data acquisition module, a data fusion and collaboration module, a dynamic terrain high-precision inversion module, a real-time error correction module, and an intelligent visualization application module; wherein, the multimodal remote sensing data acquisition module is used to simultaneously acquire satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data, and ground IoT sensor data; the data fusion and collaboration module is used to perform geographic coordinate calibration and location encoding on the satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data, and ground IoT sensor data, and to perform feature extraction and nonlinear fusion by combining an improved Mamba fusion model, through selective state space... The gating mechanism strengthens the weight of key geographic features and generates a mapping dataset with uniform accuracy. The dynamic terrain high-precision inversion module performs 3D terrain reconstruction based on the mapping dataset, combines InSAR interferometry technology and time series analysis algorithms to complete terrain deformation monitoring, and uses terrain segmentation algorithms to dynamically update the terrain. The real-time error correction module is used to identify systematic and random errors generated in the terrain inversion process of the mapping dataset, and constructs a BP neural network optimized by the ISSA algorithm to dynamically correct the 3D terrain model and terrain deformation results. The intelligent visualization application module is used to construct a digital twin simulation model, perform AR real-scene overlay, provide interactive geographic information queries, and automatically generate standardized mapping result reports.

[0006] Preferably, the multimodal remote sensing data acquisition module includes a satellite image acquisition unit, a LiDAR point cloud acquisition unit, an oblique photography acquisition unit, and a ground sensor acquisition unit. The satellite image acquisition unit is used to interface with a multi-resolution satellite remote sensing platform to simultaneously acquire multispectral and panchromatic satellite image data. The LiDAR point cloud acquisition unit is used to acquire high-precision point cloud data of the terrain surface using an airborne LiDAR device. The oblique photography acquisition unit is used to acquire oblique photographic images of the terrain using a multi-view camera from an unmanned aerial vehicle (UAV). The ground sensor acquisition unit is used to acquire data on terrain surface temperature, humidity, and settlement monitoring points using deployed IoT sensors.

[0007] Preferably, the data fusion collaboration module includes a geocoding unit, a feature extraction unit, a fusion processing unit, and a dataset output unit. The geocoding unit performs geographic coordinate calibration on multi-source remote sensing data, introduces geographic feature location codes to identify data spatial attributes, and generates feature vectors containing location information. The feature extraction unit inputs the feature vectors into an improved Mamba model and extracts image texture, point cloud geometry, and terrain topology features in parallel through a linear attention mechanism. The fusion processing unit strengthens the weights of key geographic features through a selective state-space gating mechanism and performs nonlinear fusion of the image texture, point cloud geometry, and terrain topology features. The dataset output unit standardizes the fused feature data to generate a mapping dataset with uniform accuracy.

[0008] Preferably, the dynamic terrain high-precision inversion module includes a 3D reconstruction unit, a deformation monitoring unit, and a dynamic update unit. The 3D reconstruction unit uses LiDAR point cloud and oblique photogrammetric image data from the mapping dataset, and after denoising, registration, 3D grid construction, and texture bonding, constructs a basic 3D terrain model. The deformation monitoring unit uses multiple periods of InSAR imagery and LiDAR elevation data from the mapping dataset, and uses InSAR interferometry to register, generate interferograms, and unwrap the phases of the multiple periods of InSAR imagery, extracting initial terrain deformation phase information. It then uses a time-series analysis algorithm to perform temporal correlation analysis on the multiple periods of deformation phase information, calculates the long-term deformation rate and cumulative deformation value of the terrain, and compares and verifies it with ground sensor data, outputting accurate terrain deformation results. The dynamic update unit uses the deformation results to identify areas of terrain change through a terrain segmentation algorithm, dynamically optimizes the basic 3D terrain model, obtains a high-precision 3D terrain model, and synchronously updates the terrain deformation time-series data.

[0009] Preferably, the real-time error correction module includes an error identification unit, a model optimization unit, and a correction execution unit. The error identification unit is used to identify systematic and random errors generated in the topographic inversion process of the surveying dataset through statistical analysis and feature comparison. The model optimization unit is used to optimize the initial weights and structural parameters of the BP neural network using the ISSA algorithm to improve the model correction accuracy. The correction execution unit is used to input the error data into the optimized BP neural network to dynamically correct the high-precision three-dimensional topographic model and topographic deformation results.

[0010] Preferably, the intelligent visualization application module includes a digital twin construction unit, an AR overlay unit, a query interaction unit, and a report generation unit. The digital twin construction unit, based on the high-precision 3D terrain model, terrain deformation results, and surveying dataset, constructs a 1:1 mapping digital twin simulation model of the terrain and landforms, synchronously linking terrain deformation time-series data with land cover characteristics. The AR overlay unit is used to overlay the digital twin simulation model with real-world images, achieving real-time alignment and interactive browsing between the virtual model and the real terrain. The query interaction unit provides multi-condition interactive query functions for geographic information. The report generation unit integrates surveying data, inversion results, and error correction information using structured algorithms to generate a standardized surveying results report.

[0011] The second aspect of this application provides a method for a mapping and geographic system based on remote sensing technology, comprising: simultaneously acquiring satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data, and ground-based IoT sensor data; performing geographic coordinate calibration and location encoding on the satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data, and ground-based IoT sensor data; extracting image texture, point cloud geometry, and terrain topology features in parallel through an improved Mamba fusion model, and performing nonlinear fusion through a selective state-space gating mechanism to generate a mapping dataset of uniform accuracy; constructing a basic three-dimensional terrain model based on the mapping dataset; obtaining terrain deformation results by combining InSAR interferometry technology and time series analysis algorithms; and utilizing... The terrain segmentation algorithm identifies areas of terrain change, optimizes the basic 3D terrain model based on the terrain deformation results and change area information, and obtains a high-precision 3D terrain model while simultaneously updating the terrain deformation time series data. Through statistical analysis and feature comparison, the algorithm identifies systematic and random errors generated in the terrain inversion process of the surveying dataset, optimizes the BP neural network using the ISSA algorithm, and dynamically corrects the high-precision 3D terrain model and terrain deformation results. Based on the corrected high-precision 3D terrain model, terrain deformation results, and the surveying dataset, a digital twin simulation model is constructed, enabling AR real-scene overlay and multi-condition interactive query of geographic information. Finally, a structured algorithm is used to integrate surveying data, inversion results, and error correction information to generate a standardized surveying results report.

[0012] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement a method for a remote sensing-based mapping geographic system as described in the above embodiments.

[0013] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a method for a remote sensing-based mapping geographic system as described in the above embodiments.

[0014] The fifth aspect of this application provides a computer program product, including a computer program or instructions, for implementing a method of a remote sensing-based mapping and geographic system as described in the above embodiments.

[0015] Therefore, this application has the following beneficial effects: This application embodiment utilizes a multimodal remote sensing data acquisition module to simultaneously acquire satellite remote sensing imagery, aerial LiDAR point clouds, UAV oblique photography, and ground-based IoT sensor data. This overcomes the limitations of traditional single-source data coverage, expands the acquisition dimensions and spatial coverage of mapping data, ensures the comprehensiveness of terrain information representation, and provides a multi-dimensional data foundation for subsequent mapping analysis. Furthermore, leveraging a data fusion and collaboration module, an improved Mamba fusion model is used to extract image texture, point cloud geometry, and terrain topology features from multiple sources in parallel. A selective state-space gating mechanism is employed to strengthen the weight of key geographic features, performing nonlinear fusion and generating a unified precision mapping dataset. This achieves efficient feature extraction and accurate nonlinear fusion of multi-source remote sensing data, enhancing the weight ratio of key geographic features and ensuring the comprehensiveness of mapping data. The system achieves unified accuracy; the dynamic terrain high-precision inversion module combines InSAR interferometry technology with time series analysis algorithms to complete terrain deformation monitoring, dynamically updates the terrain and landforms using terrain segmentation algorithms, and achieves high-precision reconstruction of 3D terrain and real-time monitoring of terrain deformation, providing timely results for terrain change analysis; the real-time error correction module identifies system and random errors, and dynamically corrects the model and deformation results through a BP neural network optimized by the ISSA algorithm, enhancing the accuracy and reliability of terrain data; the intelligent visualization application module constructs a high-fidelity terrain digital twin simulation model and completes AR real-scene overlay, providing flexible interactive geographic information queries and enriching the application scenarios and display forms of surveying and mapping results; thus, it solves the problems of weak multi-source data fusion and low terrain inversion accuracy in existing technologies.

[0016] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0017] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the structure of a remote sensing-based mapping and geographic system according to an embodiment of this application; Figure 2This is a schematic diagram of a multimodal remote sensing data acquisition module according to an embodiment of this application; Figure 3 This is a schematic diagram of a dynamic data fusion and collaboration module provided according to an embodiment of this application; Figure 4 This is a schematic diagram of a dynamic terrain high-precision inversion module provided according to an embodiment of this application; Figure 5 This is a schematic diagram of a real-time error correction module according to an embodiment of this application; Figure 6 This is a schematic diagram of an intelligent visualization application module provided according to an embodiment of this application; Figure 7 This is a flowchart of a remote sensing-based mapping and geographic system provided according to an embodiment of this application; Figure 8 This is a flowchart illustrating a method for mapping a geographic system based on remote sensing technology according to an embodiment of this application; Figure 9 This is a schematic diagram of a method for mapping a geographic system based on remote sensing technology according to an embodiment of this application; Figure 10 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The following description, with reference to the accompanying drawings, illustrates a remote sensing-based mapping and geographic system according to an embodiment of this application. Addressing the low accuracy of terrain inversion mentioned in the background section, this application provides a remote sensing-based mapping and geographic system. In this system, a multimodal remote sensing data acquisition module simultaneously acquires satellite remote sensing imagery, aerial LiDAR point clouds, UAV oblique photography, and ground-based IoT sensor data. This overcomes the limitations of traditional single-source data coverage, expands the acquisition dimensions and spatial coverage of mapping data, ensures the comprehensiveness of terrain information representation, and provides a multi-dimensional data foundation for subsequent mapping analysis. Furthermore, a data fusion and collaboration module, using an improved Mamba fusion model, extracts image texture, point cloud geometry, and terrain topology features from multiple sources in parallel. A selective state-space gating mechanism strengthens the weights of key geographic features, performing nonlinear fusion and generating a unified-precision mapping dataset. This achieves efficient feature extraction and accurate nonlinear fusion of multi-source remote sensing data. The system integrates multiple data sources to enhance the weighting of key geographic features and ensure consistent accuracy of surveying and mapping datasets. A dynamic high-precision terrain inversion module combines InSAR interferometry with time-series analysis algorithms to monitor terrain deformation. It dynamically updates the terrain using terrain segmentation algorithms, achieving high-precision 3D terrain reconstruction and real-time monitoring of terrain deformation, providing timely results for terrain change analysis. A real-time error correction module identifies system and random errors and dynamically corrects the model and deformation results using a BP neural network optimized with the ISSA algorithm, enhancing the accuracy and reliability of terrain data. An intelligent visualization application module constructs a high-fidelity digital twin simulation model of the terrain and performs AR overlay, providing flexible interactive geographic information queries and enriching the application scenarios and display formats of surveying and mapping results. This solves the problems of weak multi-source data fusion and low terrain inversion accuracy in existing technologies. Figure 1 This is a schematic diagram of the structure of a surveying and mapping geographic system based on remote sensing technology, provided as an embodiment of this application.

[0020] This application provides a mapping and geographic system based on remote sensing technology. The system 10 includes: Multimodal remote sensing data acquisition module 100, data fusion and collaboration module 200, dynamic terrain high-precision inversion module 300, real-time error correction module 400, and intelligent visualization application module 500.

[0021] The multimodal remote sensing data acquisition module 100 is used to simultaneously acquire satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data, and ground IoT sensor data; the data fusion and collaboration module 200 is used to perform geographic coordinate calibration and location encoding on satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data, and ground IoT sensor data, and to perform feature extraction and nonlinear fusion using an improved Mamba fusion model. It also strengthens the weights of key geographic features through a selective state-space gating mechanism to generate a mapping dataset with uniform accuracy; the dynamic terrain high-precision inversion module 30... The system performs 3D terrain reconstruction based on surveying datasets, combines InSAR interferometry technology with time series analysis algorithms to monitor terrain deformation, and uses terrain segmentation algorithms for dynamic updates of terrain features; the real-time error correction module 400 identifies systematic and random errors generated in the terrain inversion process of the surveying dataset, and constructs a BP neural network optimized by the ISSA algorithm to dynamically correct the 3D terrain model and terrain deformation results; the intelligent visualization application module 500 constructs a digital twin simulation model, performs AR real-scene overlay, provides interactive geographic information queries, and automatically generates standardized surveying results reports.

[0022] It is understood that, in this embodiment, the multimodal remote sensing data acquisition module simultaneously acquires satellite remote sensing images, aerial LiDAR point clouds, UAV oblique photography, and ground IoT sensor data, breaking through the traditional single-source data coverage limitations, expanding the acquisition dimensions and spatial coverage of mapping data, ensuring the comprehensiveness of terrain information representation, and providing a multi-dimensional data foundation for subsequent mapping analysis. With the help of the data fusion and collaboration module, through an improved Mamba fusion model, image texture, point cloud geometry, and terrain topology features of multi-source data are extracted in parallel. A selective state-space gating mechanism is used to strengthen the weight of key geographic features, performing nonlinear fusion and generating a unified precision mapping dataset. This achieves efficient feature extraction and accurate nonlinear fusion of multi-source remote sensing data, strengthening the weight ratio of key geographic features, and ensuring... The system achieves unified accuracy in obstacle mapping datasets; a dynamic terrain high-precision inversion module combines InSAR interferometry technology with time series analysis algorithms to monitor terrain deformation, dynamically updates terrain features using terrain segmentation algorithms, and achieves high-precision 3D terrain reconstruction and real-time monitoring of terrain deformation, providing timely results for terrain change analysis; a real-time error correction module identifies system and random errors, and dynamically corrects the model and deformation results through a BP neural network optimized by the ISSA algorithm, enhancing the accuracy and reliability of terrain data; an intelligent visualization application module constructs a high-fidelity terrain digital twin simulation model and completes AR real-scene overlay, providing flexible interactive geographic information queries and enriching the application scenarios and display formats of surveying and mapping results; thus, it solves the problems of weak multi-source data fusion and low terrain inversion accuracy in existing technologies. In this embodiment of the application, the multimodal remote sensing data 100 includes: such as Figure 2 As shown, there are satellite image acquisition unit, LiDAR point cloud acquisition unit, oblique photography acquisition unit, and ground sensor acquisition unit.

[0023] The satellite image acquisition unit is used to interface with a multi-resolution satellite remote sensing platform to simultaneously acquire multispectral and panchromatic satellite image data; the LiDAR point cloud acquisition unit is used to acquire high-precision point cloud data of the terrain surface through airborne LiDAR equipment; the oblique photography acquisition unit is used to acquire oblique photography images of the terrain and landforms through multi-view cameras of UAVs; and the ground sensor acquisition unit is used to acquire data on terrain surface temperature, humidity, and settlement monitoring points through deployed IoT sensors.

[0024] It is understood that the embodiments of this application connect to a multi-resolution satellite remote sensing platform through a satellite image acquisition unit to simultaneously acquire multispectral and panchromatic satellite image data, providing large-scale terrain contour and surface cover basic information for topographic mapping, meeting the need for capturing overall terrain features at a macro scale; through a LiDAR point cloud acquisition unit, high-precision point cloud data of the terrain surface is accurately acquired using airborne LiDAR equipment, providing accurate three-dimensional spatial data support for subsequent terrain elevation calculation and undulation feature analysis; through an oblique photography acquisition unit, oblique photography images of the terrain and landforms are meticulously acquired using a multi-view camera from an unmanned aerial vehicle, clearly restoring the detailed textures of surface buildings, vegetation, etc., filling the gap in the microscopic feature representation of macroscopic images; and through a ground sensing acquisition unit, IoT sensors are deployed to acquire real-time data on terrain surface temperature, humidity, and settlement monitoring points, supplementing topographic mapping with dynamic information on surface environment and settlement changes.

[0025] For example, in the mountainous watershed ecological protection and topographic monitoring project, the multi-source data acquisition phase fully leveraged the capabilities of each unit: the satellite image acquisition unit connected to the Gaofen series multi-resolution remote sensing platform to simultaneously acquire multispectral and panchromatic band data, clearly presenting the vegetation cover distribution and macroscopic topographic contours within a 50km² area of ​​the watershed; the LiDAR point cloud acquisition unit used laser scanning equipment mounted on a fixed-wing aircraft to collect topographic surface point clouds at a density of 10 points / square meter, accurately capturing elevation information such as valley depth and slope undulation; the oblique photography acquisition unit used a multi-rotor UAV equipped with a five-lens camera to complete more than 8,000 image acquisitions at a flight altitude of 300m, meticulously restoring the texture of riverbank vegetation and the details of village buildings; and the ground sensor acquisition unit deployed dozens of IoT sensors around landslide hazard points and hydrological observation stations within the watershed, transmitting real-time surface temperature and humidity data and settlement monitoring information at key locations. These four types of data complemented each other from different dimensions, providing comprehensive and accurate raw data support for subsequent data collection.

[0026] In this embodiment of the application, the data fusion and collaboration module 200 includes: Figure 3 As shown, the data consists of a geocoding unit, a feature extraction unit, a fusion processing unit, and a dataset output unit.

[0027] The system comprises the following components: a geocoding unit for calibrating multi-source remote sensing data by geographic coordinates, introducing geographic feature location codes to identify spatial attributes of the data, and generating feature vectors containing location information; a feature extraction unit for inputting feature vectors into an improved Mamba model, and extracting image texture, point cloud geometry, and terrain topology features in parallel through a linear attention mechanism; a fusion processing unit for nonlinearly fusing image texture, point cloud geometry, and terrain topology features by strengthening the weights of key geographic features through a selective state-space gating mechanism; and a dataset output unit for standardizing the fused feature data to generate a mapping dataset with uniform accuracy.

[0028] It is understood that the embodiments of this application perform geographic coordinate calibration on multi-source remote sensing data through a geocoding unit and introduce geographic feature location encoding to avoid "distortion" in subsequent fusion due to coordinate misalignment, generating feature vectors containing location information to ensure that each type of data corresponds to the real terrain spatial location; the feature extraction unit inputs the feature vectors into the linear attention mechanism of the improved Mamba model to extract image texture, point cloud geometry, and terrain topology features in parallel to avoid missing key terrain features; the fusion processing unit strengthens the weight of key geographic features through a selective state space gating mechanism to avoid interference from redundant information and performs nonlinear fusion so that the fused data retains the advantages of various features and forms an organic whole; the standardization processing of the dataset output unit ensures the uniformity of the accuracy of the surveying and mapping dataset, provides data support for subsequent dynamic terrain inversion, and avoids the inversion results being affected by differences in data format or accuracy.

[0029] It should be noted that the geocoding unit uses the WGS84 coordinate system as the reference. Four evenly distributed ground control points are selected in the survey area, and WGS84 latitude, longitude, and geodetic coordinates with an accuracy of ±0.05m are obtained through GNSSRTK static measurement. The RANSAC algorithm is used to complete the coordinate matching of multi-source data, and outliers with matching errors greater than 0.3m are removed, so that the plane error of the calibrated data is ≤0.5 pixels and the elevation error is ≤0.1m. The geographic feature location encoding adopts a structure of hemispherical code + latitude and longitude zone code + 10m grid coordinate code + geodetic height code, which is concatenated with the inherent features of the data to form a 64-dimensional feature vector. The first 30 bits are the location code and the last 34 bits are the original data features. Before output, the control point coordinates are verified to ensure that the deviation between the grid position and the actual terrain is ≤10m. The improved Mamba model refers to a customized deep learning model based on a state-space model, employing a selective state-space architecture and processing long-sequence data with linear computational complexity. It also optimizes the linear attention mechanism to adapt to the feature extraction requirements of multi-source remote sensing data and adds a selective state-space gating mechanism to achieve weighted nonlinear fusion of geographic features. The formula is:

[0030]

[0031]

[0032]

[0033]

[0034]

[0035]

[0036]

[0037] in, This is the normalized 64-dimensional feature vector; For the layer normalization function, use Numerical stability parameters; The feature vector output by the geocoding unit; It is a linear attention weight matrix; For row-normalized exponential functions; It is a linear attention activation function; For attention projection matrix; For attention projection matrix; This is a matrix transpose operation; The scaling factor is the square root of the feature dimension; For extracted image texture features; Extract weight matrices for texture features; Output matrix for texture features; The extracted point cloud geometric features; Extract weight matrices for geometric features; Output matrix for geometric features; For extracted terrain topological features; Extract weight matrices for topological features; Output matrix for topological features; This represents the gating coefficient vector for key geographic features; For activation functions; It is a two-layer perceptron; This is the weighted feature splicing result; for Texture feature gating coefficients split by dimension; for Geometric feature gate coefficients split by dimension; for Topological feature gating coefficients split by dimension; The final geographical features after integration; The activation function for the Gaussian error linear unit; For the fusion weight matrix; This is the fusion bias vector.

[0038] For example, taking a preliminary surveying and mapping project for a highway construction in the southwestern mountainous area as an example, this scenario requires the accurate integration of multi-source data to support highway alignment and slope stability analysis, and the data fusion and collaboration module plays a core role in this. First, staff simultaneously acquired 1m resolution satellite imagery, aerial LiDAR point cloud data, UAV oblique photography data, and ground subsidence sensor data for the survey area. The module then initiated full-process processing: the geocoding unit, using the WGS84 coordinate system as a reference, selected four evenly distributed ground control points, obtained ±0.05m accuracy coordinates via GNSSRTK, used the RANSAC algorithm for coordinate matching and eliminated error points greater than 0.3m, and combined hemispherical and grid coding to generate a 64-dimensional feature vector containing location information; the feature extraction unit input the vector into an improved Mamba model, using a linear attention mechanism to extract surface texture from satellite imagery, elevation geometry from LiDAR point clouds, and topographic features from oblique photography in parallel; the fusion processing unit, through a selective gating mechanism, strengthened the elevation and topographic feature weights of slope areas while weakening redundant information from vegetation occlusion; the dataset output unit, after standardization, generated a unified precision mapping dataset with a planar error ≤0.5 pixels, fully preserving key information such as terrain undulations and geological structures along the highway, providing reliable data for subsequent route optimization and disaster risk assessment.

[0039] In this embodiment of the application, the dynamic terrain high-precision inversion module 300 includes: as follows Figure 4 As shown, there are three-dimensional reconstruction units, deformation monitoring units, and dynamic update units.

[0040] The system comprises the following components: a 3D reconstruction unit, which uses LiDAR point cloud and oblique photogrammetric image data from the mapping dataset to construct a basic 3D terrain model after denoising, registration, 3D grid construction, and texture bonding; a deformation monitoring unit, which uses multiple periods of InSAR imagery and LiDAR elevation data from the mapping dataset to register, generate interferograms, and unwrap phases of the InSAR images using InSAR interferometry technology, extracting initial deformation phase information of the terrain, and performing time-series correlation analysis on the deformation phase information of multiple periods using time-series analysis algorithms to calculate the long-term deformation rate and cumulative deformation value of the terrain, while comparing and verifying with ground sensor data to output accurate terrain deformation results; and a dynamic update unit, which identifies areas of terrain change using terrain segmentation algorithms based on the deformation results, dynamically optimizes the basic 3D terrain model to obtain a high-precision 3D terrain model, and synchronously updates the terrain deformation time-series data.

[0041] It is understood that the 3D reconstruction unit in this application calls LiDAR point cloud and oblique photogrammetric images from the mapping dataset, and after denoising and registration, 3D grid construction and texture fitting, provides a benchmark template for subsequent deformation monitoring; the deformation monitoring unit calls multiple periods of InSAR images and LiDAR elevation data, and completes image registration, interferogram generation and phase unwrapping through InSAR interferometry technology. Combined with time series analysis algorithms, it correlates multiple periods of phase data to calculate the long-period deformation rate and cumulative deformation value. At the same time, it compares and verifies with ground sensor data to ensure the accuracy of deformation results and avoid the bias of single monitoring; the dynamic update unit accurately identifies the terrain change area through terrain segmentation algorithm based on the deformation results. It can not only efficiently optimize the basic terrain 3D model to obtain a high-precision version, but also update the deformation time series data simultaneously, so that the terrain inversion always fits the real terrain changes, providing dynamic and accurate inversion results for subsequent error correction.

[0042] It should be noted that the 3D reconstruction unit retrieves LiDAR point cloud and UAV oblique photogrammetric image data from the surveying dataset. After completing the denoising and registration preprocessing, the grid resolution is set according to the terrain complexity and surveying accuracy requirements of the survey area. An irregular triangular mesh (TIN) is used to build a 3D terrain grid skeleton. The grid elevation information is completed by Kriging interpolation algorithm to restore the terrain undulation and spatial structure. Subsequently, the texture information of the oblique photogrammetric image is extracted, and the texture is accurately attached to the 3D grid surface according to the mapping relationship between the image projection coordinates and the grid spatial coordinates. Color balancing and seam smoothing are performed at the texture splicing points. Finally, a basic 3D terrain model with both geometric accuracy and texture effect is constructed, which fully reproduces the 3D spatial characteristics of the survey area terrain.

[0043] Kriging interpolation is a statistical interpolation method based on the principle of spatial autocorrelation. It estimates the elevation of unknown points by weighting discrete terrain sampling points in a survey area. The formula is:

[0044]

[0045]

[0046] in, Unknown points in the terrain grid Elevation estimate; The sequence number of the known sampling point; The number of known sampling points in the surrounding LiDAR point cloud used for interpolation; For the first Weighting coefficients for each known sampling point; For the first Measured elevation values ​​of one known sampling point; The spatial location of an unknown point in a terrain grid; For the first The spatial location of a known sampling point; The sum of the weight coefficients representing all known sampling points is 1; Another index for the known sampling points, used for... From 1 to The items are accumulated; For the first Weighting coefficients for each known sampling point; For known sampling points and Covariance between them; These are Lagrange multipliers used to satisfy unbiasedness constraints; For known sampling points and the unknown Covariance between them; For the first The spatial location of a known sampling point.

[0047] When the deformation monitoring unit is working, it first retrieves multiple periods of InSAR imagery and LiDAR elevation data from the mapping dataset. Using InSAR interferometry technology, it first completes the spatial registration of the multiple periods of InSAR imagery through a feature point matching algorithm, and then generates an interferogram through filtering. Next, it performs phase unwrapping operation using the minimum cost flow method to extract the initial deformation phase information of the terrain. Subsequently, it uses a time series analysis algorithm to screen effective scatterers and perform time series correlation analysis on the deformation phase information of the multiple periods. Through fitting calculation, it obtains the long-term deformation rate and cumulative deformation value of the terrain. Finally, it verifies the above calculation results with ground sensor data for errors, and finally outputs the terrain deformation results containing core deformation parameters and accuracy indicators.

[0048] Feature point matching algorithm formula:

[0049]

[0050]

[0051]

[0052] in, It is a scale space function; The horizontal coordinate of the image pixel; The spatial ordinate of the image pixels; is the scale factor of the Gaussian kernel; It is a two-dimensional Gaussian kernel function; The pixel grayscale value of the InSAR image; The values ​​are Gaussian difference scale space values; The scaling factor for adjacent scale layers; For scale The scale space value below; The angle of the principal direction of the feature point; The gradient magnitude of the pixels in the neighborhood of the feature point; The gradient direction of pixels within the neighborhood of the feature point; It is the arctangent function; It is a sine function; It is a cosine function; The Euclidean distance between two feature point descriptors; The dimension number of the 128-dimensional SIFT feature descriptor; The first feature point of the reference image dimensional descriptor; The first feature point of the image to be registered Dimensional descriptor.

[0053] Minimum cost flow formula:

[0054]

[0055]

[0056]

[0057] in, For pixels Phase gradient at; The horizontal coordinate of the image pixel; The spatial ordinate of the image pixels; For pixels in the interferogram The entanglement phase value; For the edge of the network The cost; For edges in the network; For the edge The corresponding phase difference; For the edge The flow on; A node in the network; For inflow node The set of edges; outflow node edge set For nodes Supply and demand; The goal is to minimize costs; Let be the set of all edges in the network.

[0058] Time series analysis algorithm formula:

[0059]

[0060]

[0061]

[0062] in, For the first Interference phase of periodic images; This refers to the image period number; This refers to the image period number; For the first Deformation phase components corresponding to the periodic image; For the first Topographic phase components corresponding to the periodic images; For the first Noise phase components corresponding to the periodic images; The rate of terrain deformation; For the time baseline matrix; This is the matrix transpose symbol; The deformation phase vector; for The cumulative terrain deformation value at any given time; For monitoring time; For monitoring time intervals; The interference coherence coefficient; The pixel index within the window; This represents the number of pixels within the window. For the baseline period image The grayscale value of each pixel; For the first image to be registered The conjugate of the grayscale values ​​of each pixel.

[0063] For example, taking a dynamic topographic monitoring project in an open-pit coal mine as an example, the 3D reconstruction unit first calls up the LiDAR point cloud (point cloud density 25 points / ㎡) and oblique photogrammetric image (spatial resolution 3cm) of the area. After radius filtering and noise reduction, and point cloud-image co-registration based on feature matching, a 2m resolution 3D grid is constructed and fitted with high-fidelity textures to form a basic 3D topographic model covering the entire mining area and surrounding buffer zone. The deformation monitoring unit simultaneously calls up multiple periods of InSAR imagery and LiDAR elevation data, and uses InSAR interferometry technology to complete image registration, interferogram generation, and phase unwrapping. It extracts the initial deformation phase information of the terrain, and combines time series analysis algorithms to perform time-series correlation analysis on the phase information of multiple periods to obtain the long-term deformation rate and cumulative deformation value of the terrain in the mining area. Then, it compares and verifies the data with the ground sensor data deployed in the mining area, and outputs accurate terrain deformation results. Based on the deformation results, the dynamic update unit identifies multiple areas of terrain and landform change, such as mining slopes and spoil heaps, through terrain segmentation algorithms. It then optimizes the geometric and texture information of the changed areas in the basic 3D terrain model to generate a high-precision 3D terrain model. Simultaneously, it updates the terrain deformation time series data of the area, providing intuitive and accurate data support for mine slope stability assessment and safe production scheduling.

[0064] In this embodiment of the application, the real-time error correction module 400 includes, as follows: Figure 5 As shown, there are an error identification unit, a model optimization unit, and a correction execution unit.

[0065] The error identification unit is used to identify systematic and random errors generated in the topographic inversion process of the surveying dataset through statistical analysis and feature comparison; the model optimization unit is used to optimize the initial weights and structural parameters of the BP neural network using the ISSA algorithm to improve the accuracy of model correction; the correction execution unit is used to input the error data into the optimized BP neural network to dynamically correct the high-precision 3D topographic model and topographic deformation results.

[0066] It is understood that the error identification unit in this application embodiment accurately distinguishes between systematic errors and random errors through statistical analysis and feature comparison, providing a clear target for subsequent correction; the model optimization unit uses the ISSA algorithm to optimize the initial weights and structural parameters of the BP neural network, solving the problem that the BP neural network is prone to getting trapped in local optima and the instability of correction accuracy caused by random initial weights. Through the global search capability of the ISSA algorithm, a better parameter configuration is found for the neural network, improving the model's adaptability to different types of errors; the correction execution unit inputs the error data into the optimized BP neural network, dynamically corrects the high-precision 3D terrain model and terrain deformation results, ensuring the accuracy and reliability of the entire surveying and mapping output results, and providing a high-quality data foundation for intelligent visualization applications.

[0067] It should be noted that the chaotic mapping of the model optimization unit improves the uniformity of population initialization and the search capability of dynamic inertia weight balance. The fitness function is designed with the prediction error of the BP neural network as the optimization target. The method for determining the structural parameters of the three-layer BP network is clarified. The weight update is achieved through ISSA population iteration and BP gradient descent, which solves the problems of slow convergence and easy local optima in traditional algorithms.

[0068] ISSA algorithm formula:

[0069]

[0070] in, For the first The first sparrow Dimensional position; These are the parameters for controlling chaos. This is the position of the previous iteration; Inertial weight; This represents the upper limit of the weight, with a value of 0.9. This is the lower limit of the weight, with a value of 0.4; This represents the current iteration number; This represents the maximum number of iterations.

[0071] Fitness function design formula:

[0072] in, To optimize the objective; The number of samples; This represents the true value of the error. This represents the error value predicted by the BP neural network.

[0073] Formula for determining the structure parameters of a BP neural network:

[0074] in This represents the number of hidden layer nodes. This represents the number of nodes in the input layer. This represents the number of nodes in the output layer. It is an empirical constant ranging from 1 to 10 (taken as 5).

[0075] Formula for weight update iteration rule:

[0076]

[0077] in, For the first The position of the sparrow in the iteration; This is the current iteration position; A random number between 0 and 1; T is the maximum number of iterations; The serial number of the sparrow This is the optimal position for the current population. Inertial weight; For the first Iterative network weights; The current weight; The learning rate is set to 0.01. For error Partial derivatives with respect to the weights; The momentum factor is 0.9. This represents the weight change amount in the previous iteration.

[0078] For example, taking a large open-pit iron ore mining area terrain dynamic monitoring project as an example, the error identification unit first retrieves the LiDAR point cloud (density 30 points / ㎡), multiple periods of Sentinel-1 InSAR images, and measured elevation data of the mining area. Through statistical analysis, it calculates the mean (128.7m) and standard deviation (±9.2mm) of the point cloud elevation data, identifying the systematic phase deviation of the InSAR image caused by mining dust interference. Then, through feature comparison, it extracts terrain features such as the slope of the mining area (average 42°) and the elevation of the spoil heap platform, and compares them with the baseline features of the previous month to locate the local terrain texture anomalies caused by random noise induced by blasting vibration. The model optimization unit then uses the ISSA algorithm to optimize the BP neural network, using the error value as the fitness function, and iteratively adjusts the initial weights and the number of hidden layer nodes of the network (finally determining a 4-11-1 structure), which improves the model convergence speed by 38%. The correction execution unit inputs the identified systematic errors (mean deviation 6.8mm) and random errors into the optimized BP neural network, outputting targeted corrections to dynamically adjust the elevation and texture information of the bottom depression area and the edge of the spoil heap in the basic terrain 3D model. At the same time, it corrects the temporal deformation rate data. After correction, the model is compared with the measured data in the mining area. The elevation error is reduced from ±9.2mm to ±2.3mm, and the deformation rate accuracy is improved to ±1.6mm / month, providing accurate data support for mine slope stability assessment, mining progress planning, and safety protection scheme formulation.

[0079] In this embodiment of the application, the intelligent visualization application module 500 includes, for example: Figure 6 As shown, there are digital twin construction unit, AR overlay unit, query interaction unit, and report generation unit.

[0080] The digital twin construction unit, based on a high-precision 3D terrain model, terrain deformation results, and surveying datasets, constructs a 1:1 mapping of the physical terrain into a digital twin simulation model, synchronously linking terrain deformation time-series data with land cover characteristics. The AR overlay unit is used to overlay the digital twin simulation model with real-world images, achieving real-time alignment and interactive browsing between the virtual model and the real terrain. The query and interaction unit provides multi-condition interactive query functions for geographic information. The report generation unit integrates surveying data, inversion results, and error correction information through structured algorithms to generate standardized surveying results reports.

[0081] Understandably, the digital twin construction unit of this application constructs a simulation model that maps to the physical terrain in a 1:1 ratio based on a high-precision 3D terrain model, deformation results, and surveying dataset. It also synchronously links deformation time-series data with land cover characteristics, enabling the model to carry dynamic deformation history and land surface attribute information, becoming a digital mirror of the terrain. This provides intuitive and comprehensive digital evidence for subsequent terrain change trend analysis and disaster risk prediction scenarios. The AR overlay unit aligns and overlays the digital twin model with real-world images in real time, allowing for interactive browsing of the virtual model and the real terrain, facilitating rapid location of key terrain areas. The query and interaction unit provides multi-condition interactive query functions, allowing users to accurately filter information based on terrain location, deformation period, land surface type, and other conditions. The report generation unit integrates surveying data, inversion results, and error correction information through structured algorithms to generate standardized surveying results reports. These reports not only have a unified format for easy archiving but also clearly present the complete chain, making the application of surveying results more evidence-based and reducing the difficulty of understanding for non-professionals.

[0082] It should be noted that the formula for constructing a digital twin is:

[0083]

[0084]

[0085]

[0086]

[0087]

[0088] in, The coordinates of terrain points in the target unified coordinate system; These are the three-dimensional coordinate components of the target coordinate system; This is a seven-parameter coordinate transformation matrix; The coordinates of the terrain points in the original coordinate system of the survey data; These are the three-dimensional coordinate components of the original coordinate system; The root mean square error of the digital twin model; This represents the total number of accuracy verification points. For the first in the digital twin model The coordinates of each verification point; For the first GNSS measured coordinates of each verification point; The serial number of the verification point; for The optimal correlation between time-varying deformation data and model partitions; This is a function to find the maximum value. This is the function for calculating the correlation coefficient (value range [-1, 1]). for Temporal data of terrain deformation at any given time; For the first digital twin model One geographical region; For timestamps; This represents the sequence number of the geographic partition of the model; The rotation matrix of the virtual model relative to the real scene; This is the translation vector of the virtual model relative to the real scene; This is a pose calculation algorithm based on feature points; The world coordinates of the feature points in the digital twin model; These are the pixel coordinates of the corresponding feature points in the real-world image; The index of the feature point; This is the coordinate deviation correction amount for AR overlay; For adaptive correction coefficients; The real-time coordinates of the digital twin model after AR overlay; The reference coordinates for the terrain feature points; The overall matching degree between query conditions and geographic information data; This represents the total number of query conditions. For the first The weight of each query condition; A function for calculating the matching degree; For the first Individual user query criteria; For geographic information data and the first The attribute values ​​corresponding to each query condition; The index is the sequence number of the query condition.

[0089] For example, taking a smart terrain management project in a new urban area as an example, the digital twin construction unit constructs a digital twin simulation model of the terrain based on a high-precision 3D terrain model of the area (LiDAR acquisition accuracy of 5cm), terrain deformation results from surface subsidence monitoring (InSAR monitoring once a month, subsidence rate of 3mm / month in the municipal square area), and a surveying dataset (including information on underground pipe networks, building foundations, etc.). This model is mapped 1:1 to the physical terrain, and the monthly subsidence time series data is synchronously associated with the surface building, green space, and other coverage features. The AR overlay unit overlays this digital twin model with UAV real-scene images (resolution 0.1m) for AR display, using visual S... LAM technology enables real-time alignment between the virtual model and the real terrain (accuracy ≤ 2cm). On-site staff can view the correspondence between the underground pipeline network in the model and the actual ground. The query and interaction unit provides multi-condition interactive query functions for geographic information. For example, by entering "settlement > 5mm / month and underground pipeline burial depth < 3m", the risk area of ​​a certain road section can be quickly located. The report generation unit integrates surveying data, settlement inversion results and error correction information through structured algorithms to automatically generate a new area terrain deformation and pipeline safety assessment report, including settlement hotspot distribution map, pipeline risk level table and other content, providing accurate basis for urban management departments to formulate maintenance plans.

[0090] This application proposes a remote sensing-based mapping and geographic system. Through a multimodal remote sensing data acquisition module, it simultaneously acquires satellite remote sensing imagery, aerial LiDAR point clouds, UAV oblique photography, and ground-based IoT sensor data. This overcomes the limitations of traditional single-source data coverage, expands the acquisition dimensions and spatial coverage of mapping data, ensures the comprehensiveness of terrain information representation, and provides a multi-dimensional data foundation for subsequent mapping analysis. Utilizing a data fusion and collaboration module, and through an improved Mamba fusion model, it extracts image texture, point cloud geometry, and terrain topology features from multiple sources in parallel. A selective state-space gating mechanism strengthens the weights of key geographic features, performing nonlinear fusion and generating a unified precision mapping dataset. This achieves efficient feature extraction and accurate nonlinear fusion of multi-source remote sensing data, enhancing the representation of key geographic features. The weighting of data ensures consistent accuracy across the mapping dataset; the dynamic terrain high-precision inversion module combines InSAR interferometry with time-series analysis algorithms to monitor terrain deformation, dynamically updating topography using terrain segmentation algorithms to achieve high-precision 3D terrain reconstruction and real-time monitoring of terrain deformation, providing timely results for terrain change analysis; the real-time error correction module identifies system and random errors, dynamically correcting the model and deformation results through a BP neural network optimized by the ISSA algorithm, enhancing the accuracy and reliability of terrain data; the intelligent visualization application module constructs a high-fidelity digital twin simulation model of terrain and completes AR real-scene overlay, providing flexible interactive geographic information queries and enriching the application scenarios and display formats of mapping results; thus, it solves the problems of weak multi-source data fusion and low terrain inversion accuracy in existing technologies. The following will illustrate a mapping and geographic system based on remote sensing technology through a specific embodiment, such as... Figure 7 As shown, it includes: Using a dynamic monitoring project of slope topography along a highway in the southwestern mountainous region as an application scenario, the multimodal remote sensing data acquisition module fully activated the functions of each unit to carry out data acquisition. The satellite image acquisition unit pre-connected to multi-resolution satellite remote sensing platforms such as Gaofen-3 and Sentinel-2, and set synchronous acquisition commands according to the characteristics of the mountainous terrain to accurately acquire multispectral and panchromatic satellite image data covering the monitoring area. The panchromatic image resolution reached 0.8 meters, clearly capturing subtle features such as slope surface cracks and vegetation cover changes. The LiDAR point cloud acquisition unit, using a RIEGLVQ-1560i LiDAR device mounted on a general aviation aircraft, conducted a flight-planned scan of a 50-kilometer area along the highway at a flight altitude of 1000 meters and a point cloud density of 50 points / square meter, acquiring high-precision point clouds including slope surfaces, vegetation, and highway subgrade. The data, with elevation accuracy better than 5 centimeters, was collected using three drones equipped with five-lens cameras in a crisscross flight pattern. These drones, at an altitude of 200 meters, captured multi-view images of key road sections (such as steep slopes and bridge approach areas). Each flight acquired over 2,000 oblique images, with an overlap of over 80% to ensure modeling accuracy. The ground-based sensor acquisition unit deployed 120 IoT sensors at 500-meter intervals within the monitoring area to collect real-time data on slope surface temperature and humidity, as well as displacement information from 15 fixed settlement monitoring points. Data sampling was conducted hourly to capture subtle deformations caused by rainfall and soil weathering in the mountainous area. The data collected by each unit was aggregated through a unified interface to form a multi-source raw dataset containing images, point clouds, and sensor data, providing comprehensive data support for subsequent processing.

[0091] After receiving the raw data from multiple sources, the data fusion and collaboration module first performs coordinate unification by the geocoding unit. This unit uses the WGS84 coordinate system as a reference and calibrates data from different coordinate systems, such as satellite imagery and LiDAR point clouds, using a seven-parameter coordinate transformation method to eliminate spatial bias caused by coordinate heterogeneity. Simultaneously, it adds geographic feature location codes containing longitude, latitude, and elevation information to each type of data, transforming discrete data into standardized feature vectors containing spatial attributes. Subsequently, the feature extraction unit inputs these feature vectors into an improved Mamba model, utilizing the model's linear attention mechanism to process the multi-source data in parallel—focusing on extracting texture features such as slope vegetation cover and surface reflectivity from satellite imagery, extracting three-dimensional features such as slope gradient, aspect, and roadbed geometry from LiDAR point clouds, and extracting attribute features such as temperature change curves and humidity thresholds from ground sensor data, achieving efficient separation and extraction of multi-dimensional features. The fusion processing unit initiates a selective state-space gating mechanism, assigning weights to the extracted features based on the core requirements of highway monitoring. The weights of slope sliding interface features and roadbed settlement correlation features are increased to 0.3, while the weights of irrelevant vegetation noise features are reduced. A nonlinear fusion algorithm organically integrates multi-dimensional features, eliminating data redundancy and conflicts. Finally, the dataset output unit normalizes the fused feature data, uniformly calibrating the data accuracy to the centimeter level, generating a unified precision mapping dataset containing spatial and attribute information, providing high-quality data input for terrain inversion.

[0092] The dynamic terrain high-precision inversion module, based on the generated unified precision mapping dataset, carries out terrain inversion and updating work in stages. The 3D reconstruction unit retrieves LiDAR point cloud and oblique photogrammetric image data from the mapping dataset. After noise point removal preprocessing, grid parameters are set according to the terrain complexity and mapping accuracy requirements of the survey area. An irregular triangular mesh (TIN) is used to build a 3D terrain grid skeleton. The grid elevation information is completed by Kriging interpolation algorithm to restore the terrain undulation and spatial structure. Subsequently, the texture information of the oblique photogrammetric image is extracted, and the texture is accurately attached to the 3D grid surface according to the mapping relationship between the image projection coordinates and the grid spatial coordinates. Color balancing and seam smoothing are performed at the texture splicing points. Finally, a basic 3D terrain model with both geometric accuracy and texture effect is constructed, which fully reproduces the 3D spatial characteristics of the survey area terrain. When the deformation monitoring unit operates, it first retrieves multiple periods of InSAR imagery and LiDAR elevation data from the mapping dataset. A feature point matching algorithm is used to register the spatial positions of the multiple InSAR images. After filtering and generating interferograms, the minimum cost flow method is used to perform phase unwrapping to extract the initial deformation phase information of the terrain. Subsequently, a time series analysis algorithm is used to fit the trend of the deformation phase information across multiple periods, calculating the long-term deformation rate and cumulative deformation value of the slopes along the highway. Finally, this result is compared and verified with settlement monitoring data from ground sensors, outputting terrain deformation results containing core deformation parameters and accuracy indicators. The dynamic update unit initiates a terrain segmentation algorithm based on superpixel segmentation according to the deformation results, automatically identifying areas of slope slippage, soil and rock collapse, and other changes. It then performs mesh reconstruction and texture updates on the changed areas in the basic 3D terrain model, obtaining a high-precision 3D terrain model that closely matches the actual terrain. Simultaneously, the terrain deformation time series database is updated to ensure data timeliness.

[0093] The real-time error correction module initiates a correction process to address data deviation issues during terrain inversion. The error identification unit first processes the mapping dataset and inversion results, calculating the mean, standard deviation, and probability distribution of the LiDAR point cloud elevation data through statistical analysis. This identifies the overall elevation shift (mean deviation 8 mm) caused by systematic errors in the LiDAR equipment. Simultaneously, by calculating the residuals between InSAR deformation data and ground sensor data, it locates local errors caused by random noise. The feature comparison unit extracts the slope topographic geometric features (such as slope and slip surface depth) from the inversion model and compares them one by one with the standard features in the historical benchmark feature database. This reveals feature matching deviations caused by differences in data acquisition time periods, further clarifying the error type and distribution range. The model optimization unit then activated the ISSA algorithm to optimize the BP neural network. Using the error identification result as the fitness function, the initial weights and the number of hidden layer nodes of the BP neural network were iteratively adjusted based on the exploration and following behavior of the sparrow population. After 50 iterations, the optimal network structure was determined to be 4-12-1, which improved the model convergence speed by 35%. The optimized BP neural network significantly improved the fitting accuracy of the error and could accurately map the correlation between the error and the terrain parameters. The correction execution unit input the identified systematic error and random error data into the optimized BP neural network. The network output targeted elevation correction and deformation correction coefficients, which dynamically adjusted the roadbed elevation values ​​in the high-precision terrain 3D model. At the same time, it corrected outliers in the terrain deformation time series data. After correction, the model was compared with the measured data, and the elevation error was reduced from ±8 mm to ±1.8 mm. The deformation monitoring accuracy was further improved, providing reliable data for highway safety assessment.

[0094] The intelligent visualization application module transforms processed high-precision data into intuitive and usable monitoring results. The digital twin construction unit, based on a corrected high-precision 3D terrain model, integrates terrain deformation time-series data, land cover characteristics, and highway engineering attribute data. Using the Unity3D engine, it constructs a 1:1 mapping of the physical terrain into a digital twin simulation model. The model clearly displays the spatial relationship between the highway subgrade and slopes, and allows users to trace the terrain deformation process at any time using a timeline control. Vegetation, soil, and rock elements in the land cover features are precisely associated with the model; clicking on the corresponding area displays detailed attribute information. The AR overlay unit uses GPS and visual SLAM technology on mobile terminals to overlay the digital twin simulation model with real-world images in real time. A feature point matching algorithm achieves precise alignment between the virtual model and the real terrain, with an alignment accuracy of 3 centimeters. Highway maintenance personnel can intuitively see virtual annotations of slope slippage trends on a tablet computer, enabling interactive browsing of the virtual model and the real terrain. The query interaction unit features a multi-condition combination query function. Staff can input combinations such as "slope deformation > 8 mm + soil moisture > 60%", and the system quickly retrieves and highlights risk areas that meet the conditions, while simultaneously outputting detailed monitoring data and deformation trend analysis for that area. The report generation unit automatically integrates surveying data, inversion results, and error correction information using a structured algorithm, generating standardized surveying results reports according to a fixed structure of "project overview—data acquisition description—topographic inversion results—error analysis—safety assessment". The reports include screenshots of the 3D topographic model, deformation trend curves, and risk level classification tables, directly providing accurate data support and decision-making basis for slope maintenance and risk early warning by highway construction and management departments.

[0095] In summary, this embodiment acquires multi-dimensional data such as satellite imagery and LiDAR point clouds through a multi-modal remote sensing data acquisition module. A data fusion and collaboration module calibrates the data using the WGS84 coordinate system and fuses it through an improved Mamba model to generate a unified dataset. A dynamic inversion module completes 3D modeling and deformation monitoring. A real-time correction module accurately identifies and corrects errors. An intelligent visualization module outputs a twin model and a standardized report. This significantly improves monitoring accuracy and efficiency, providing reliable data support and decision-making basis for highway slope maintenance and risk early warning.

[0096] Next, referring to the accompanying drawings, a method for a surveying and mapping geographic system based on remote sensing technology is described according to an embodiment of this application.

[0097] like Figure 8 As shown, the method for a mapping and geographic system based on remote sensing technology includes the following steps: In step S101, satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data, and ground IoT sensor data are acquired simultaneously.

[0098] It is understood that the embodiments of this application achieve multi-dimensional terrain information coverage from macro to micro and from air to ground by simultaneously acquiring four types of multi-source data: satellite remote sensing imagery, aerial LiDAR point clouds, UAV oblique photography, and ground-based IoT sensing. Satellite data provides a wide-ranging macroscopic perspective, LiDAR and oblique photography supplement three-dimensional structure and texture details, and ground sensing enhances local dynamic monitoring. The complementary advantages of multi-source data provide multimodal raw information support for subsequent geographic coordinate calibration, feature fusion, and other processes, ensuring that the surveying and mapping geographic system has a comprehensive perception and accurate depiction of the terrain.

[0099] In step S102, the satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data and ground IoT sensor data are calibrated and encoded in geographic coordinates. By improving the Mamba fusion model, image texture, point cloud geometry and terrain topology features are extracted in parallel. Then, nonlinear fusion is performed through a selective state space gating mechanism to generate a mapping dataset with uniform accuracy.

[0100] Among them, the mapping dataset refers to a high-precision dataset containing multimodal features of satellite remote sensing images, aerial LiDAR point clouds, UAV oblique photography and ground IoT sensor data that have been calibrated with geographic coordinates and coded with location, which is used for terrain 3D reconstruction, deformation monitoring and geographic information visualization.

[0101] It is understood that the embodiments of this application generate a mapping dataset with uniform accuracy by performing geographic coordinate calibration, location encoding and feature fusion on multi-source remote sensing data. This integrates the scattered multi-source raw data into a standardized set containing image texture, point cloud geometry and terrain topology features, supporting the operation of modules such as dynamic terrain inversion, real-time error correction and intelligent visualization, improving the consistency and reliability of mapping data, effectively supporting terrain management and disaster early warning scenarios, and promoting the efficiency improvement of mapping and geographic systems in engineering construction and geographic monitoring.

[0102] In step S103, a basic three-dimensional terrain model is constructed based on the mapping dataset. The terrain deformation results are obtained by combining InSAR interferometry technology and time series analysis algorithm. The terrain segmentation algorithm is used to identify the terrain and landform change areas. The basic three-dimensional terrain model is optimized based on the terrain deformation results and change area information to obtain a high-precision three-dimensional terrain model. The terrain deformation time series data is updated synchronously.

[0103] Among them, the terrain segmentation algorithm, based on terrain deformation results and terrain features, divides the survey area's terrain into different units through spatial clustering or threshold division, thereby accurately identifying areas where terrain and landforms have changed. This provides a regional positioning basis for the dynamic optimization of the basic 3D terrain model. The formula is:

[0104]

[0105]

[0106]

[0107] in, For terrain grid pixels Deformation screening results; The horizontal coordinate of the terrain grid pixel; The spatial ordinate of the terrain grid pixel; For pixels Cumulative topographic deformation at the location; The deformation threshold is set based on the topographic stability of the survey area; To indicate that the pixel belongs to the deformation exceeding the threshold region, M=1; N=0 is the value used to determine whether the pixel belongs to a non-deformation region. This is the final determination result for the terrain change area; For pixels The current terrain slope value at the location; For pixels The original terrain slope value at the location; This is the threshold for slope variation; A set of seed points for growth in areas of terrain change; For pixels Deformation gradient value at; This represents the maximum value of the topographic deformation gradient within the survey area; For growth point With seed point The similarity determination results; This serves as an identifier for the growth point; This serves as the identifier for the seed point; P=1 is the criterion value representing the similarity of the deformation characteristics of the two and their potential for merging and growth. Q=0 is the value representing the determination that the two deformation characteristics are dissimilar and therefore cannot grow. This represents the cumulative deformation value of the growth point. This represents the cumulative deformation value of the seed point; The threshold for determining deformation similarity.

[0108] It is understood that the embodiments of this application, through terrain segmentation algorithms, can accurately identify specific areas where terrain and landforms have changed based on terrain deformation results, clarify the optimization range of the basic terrain 3D model, and improve the efficiency and pertinence of dynamic model optimization. At the same time, it accurately defines the boundaries and ranges of the changed areas, so that the optimization of the basic model focuses only on the actual deformation areas, ensuring the matching degree between the high-precision terrain 3D model and the actual terrain conditions, and also enabling the update of terrain deformation time series data to be associated with specific changed areas, further enhancing the application value of the data.

[0109] For example, in a geological monitoring project on a high slope in the southwestern mountainous region, technicians first used LiDAR point clouds (point cloud density 20 points / ㎡) and oblique photogrammetric images of the area to construct a basic 3D topographic model. Then, using multiple InSAR images combined with time-series analysis algorithms, they obtained the long-term deformation rate and cumulative deformation value of the slope. After verification by comparing with ground crack gauge data, accurate deformation results were output. Subsequently, using a terrain segmentation algorithm, with 8mm as the deformation threshold and 3° as the slope change threshold, the deformation results and slope topographic features were jointly analyzed to accurately identify three crack propagation zones at the rear edge of the slope and two slope slip zones in the middle, clearly defining the boundaries and extent of each change zone. Based on this information, technicians only optimized the elevation and texture of the change zones in the model, quickly generating a high-precision 3D topographic model. At the same time, the deformation data of each change zone was synchronously updated to the time-series database, providing accurate basis for geological disaster early warning and slope reinforcement scheme formulation.

[0110] In step S104, through statistical analysis and feature comparison, the systematic and random errors generated in the topographic inversion process of the surveying dataset are identified. The ISSA algorithm is used to optimize the BP neural network and dynamically correct the high-precision three-dimensional topographic model and topographic deformation results.

[0111] Among them, statistical analysis refers to the quantitative identification of the central tendency, dispersion and distribution pattern of errors in the terrain inversion process by calculating the statistical measures of the mean, standard deviation and probability distribution of surveying and mapping data, thereby distinguishing between systematic errors and random errors.

[0112] Feature comparison refers to extracting the topographic geometric features and texture features of surveying and mapping data, and comparing them one by one with standard features or historical features to identify feature differences caused by data deviations, thereby locating errors.

[0113] It is understood that the statistical analysis in this application, by calculating the mean, standard deviation, and probability distribution of the surveying data, quantifies the central tendency and dispersion of errors, clearly distinguishing between systematic and random errors. Feature comparison extracts the topographic geometric features and texture features corresponding to the surveying data and compares them one by one with standard features or historical benchmark features to accurately locate feature anomalies caused by data deviations. The combination of these two methods avoids the risk of misjudgment associated with a single method, clarifies the source and type of error, provides precise guidance for subsequent model correction, and ensures the credibility of the 3D terrain model and deformation results.

[0114] In step S105, based on the corrected high-precision 3D terrain model, terrain deformation results, and surveying dataset, a digital twin simulation model is constructed. AR real-scene overlay and multi-condition interactive query of geographic information are performed. The surveying data, inversion results, and error correction information are integrated through structured algorithms to generate a standardized surveying results report.

[0115] Among them, the structured algorithm integrates surveying and mapping data, inversion results, and error correction information through preset data classification rules and logical association structures to automatically generate a standardized surveying and mapping report containing multi-dimensional analysis content. The formula is as follows:

[0116]

[0117] in, This represents the overall weighted score for a specific section of the report. Number of data types; For the first Weighting coefficients for class data (surveying data) Inversion results Error correction ); For the first Standardized scores (values) for class data ); The content modules of the final report are combined; This is a function for concatenating modules; For the first One content module; A module sorting function; For the first The priority of each module; This is the output format for the final report; For format template functions; These are format style parameters.

[0118] It is understood that the structured algorithm in this application's embodiments classifies and integrates multi-source information such as surveying data and inversion results through modular design, allocates data priorities according to preset weight rules, and sequentially connects the various content modules of the report. With its clear execution logic, this algorithm avoids data clutter and chaos, improving report generation efficiency while ensuring a unified content structure and accurate data correlation. Furthermore, its maintainability facilitates subsequent functional expansion, providing standardized and highly reliable results for terrain management decisions.

[0119] For example, taking a mountain highway topographic surveying project as an example, after receiving LiDAR point cloud data, topographic inversion results, and error correction records, the structured algorithm first classifies and sorts the data according to preset rules, assigning weights of 0.4 to the surveying data, 0.35 to the inversion results, and 0.25 to the error information using a weighting formula. Then, it automatically integrates the data into a standardized content framework according to the priority modules of "basic data—core analysis—accuracy description," and uses templates to generate a report containing elevation statistics tables and deformation trend maps. The entire process requires no manual typesetting, improving generation efficiency by 50%. The report has a unified structure and accurate data correlation, providing highly reliable standardized results for highway alignment and safety assessment.

[0120] According to the embodiments of this application, a method for a remote sensing-based mapping and geographic system is proposed. This method utilizes a multimodal remote sensing data acquisition module to simultaneously acquire satellite remote sensing imagery, aerial LiDAR point clouds, UAV oblique photography, and ground-based IoT sensor data. This overcomes the limitations of traditional single-source data coverage, expands the acquisition dimensions and spatial coverage of mapping data, ensures the comprehensiveness of terrain information representation, and provides a multi-dimensional data foundation for subsequent mapping analysis. Furthermore, a data fusion and collaboration module, through an improved Mamba fusion model, extracts image texture, point cloud geometry, and terrain topology features from multiple sources in parallel. A selective state-space gating mechanism is used to strengthen the weights of key geographic features, performing nonlinear fusion and generating a unified precision mapping dataset. This achieves efficient feature extraction and accurate nonlinear fusion of multi-source remote sensing data, enhancing the representation of key geographic features. The weighting of features ensures the uniform accuracy of the surveying and mapping dataset; the dynamic terrain high-precision inversion module combines InSAR interferometry technology with time series analysis algorithms to complete terrain deformation monitoring, dynamically updates the terrain and landforms using terrain segmentation algorithms, achieves high-precision reconstruction of 3D terrain and real-time monitoring of terrain deformation, and provides timely results for terrain change analysis; the real-time error correction module identifies system and random errors, and dynamically corrects the model and deformation results through a BP neural network optimized by the ISSA algorithm, enhancing the accuracy and reliability of terrain data; the intelligent visualization application module constructs a high-fidelity terrain digital twin simulation model and completes AR real-scene overlay, providing flexible interactive geographic information queries and enriching the application scenarios and display forms of surveying and mapping results; thus, it solves the problems of weak multi-source data fusion and low terrain inversion accuracy in existing technologies.

[0121] The following will illustrate a method for a remote sensing-based mapping and geographic system through a specific embodiment, such as... Figure 9 As shown, it includes: Taking a large-scale coal mine ecological restoration and topographic monitoring project as an example, the technical team strictly followed this surveying method to carry out the entire process, accurately capturing the topographic changes and ecological restoration effects caused by coal mining. First, multi-source data synchronous acquisition was initiated: panchromatic and multispectral images covering the entire mining area and a surrounding 100 square kilometers were acquired using the Gaofen-2 satellite. The panchromatic band resolution reached 0.8 meters, clearly identifying the distribution of vegetation seedlings in the reclaimed area. A fixed-wing aircraft equipped with a RIEGLVQ-120 LiDAR device was deployed at a flight altitude of 1200 meters and a point cloud density of 35 points / square meter to scan key areas such as coal mining subsidence areas and spoil heaps, acquiring high-precision three-dimensional data including surface undulations and coal seam mining boundaries. Data from Weidian Cloud; Four drones equipped with five-lens tilting cameras were deployed to conduct low-altitude photography of three reclamation demonstration areas using a "spiral" flight mode. The flight altitude was 150 meters, and each flight acquired 1,800 images. The image overlap was maintained at more than 85% to ensure modeling details. Eighty IoT sensors were deployed at the edge of the subsidence area and the foot of the spoil heap to collect real-time data on surface subsidence, soil moisture, and vegetation growth environment. The sampling frequency was set to once every 20 minutes, with a focus on monitoring the terrain stability during the rainy season (July-September).

[0122] After data acquisition, the data enters the fusion processing stage. Technicians first use the WGS84 coordinate system as a reference and employ a seven-parameter transformation method to calibrate the coordinates of data from different sources, including satellite imagery and LiDAR point clouds, eliminating spatial deviations caused by differences in equipment coordinate systems. Simultaneously, a location code containing latitude, longitude, elevation, and acquisition time is added to each data point, generating a structured feature vector. This feature vector is then input into an improved Mamba fusion model. The model processes multi-source data in parallel using a linear attention mechanism: extracting texture features such as vegetation cover and soil exposure rate from satellite imagery; extracting geometric features such as slope of subsidence areas and height of spoil heaps from LiDAR point clouds; and extracting attribute features such as settlement rate and soil moisture content from sensor data. During the fusion process, a selective state-space gating mechanism is activated, increasing the weight of key features such as coal mining subsidence boundaries and vegetation root distribution areas in reclaimed areas to 0.35, while weakening irrelevant background noise features. After integration by a nonlinear fusion algorithm, a standardized mapping dataset with a uniform accuracy of ±3 cm is output from the data processing terminal.

[0123] Based on this dataset, the technical team carried out terrain inversion and model construction: 3D reconstruction retrieved LiDAR point cloud and UAV oblique photogrammetric image data from the surveying dataset. After denoising and registration preprocessing, the grid resolution was set according to the terrain complexity and surveying accuracy requirements of the survey area. An irregular triangular mesh (TIN) was used to build a 3D terrain grid skeleton. The grid elevation information was completed using the Kriging interpolation algorithm to restore the terrain undulation and spatial structure. Subsequently, the texture information of the oblique photogrammetric image was extracted, and the texture was accurately fitted to the 3D grid surface according to the mapping relationship between the image projection coordinates and the grid spatial coordinates. Color balancing and seam smoothing were performed at the texture splicing points. Finally, a basic 3D terrain model with both geometric accuracy and texture effect was constructed, which fully reproduced the 3D spatial characteristics of the survey area terrain. When carrying out deformation monitoring, firstly, multiple periods of InSAR imagery and LiDAR elevation data are retrieved from the mapping dataset. Using InSAR interferometry technology, the spatial location registration of the multiple periods of InSAR imagery is completed first through a feature point matching algorithm, and then an interferogram is generated through filtering. Next, the phase unwrapping operation is carried out using the minimum cost flow method to extract the initial deformation phase information of the terrain. Subsequently, a time series analysis algorithm is used to screen effective scatterers and perform time series correlation analysis on the deformation phase information of the multiple periods. The long-term deformation rate and cumulative deformation value of the terrain are obtained through fitting calculation. Finally, the above calculation results are verified with ground sensor data for error, and the final output is the terrain deformation result containing core deformation parameters and accuracy indicators.

[0124] In the error correction phase, technicians used statistical analysis to calculate the mean (1256.8 meters) and standard deviation (±9.2 mm) of the LiDAR point cloud elevation data, identifying a 6 mm overall elevation shift caused by temperature drift in the LiDAR equipment. By comparing the topographic features of the subsidence area in the inversion model with standard features in historical exploration data, they located random errors caused by sudden fluctuations in mining activities. Subsequently, the ISSA algorithm was used to optimize the BP neural network, using the error value as the fitness function. The network parameters were iteratively adjusted based on the exploration and updating behavior of a sparrow population. After 60 iterations, the optimal network structure was determined to be 5-13-1, improving the model convergence speed by 40% compared to the unoptimized version. The error data was input into the optimized network, which outputs targeted corrections to dynamically adjust the subsidence area elevation and deformation time-series data in the high-precision model. After correction, the model was compared with GNSS measured data, and the elevation error was reduced to ±2.1 mm.

[0125] Finally, in the application phase, the technical team, based on the revised model and data, used the UE5 engine to construct a digital twin simulation model that maps 1:1 to the physical topography of the mining area. This model was linked to three years of deformation time-series data and vegetation growth information in the reclaimed area. A timeline control allows for a direct review of the entire process of subsidence area restoration from mining to completion. During on-site surveys, GPS and visual SLAM technology on mobile terminals enabled AR overlay of the digital twin model with real-world images, achieving an alignment accuracy of 2 centimeters. Staff could click on reclaimed terraces in the real-world scene to view detailed data such as soil fertility and vegetation survival rate on the screen. The query system supports multi-condition searches; entering "subsidence rate > 10 mm / year + vegetation coverage < 30%" quickly locates three areas with weak ecological restoration capabilities. The report generation system automatically integrates all data and analysis results using a structured algorithm, generating standardized reports according to the logic of "mining area overview—data acquisition—topography inversion—error analysis—restoration recommendations." This provides precise data support for optimizing ecological restoration plans for coal mining enterprises and for acceptance evaluation by regulatory departments.

[0126] In summary, this application's embodiments acquire multi-source data, including satellite imagery and LiDAR point clouds, and after calibration in the WGS84 coordinate system, use an improved Mamba model to generate centimeter-level datasets. Then, combining InSAR technology with time-series analysis, a 3D model is constructed and optimized, with errors corrected using an ISSA-optimized BP neural network. Finally, a digital twin model is built, enabling AR overlay, multi-condition queries, and standardized report generation. This accurately captures the dynamics of subsidence and restoration, providing reliable support for ecological restoration optimization and regulatory acceptance.

[0127] Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 1001, the processor 1002, and the computer program stored on the memory 1001 and capable of running on the processor 1002.

[0128] When the processor 1002 executes the program, it implements a method for a surveying and mapping geographic system based on remote sensing technology provided in the above embodiments.

[0129] Furthermore, electronic devices also include: Communication interface 1003 is used for communication between memory 1001 and processor 1002.

[0130] The memory 1001 is used to store computer programs that can run on the processor 1002.

[0131] The memory 1001 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0132] If the memory 1001, processor 1002, and communication interface 1003 are implemented independently, then the communication interface 1003, memory 1001, and processor 1002 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0133] Optionally, in a specific implementation, if the memory 1001, processor 1002, and communication interface 1003 are integrated on a single chip, then the memory 1001, processor 1002, and communication interface 1003 can communicate with each other through an internal interface.

[0134] The processor 1002 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0135] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for a mapping and geographic system based on remote sensing technology.

[0136] Furthermore, embodiments of this application also provide a computer program product, including a computer program or instructions, which, when executed, implement the aforementioned method for a remote sensing-based mapping and geographic system.

[0137] In the description of this specification, the references to "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0138] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0139] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0140] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0141] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0142] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A mapping and geographic system based on remote sensing technology, characterized in that, include: The system includes a multimodal remote sensing data acquisition module, a data fusion and collaboration module, a dynamic terrain high-precision inversion module, a real-time error correction module, and an intelligent visualization application module; among which... The multimodal remote sensing data acquisition module is used to simultaneously acquire satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data, and ground IoT sensor data. The data fusion and collaboration module is used to perform geographic coordinate calibration and location encoding on the satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data and ground IoT sensor data. It combines the improved Mamba fusion model to perform feature extraction and nonlinear fusion, and strengthens the weight of key geographic features through a selective state space gating mechanism to generate a mapping dataset with uniform accuracy. The dynamic terrain high-precision inversion module performs three-dimensional terrain reconstruction based on the mapping dataset, combines InSAR interferometric measurement technology and time series analysis algorithm to complete terrain deformation monitoring, and uses terrain segmentation algorithm to dynamically update the terrain and landform. The real-time error correction module is used to identify the systematic and random errors generated by the mapping dataset during the terrain inversion process, and to construct a BP neural network optimized by the ISSA algorithm to dynamically correct the three-dimensional terrain model and terrain deformation results. The intelligent visualization application module is used to construct digital twin simulation models, perform AR real-scene overlay, provide interactive geographic information queries, and automatically generate standardized surveying and mapping results reports.

2. The mapping and geographic system based on remote sensing technology according to claim 1, characterized in that, The multimodal remote sensing data acquisition module includes a satellite image acquisition unit, a LiDAR point cloud acquisition unit, an oblique photography acquisition unit, and a ground sensor acquisition unit. The satellite image acquisition unit interfaces with a multi-resolution satellite remote sensing platform to simultaneously acquire multispectral and panchromatic satellite image data. The LiDAR point cloud acquisition unit acquires high-precision point cloud data of the terrain surface using an airborne LiDAR device. The oblique photography acquisition unit acquires oblique photographic images of the terrain using a multi-view camera from an unmanned aerial vehicle (UAV). The ground sensor acquisition unit acquires data on terrain surface temperature, humidity, and settlement monitoring points using deployed IoT sensors.

3. A mapping and geographic system based on remote sensing technology according to claim 1, characterized in that, The data fusion and collaboration module includes a geocoding unit, a feature extraction unit, a fusion processing unit, and a dataset output unit. The geocoding unit performs geographic coordinate calibration on multi-source remote sensing data, introduces geographic feature location codes to identify data spatial attributes, and generates feature vectors containing location information. The feature extraction unit inputs the feature vectors into an improved Mamba model and extracts image texture, point cloud geometry, and terrain topology features in parallel through a linear attention mechanism. The fusion processing unit strengthens the weights of key geographic features through a selective state-space gating mechanism and performs nonlinear fusion of the image texture, point cloud geometry, and terrain topology features. The dataset output unit standardizes the fused feature data to generate a mapping dataset with uniform accuracy.

4. A mapping and geographic system based on remote sensing technology according to claim 1, characterized in that, The dynamic terrain high-precision inversion module includes a 3D reconstruction unit, a deformation monitoring unit, and a dynamic update unit. The 3D reconstruction unit uses LiDAR point cloud and oblique photogrammetric image data from the mapping dataset. After denoising, registration, 3D grid construction, and texture bonding, it constructs a basic 3D terrain model. The deformation monitoring unit uses multiple periods of InSAR imagery and LiDAR elevation data from the mapping dataset. It uses InSAR interferometry to register, generate interferograms, and unwrap the phases of the multiple periods of InSAR imagery, extracting initial terrain deformation phase information. It then uses a time-series analysis algorithm to perform temporal correlation analysis on the multi-period deformation phase information, calculating the long-term deformation rate and cumulative deformation value of the terrain. Simultaneously, it compares and verifies the results with ground sensor data, outputting accurate terrain deformation results. The dynamic update unit uses the deformation results to identify areas of terrain change through terrain segmentation algorithms, dynamically optimizes the basic 3D terrain model, obtains a high-precision 3D terrain model, and synchronously updates the terrain deformation time-series data.

5. A mapping and geographic system based on remote sensing technology according to claim 1, characterized in that, The real-time error correction module includes an error identification unit, a model optimization unit, and a correction execution unit. The error identification unit is used to identify systematic and random errors generated in the topographic inversion process of the surveying dataset through statistical analysis and feature comparison. The model optimization unit is used to optimize the initial weights and structural parameters of the BP neural network using the ISSA algorithm to improve the model correction accuracy. The correction execution unit is used to input the error data into the optimized BP neural network to dynamically correct the high-precision 3D topographic model and topographic deformation results.

6. A mapping and geographic system based on remote sensing technology according to claim 1, characterized in that, The intelligent visualization application module includes a digital twin construction unit, an AR overlay unit, a query interaction unit, and a report generation unit. The digital twin construction unit, based on the high-precision 3D terrain model, terrain deformation results, and surveying dataset, constructs a 1:1 mapping digital twin simulation model of the terrain and landforms, synchronously linking terrain deformation time-series data with land cover characteristics. The AR overlay unit is used to overlay the digital twin simulation model with real-world images, achieving real-time alignment and interactive browsing between the virtual model and the real terrain. The query interaction unit provides multi-condition interactive query functions for geographic information. The report generation unit integrates surveying data, inversion results, and error correction information using structured algorithms to generate standardized surveying results reports.

7. A method for applying to a remote sensing-based mapping geographic system according to any one of claims 1-6, characterized in that, The method includes: Simultaneously acquire satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data, and ground-based IoT sensor data; The satellite remote sensing image data, aerial LiDAR point cloud data, UAV oblique photography data and ground IoT sensor data are subjected to geographic coordinate calibration and location encoding. By improving the Mamba fusion model, image texture, point cloud geometry and terrain topology features are extracted in parallel. Nonlinear fusion is performed through a selective state space gating mechanism to generate a mapping dataset with uniform accuracy. A basic three-dimensional terrain model is constructed based on the mapping dataset. The terrain deformation results are obtained by combining InSAR interferometry technology and time series analysis algorithm. The terrain segmentation algorithm is used to identify the terrain and landform change areas. The basic three-dimensional terrain model is optimized based on the terrain deformation results and change area information to obtain a high-precision three-dimensional terrain model. The terrain deformation time series data is updated synchronously. Through statistical analysis and feature comparison, the systematic and random errors generated in the topography inversion process of the surveying dataset are identified. The ISSA algorithm is used to optimize the BP neural network and dynamically correct the high-precision three-dimensional topography model and topography deformation results. Based on the corrected high-precision 3D terrain model, terrain deformation results, and the surveying dataset, a digital twin simulation model is constructed. AR real-scene overlay and multi-condition interactive query of geographic information are performed. The surveying data, inversion results, and error correction information are integrated through structured algorithms to generate a standardized surveying results report.

8. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method of a remote sensing-based mapping geographic system as claimed in claim 7.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When a computer program or instruction is executed, it implements a method for a remote sensing-based mapping geographic system as claimed in claim 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When a computer program or instruction is executed, it implements a method for a remote sensing-based mapping geographic system as claimed in claim 7.

Images