A seamless splicing method and system for environment data based on micro scenes

Abstract

The present application belongs to the technical field of earth space environment monitoring, and specifically discloses a seamless splicing method for environment data based on micro scenes, which comprises the following steps: obtaining three-dimensional grid unit environment data to be spliced; extracting edge features of adjacent surfaces; dynamically matching and correcting the space coordinates of the features of adjacent grids; selecting a physical equation as a constraint condition, and calculating an edge point interpolation result in combination with a credibility weight; checking the spliced edge data, identifying abnormal points, and generating a checking report; if the checking fails, feeding back to the feature extraction step for reprocessing until the threshold is met; and outputting integrated environment data after seamless splicing. The present application realizes accurate alignment of the spatial positions of the edges of three-dimensional grid units, natural transition of physical properties, and time dynamic consistency, and significantly improves the accuracy and reliability of environment data splicing.

Description

Technical Field

[0001] This invention belongs to the field of Earth space environment monitoring technology, and specifically relates to a method and system for seamless stitching of environmental data based on micro-scenes. Background Technology

[0002] In the fields of geospatial environment monitoring, numerical simulation, and 3D scene construction, the storage and dynamic expansion of large-scale environmental data face significant challenges due to the continuous expansion of environmental data dimensions and the increasing demands for accuracy. Environmental data encompasses various parameters, including atmospheric, hydrological, geological, and ecological data, and its resolution requirements have gradually increased from the kilometer level to the meter level, resulting in a geometric increase in data volume. Traditional large-scale environmental data management methods typically employ a holistic modeling and storage strategy, which involves collecting data across the entire target area and directly generating a single 3D data model. While this holistic approach ensures the spatial continuity of the data, it leads to two major drawbacks: a dramatic increase in storage space requirements and poor flexibility in scene expansion. These shortcomings make it difficult to meet the practical needs of modern environmental monitoring and numerical simulation for efficient management and dynamic updates of massive amounts of data.

[0003] To address the aforementioned issues, a scene management method based on 3D geospatial meshing has emerged. This method divides a large scene into several 3D mesh cells of a preset size, with each cell independently storing environmental data within its coverage area. This strategy significantly reduces the storage pressure on a single scene, and when the scene needs to be expanded, only adjacent mesh cells need to be stitched together without reconstructing the entire domain data, thus greatly improving dynamic expansion efficiency. However, during the mesh cell stitching process, due to differences in data sources between different mesh cells, slight differences in spatial coordinate systems, and inconsistent boundary condition settings during physical modeling, the edge data of adjacent meshes often exhibits discontinuity. Specifically, spatial positional deviations in the edge feature points of adjacent grids due to coordinate system transformation errors can lead to spatial discontinuities when directly spliced ​​together, such as abrupt breaks in mountain ranges at grid boundaries; unnatural jumps in physical parameters of environmental data, such as temperature, wind speed, and humidity, at grid edges violate natural laws such as atmospheric circulation and heat conduction; inconsistent temporal dynamics, with real-time updated environmental data exhibiting time discrepancies due to different collection frequencies of each grid; and the fusion of heterogeneous information, such as satellite remote sensing data and ground sensor data, in different grids, resulting in significant differences in data accuracy and error distribution, which can lead to accuracy discontinuities when directly spliced ​​together. Summary of the Invention

[0004] The purpose of this invention is to overcome the defects in the existing technology and provide a method for seamless stitching of environmental data based on micro-scenes.

[0005] A first aspect of the present invention provides a method for seamless stitching of environmental data based on micro-scenes, comprising the following steps: Step 1: Obtain environmental data for at least two 3D mesh units to be stitched together. The environmental data includes spatial coordinates, physical parameters, timestamps, and data source identifiers. Step 2: Extract key feature points, gradient features, and texture structure from the adjacent surfaces of each 3D mesh cell; Step 3: Dynamically match the multimodal edge features extracted from adjacent grid cells, construct feature twins, calculate coordinate correction based on the matching results, and perform spatial coordinate correction on the grid edge points; Step 4: Based on the type of environmental data, select the corresponding physical equation from the preset physical equation library as the constraint condition, combine the original data and confidence weight of the adjacent grid edge points, and calculate the interpolation result of the edge points through the hybrid interpolation operator to complete the physical attribute transition; Step 5: Perform spatial continuity and temporal consistency checks on the stitched edge data, identify outliers, and generate a check report; Step 6: If the verification fails, the anomaly information is fed back to Step 2, the multimodal edge features at the anomaly point are re-extracted, and Steps 3-5 are executed again until all indicators meet the preset threshold. Step 7: Output the seamlessly stitched integrated environment data.

[0006] A further solution is that the physical equation library includes the Navier-Stokes equations corresponding to fluid data, Fick's law corresponding to diffusion data, and terrain slope variation constraints corresponding to terrain data. The selection of the corresponding physical equation as a constraint based on the type of environmental data includes: selecting the Navier-Stokes equation when the environmental data is wind speed or water flow; selecting Fick's law when the environmental data is pollutant concentration or temperature; and selecting terrain slope variation constraint when the environmental data is terrain elevation.

[0007] A further approach is that, in step 4, the hybrid interpolation operator uses a weighted average method to calculate the interpolation results of the edge points, and the calculation formula is as follows: ,in , These are the original data for adjacent grid edge points. , The weights are calculated based on data credibility. The reliability of the data is dynamically adjusted based on the accuracy, timeliness, and historical errors of the data source.

[0008] A further approach is that the spatial continuity verification in step 5 includes: calculating the abrupt change index of the edge points after splicing. ,in For edge point data values, It is the average of several adjacent voxels. For standard deviation, when Points exceeding a preset threshold are marked as spatial anomalies. The time consistency verification includes: verifying the timestamp difference of environmental data of adjacent grid cells; if the difference exceeds a preset time threshold, linear interpolation is used to generate intermediate time data and the time inconsistency point is marked.

[0009] A further approach is to perform super-resolution reconstruction on the edge data of low-resolution grid cells before extracting multimodal edge features, thereby improving the accuracy to the same level as high-resolution grid cells; and to use Gaussian filtering to smooth the transition region from high-resolution grid cells to low-resolution grid cells. During the process of dynamic matching, constructing feature twins and calculating coordinate corrections, an initial confidence factor is assigned to the environmental data of each grid cell. The confidence factor is used for feature selection during feature matching and weight calculation of coordinate corrections. In the process of calculating the interpolation results of edge points using the hybrid interpolation operator, the confidence factor is dynamically adjusted according to the data update and historical error. The weights in the hybrid interpolation operator are calculated based on the adjusted confidence factor.

[0010] A second aspect of the present invention discloses a seamless environmental data stitching system based on micro-scenes, comprising: An external data interface module is used to receive environmental data of the 3D mesh cells to be stitched together; The feature extraction module is used to extract key feature points, gradient features, and texture structure from the adjacent surfaces of each 3D mesh cell; The matching and correction module is used to dynamically match the multimodal edge features extracted from adjacent grid cells, construct feature twins, calculate coordinate correction based on the matching results, and perform spatial coordinate correction on grid edge points. The constraint interpolation module is used to select the corresponding physical equation as the constraint condition according to the type of environmental data, and calculate the interpolation result of the edge point by combining the original data and confidence weight of the adjacent grid edge point through a hybrid interpolation operator. The spatiotemporal verification module is used to verify the spatial continuity and temporal consistency of the stitched edge data, identify outliers, and generate a verification report. The correction module is used to receive the abnormal point information identified by the spatiotemporal verification module and send the abnormal point information to the feature extraction module to re-extract the multimodal edge features at the abnormal point; The results output module is used to output the seamlessly stitched integrated environment data.

[0011] A further embodiment is that the feature extraction module includes: The data preprocessing unit is used to perform super-resolution reconstruction on the edge data of low-resolution grid cells and to smooth the region where high-resolution grid cells transition to low-resolution grid cells. The 3D SIFT feature extractor is used to identify key edge feature points and record their three-dimensional coordinates and attribute values. The 3D gradient calculator is used to calculate the spatial rate of change of data at the edges and obtain gradient features. Gray-level co-occurrence matrix processor, used to extract texture structure of edge regions.

[0012] A further embodiment is that the matching correction module includes: The DTW matching algorithm processor is used to dynamically match the multimodal edge features of adjacent grid cells and construct feature twins. The coordinate deviation calculation unit is used to calculate the coordinate correction amount based on the matching results; Coordinate converter, used to correct the spatial coordinates of grid edge points; The matching result evaluation unit is used to evaluate the matching accuracy. When the error exceeds a preset threshold, it triggers the re-extraction of features. A credibility assessment model is used to evaluate the credibility of environmental data for each grid cell. An initial confidence allocation unit is used to allocate an initial confidence factor to the environmental data of each grid cell. The confidence factor is used for feature filtering and weight calculation of coordinate correction during feature matching.

[0013] A further embodiment is that the constraint interpolation module includes: A multiphysics equation library is used to store the physical equations corresponding to different types of environmental data; The equation selection unit is used to select the corresponding physical equation from the multiphysics equation library based on the type of environmental data. The hybrid interpolation operator calculator is used to calculate the interpolation results of edge points by combining physical equations, raw data of adjacent grid edge points, and confidence weights.

[0014] The weighted fusion calculator is used to calculate the final fusion value of edge points based on the confidence factor.

[0015] A further embodiment is that the spatiotemporal verification module includes: The spatial mutation index calculation unit is used to calculate the mutation index of the edge points after splicing and to identify spatial anomalies. The timestamp verification unit is used to verify the timestamp difference between environmental data of adjacent grid cells. When the difference exceeds a preset time threshold, time interpolation is triggered. Anomaly marking unit is used to mark anomalies and transmit the anomaly information to the correction module.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention breaks through the limitations of traditional methods that only pursue numerical continuity. By combining feature matching with physical law constraints, it achieves deep seamless splicing from spatial location to physical attributes, ensuring that the splicing result is not only numerically consistent but also conforms to the inherent laws of the natural environment, thus completely avoiding spatial discontinuity and attribute abrupt changes.

[0017] This invention addresses the challenge of fusing multi-source heterogeneous data by employing a dynamic adaptation and weighted fusion strategy. It is compatible with environmental data from different sources, with different precision, and in different formats, effectively eliminating splicing contradictions caused by data differences and improving adaptability to complex data environments.

[0018] This invention utilizes real-time verification and closed-loop optimization mechanisms to rapidly process dynamically updated environmental data, ensuring data consistency over time, meeting the needs of real-time monitoring and dynamic analysis, and providing support for timely decision-making.

[0019] This invention eliminates the need to reconstruct the entire domain data when expanding the scene, reducing the requirements for storage device capacity and read / write speed. When the scene range needs to be expanded, there is no need to remodel the new area. Functions can be flexibly expanded according to needs, reducing the investment in secondary development.

[0020] In military applications, this invention can cover the needs of all levels, from high-level strategy to grassroots operations, providing consistent and accurate environmental data support for each level, thereby improving the scientific nature of operational decision-making and the success rate of operations. Attached Figure Description

[0021] The following figures are for illustrative purposes only and are not intended to limit the scope of the invention, wherein: Figure 1 : Schematic diagram of the splicing method of the present invention; Figure 2 : Schematic diagram of the adaptive fusion process for multi-source heterogeneous data; Figure 3 : A schematic diagram of the splicing system module connection of the present invention; Figure 4 : Block diagram of feature extraction module; Figure 5 : Block diagram of the matching and correction module; Figure 6 : Block diagram of the constraint interpolation module; Figure 7 : Block diagram of the spatiotemporal verification module; The diagram shows: 1. External data interface module; 2. Feature extraction module; 3. Matching correction module; 4. Constraint interpolation module; 5. Spatiotemporal verification module; 6. Correction module; 7. Result output module; 8. Data preprocessing unit; 9. Initial confidence allocation unit; 10. Weighted fusion calculator; 11. 3D SIFT feature extractor; 12. 3D gradient calculator; 13. Gray-level co-occurrence matrix processor; 14. DTW matching algorithm processor; 15. Coordinate deviation calculation unit; 16. Coordinate converter; 17. Matching result evaluation unit; 18. Confidence evaluation model; 19. Multiphysics equation library; 20. Equation selection unit; 21. Hybrid interpolation operator calculator; 22. Spatial catastrophe index calculation unit; 23. Timestamp verification unit; 24. Outlier marking unit. Detailed Implementation

[0022] To make the objectives, technical solutions, design methods, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0023] Example 1 This embodiment discloses a seamless stitching method for environmental data based on micro-scenes. In this embodiment, the stitching of environmental data from two adjacent three-dimensional grid cells within a certain watershed is taken as an example. Grid A covers the upstream mountainous area, and Grid B covers the downstream plain. Both grid cells store various types of environmental data such as topographic elevation, wind speed, temperature, and pollutant concentration. However, the data sources of the two are different. Grid A is mainly based on satellite remote sensing data with a spatial resolution of 100 meters, while Grid B is mainly based on ground sensor network data with a spatial resolution of 10 meters. Furthermore, there is a slight deviation in the coordinate systems used by the two.

[0024] like Figure 1 As shown, it includes the following steps: Step 1: Obtain environmental data for grid A and grid B. Each data entry includes spatial coordinates, physical parameters, timestamp, and data source identifier. The spatial coordinates are three-dimensional coordinates in the WGS84 coordinate system. The physical parameters include altitude, wind speed vector, temperature value, and PM2.5 concentration value. The timestamp records the data acquisition time, and the data source identifier distinguishes between satellite remote sensing data and ground sensor data.

[0025] Step 2: For adjacent surfaces of meshes A and B, the 3D SIFT algorithm is used to identify key feature points at the edges, calculate the spatial rate of change of data at the edges, obtain gradient features, and extract the texture structure of the edge region using a gray-level co-occurrence matrix. In this embodiment, key feature points include terrain peak points, pollutant concentration extreme points, etc., and the three-dimensional coordinates and attribute values ​​of the feature points are recorded. Gradient features include the rate of change of temperature in the vertical direction and the wind speed shear in the horizontal direction. The texture structure can be the spatial distribution texture of vegetation cover.

[0026] Step 3: A dynamic time warping algorithm is used to match the multimodal features of the edges of grids A and B. For example, a feature twin correspondence is established between a terrain peak point on the edge of grid A and a terrain peak point on the edge of grid B. Based on the matching results, the coordinate correction between the two grid cells is calculated, including translation, rotation, and scaling. The coordinate transformation is applied to all edge points of grid B, achieving sub-pixel spatial alignment with the edge points of grid A, thus eliminating the mountain range interruption caused by coordinate system differences.

[0027] Step 4: Based on the type of environmental data, select the corresponding physical equation from the preset physical equation library as the constraint condition. Combine the original data and confidence weights of adjacent grid edge points, calculate the interpolation results of the edge points using a hybrid interpolation operator to complete the physical attribute transition. In this embodiment, for wind speed data, the Navier-Stokes equation is selected as the constraint to ensure the continuity of the velocity vector at the boundary; for temperature data, Fick's law is selected as the constraint to ensure the gradient continuity of heat diffusion; for terrain elevation data, the slope variation is constrained to not exceed the maximum slope threshold of the natural terrain. The hybrid interpolation operator uses a weighted average method, and the calculation formula is as follows: ,in , These are the original data for adjacent grid edge points. , The weights are calculated based on data reliability. Since the ground sensor data of grid B has higher accuracy than the satellite remote sensing data of grid A, the reliability weight for grid B is... The weight of grid A is set to 0.7. Set it to 0.3.

[0028] Step 5: Perform spatial continuity and temporal consistency checks on the stitched edge data, identify outliers, and generate a check report. During spatial verification, calculate the abrupt change index of the stitched edge points. ,in For edge point data values, It is the average of several adjacent voxels. For standard deviation, when Points exceeding a preset threshold are marked as spatial anomalies. During time verification, the timestamp difference between grid A and grid B is checked. For example, if the data collection time for grid A is 9:00 AM and that for grid B is 9:40 AM, the difference of 40 minutes exceeds the preset 30-minute threshold. Therefore, linear interpolation is used to generate data at the intermediate time between 9:20 AM and 9:30 AM, and the time inconsistency is marked.

[0029] Step 6: If the verification fails, the anomaly information is fed back to Step 2. The multimodal edge features at the anomaly point are re-extracted, and Steps 3 to 5 are executed again until all indicators meet the preset threshold. In this embodiment, for the marked temperature anomaly points, the location information of the point is fed back to the edge feature extraction step. More refined gradient features and texture features are re-extracted near the anomaly point. Then, matching correction and physical constraint interpolation are re-executed. An adaptive optimization algorithm is used to adjust the interpolation weights, and the confidence weight of grid B is adjusted to 0.8. After recalculation, the temperature of the point becomes 26.5 degrees Celsius, and the mutation index drops to 1.2, thus passing the verification.

[0030] Step 7: Output the seamlessly stitched integrated environmental data. The output includes the stitched data file, a verification report, and coordinate correction records. The verification report details the location and processing results of spatial anomalies, as well as timestamp differences and interpolation details at points of temporal inconsistency. In the final stitched data, the terrain elevation, wind speed, temperature, and pollutant concentration at the boundary between grid A and grid B exhibit a continuous, smooth, and physically consistent transition, achieving true seamless stitching.

[0031] like Figure 2 As shown, this embodiment also incorporates a multi-source heterogeneous data adaptive fusion mechanism during implementation. Specifically, before extracting multimodal edge features in step 2, super-resolution reconstruction is performed on the edge data of low-resolution grid A. The ESRGAN model is used to increase its spatial resolution from 100 meters to 10 meters, maintaining consistency with the resolution of grid B. Gaussian filtering is used to smooth the transition region from grid B to grid A, avoiding detail anomalies caused by sudden resolution changes. In step 3, during dynamic matching and coordinate correction, an initial confidence factor is assigned to each grid cell. The initial confidence factor for grid A is 0.6, and for grid B it is 0.9. The initial confidence factor is used for feature selection during feature matching and for weighting the coordinate correction. In step 4, during the calculation of the interpolation result using the hybrid interpolation operator, the confidence factor is dynamically adjusted based on data updates and historical errors. For example, when the sensor data of grid B is updated multiple times consecutively and the error is stable, its confidence factor gradually increases to 0.95, and the interpolation weight changes dynamically accordingly.

[0032] This embodiment can directly adapt to the operational environment data needs of the entire military combat system, from the Central Military Commission (CMC) and theater commands to brigades, platoons, and individual soldiers. The CMC needs to coordinate the national and even global military strategic environment; this embodiment can seamlessly stitch together three-dimensional grid environment data from different regions, providing comprehensive and accurate environmental data for the CMC's strategic decision-making. For the tactical deployment needs at the brigade / regiment level, this embodiment can stitch together grid environment data within the operational area, providing support for tactical deployment and weapon application. For the specific tactical mission needs at the platoon level, such as reconnaissance and raids, it can quickly stitch together grid environment data of the platoon's activity area, ensuring the smooth conduct of platoon operations. For the real-time environmental perception needs at the individual soldier level, it enables rapid and seamless stitching together of micro-scene grid environment data surrounding the soldier, providing real-time environmental conditions such as temperature, humidity, and terrain undulations, assisting soldiers in avoiding dangerous areas such as steep slopes and deep ditches, and improving individual survivability and combat efficiency.

[0033] Example 2 like Figure 3 As shown in the figure, based on Embodiment 1, this embodiment discloses a seamless stitching system for environmental data based on micro-scenes, including an external data interface module 1, a feature extraction module 2, a matching correction module 3, a constraint interpolation module 4, a spatiotemporal verification module 5, a correction module 6, and a result output module 7.

[0034] Among them, the external data interface module 1 serves as the data entry point of the system, and is used to receive the environmental data of the three-dimensional mesh cells to be stitched together.

[0035] like Figure 4 As shown, the feature extraction module 2 specifically includes a data preprocessing unit 8, a 3D SIFT feature extractor 11, a 3D gradient calculator 12, and a gray-level co-occurrence matrix processor 13. The data preprocessing unit 8 performs super-resolution reconstruction on the edge data of low-resolution grid cells, using the ESRGAN model to improve the spatial resolution to the same accuracy as high-resolution grid cells, and employs Gaussian filtering to smooth the transition area from high-resolution to low-resolution grid cells, avoiding detail anomalies caused by sudden resolution changes. The 3D SIFT feature extractor 11 identifies key edge feature points and records their three-dimensional coordinates and attribute values, such as terrain peak points and pollutant concentration extreme points; the 3D gradient calculator 12 calculates the spatial rate of change of data at the edge to obtain gradient features; and the gray-level co-occurrence matrix processor 13 extracts the texture structure of the edge region.

[0036] like Figure 5As shown, the matching correction module 3 is used to dynamically match the multimodal edge features extracted from adjacent grid cells, construct feature twins, calculate coordinate corrections based on the matching results, and perform spatial coordinate correction on the grid edge points. The matching correction module 3 includes a DTW matching algorithm processor 14, a coordinate deviation calculation unit 15, a coordinate converter 16, a matching result evaluation unit 17, a credibility evaluation model 18, and an initial credibility allocation unit 9. The DTW matching algorithm processor 14 dynamically matches the multimodal edge features of adjacent grid cells to construct feature twins, for example, establishing a correspondence between a terrain peak point on the edge of grid A and a terrain peak point on the edge of grid B. The credibility evaluation model 18 evaluates the credibility of the environmental data of each grid cell, and the initial credibility allocation unit 9 assigns an initial credibility factor to the environmental data of each grid cell, which is used for feature selection and weight calculation of coordinate corrections during feature matching. The coordinate deviation calculation unit 15 calculates the coordinate correction amount based on the matching result, including translation, rotation and scaling; the coordinate converter 16 performs spatial coordinate correction on the grid edge points; the matching result evaluation unit 17 evaluates the matching accuracy, and triggers re-extraction of features when the error exceeds the preset threshold.

[0037] like Figure 6 As shown, the constraint interpolation module 4 is used to select the corresponding physical equation as the constraint condition according to the type of environmental data. Combining the original data and confidence weights of adjacent grid edge points, it calculates the interpolation result of the edge points using a hybrid interpolation operator. This module includes a multiphysics equation library 19, an equation selection unit 20, a hybrid interpolation operator calculator 21, and a weighted fusion calculator 10. The multiphysics equation library 19 stores physical equations such as the Navier-Stokes equation, Fick's law, and terrain slope variation constraints. The equation selection unit 20 automatically selects the corresponding physical equation according to the type of environmental data. For example, when the environmental data is wind speed or water flow, the Navier-Stokes equation is selected; when it is pollutant concentration or temperature, Fick's law is selected; and when it is terrain elevation, terrain slope variation constraints are selected. The hybrid interpolation operator calculator 21 combines the selected physical equation, the original data of adjacent grid edge points, and the confidence weights, and calculates the interpolation result of the edge points using a weighted average method. The calculation formula is: ,in , These are the original data for adjacent grid edge points. , The weights are calculated based on data reliability. The weighted fusion calculator 10 calculates the final fusion value of the edge points based on the reliability factor, and dynamically adjusts the reliability factor according to data updates and historical errors. Through the above hybrid interpolation operator, not only can the original data information of adjacent grids be effectively utilized, but also the weights can be reasonably allocated according to the data reliability, ensuring that the interpolation results not only conform to physical laws, but also fully reflect the true reliability of the data, providing a high-quality data foundation for subsequent spatiotemporal dynamic verification.

[0038] like Figure 7 As shown, the spatiotemporal verification module 5 is used to verify the spatial continuity and temporal consistency of the stitched edge data, identify outliers, and generate a verification report. The spatiotemporal verification module 5 includes a spatial mutation index calculation unit 22, a timestamp verification unit 23, and an outlier marking unit 24. The spatial mutation index calculation unit 22 calculates the mutation index of the stitched edge points. ,in For edge point data values, It is the average of several adjacent voxels. For standard deviation, when Points exceeding a preset threshold are marked as spatial anomalies. The timestamp verification unit 23 verifies the timestamp difference between environmental data of adjacent grid cells. If the difference exceeds a preset time threshold, linear interpolation is used to generate intermediate time data, and points of time inconsistency are marked. The anomaly marking unit 24 marks the identified anomalies and transmits the anomaly information to the correction module 6.

[0039] The correction module 6 is used to receive the abnormal point information identified by the spatiotemporal verification module 5 and send the abnormal point information to the feature extraction module 2, triggering the feature extraction module 2 to re-extract the multimodal edge features at the abnormal points. The matching correction module 3 and the constraint interpolation module 4 then re-execute the matching, correction and interpolation processes until all verification indicators meet the preset threshold.

[0040] Output module 7 is used to output the seamlessly stitched integrated environmental data. The output includes the stitched data file, a verification report, and coordinate correction records. The verification report details the location and processing results of spatial anomalies, as well as timestamp differences and interpolation details at points of temporal inconsistency. In the final output stitched data, all environmental parameters at the boundaries of adjacent grid cells exhibit a continuous, smooth, and physically consistent transition, achieving true seamless stitching.

[0041] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention. Various embodiments of the present invention have been described above. The above description is exemplary and not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical applications, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for seamlessly stitching environmental data based on micro-scenes, characterized in that, Includes the following steps: Step 1: Obtain environmental data for at least two 3D mesh units to be stitched together. The environmental data includes spatial coordinates, physical parameters, timestamps, and data source identifiers. Step 2: Extract key feature points, gradient features, and texture structure from the adjacent surfaces of each 3D mesh cell; Step 3: Dynamically match the multimodal edge features extracted from adjacent grid cells, construct feature twins, calculate coordinate correction based on the matching results, and perform spatial coordinate correction on the grid edge points; Step 4: Based on the type of environmental data, select the corresponding physical equation from the preset physical equation library as the constraint condition, combine the original data and confidence weight of the adjacent grid edge points, and calculate the interpolation result of the edge points through the hybrid interpolation operator to complete the physical attribute transition; Step 5: Perform spatial continuity and temporal consistency checks on the stitched edge data, identify outliers, and generate a check report; Step 6: If the verification fails, the anomaly information is fed back to Step 2, the multimodal edge features at the anomaly point are re-extracted, and Steps 3-5 are executed again until all indicators meet the preset threshold. Step 7: Output the seamlessly stitched integrated environment data.

2. The seamless stitching method for environmental data based on micro-scenes according to claim 1, characterized in that, The physical equation library includes the Navier-Stokes equations, Fick's law, and terrain slope variation constraints. The selection of the corresponding physical equation as a constraint based on the type of environmental data includes: selecting the Navier-Stokes equation when the environmental data is wind speed or water flow; selecting Fick's law when the environmental data is pollutant concentration or temperature; and selecting terrain slope variation constraint when the environmental data is terrain elevation.

3. The seamless stitching method for environmental data based on micro-scenes according to claim 2, characterized in that, In step 4, the hybrid interpolation operator uses a weighted average method to calculate the interpolation results of the edge points. The calculation formula is as follows: ,in , These are the original data for adjacent grid edge points. , The weights are calculated based on data credibility. The reliability of the data is dynamically adjusted based on the accuracy, timeliness, and historical errors of the data source.

4. The seamless stitching method for environmental data based on micro-scenes according to claim 3, characterized in that, The spatial continuity verification in step 5 includes: calculating the abrupt change index of the edge points after splicing. ,in For edge point data values, It is the average of several adjacent voxels. For standard deviation, when Points exceeding a preset threshold are marked as spatial anomalies. The time consistency verification includes: verifying the timestamp difference of environmental data of adjacent grid cells; if the difference exceeds a preset time threshold, linear interpolation is used to generate intermediate time data and the time inconsistency point is marked.

5. The seamless stitching method for environmental data based on micro-scenes according to claim 4, characterized in that, Before extracting multimodal edge features, super-resolution reconstruction is performed on the edge data of low-resolution grid cells to improve the accuracy to the same level as high-resolution grid cells; Gaussian filtering is used to smooth the transition region from high-resolution grid cells to low-resolution grid cells. During the process of dynamic matching, constructing feature twins and calculating coordinate corrections, an initial confidence factor is assigned to the environmental data of each grid cell. The confidence factor is used for feature selection during feature matching and weight calculation of coordinate corrections. In the process of calculating the interpolation results of edge points using the hybrid interpolation operator, the confidence factor is dynamically adjusted according to the data update and historical error. The weights in the hybrid interpolation operator are calculated based on the adjusted confidence factor.

6. A seamless environmental data stitching system based on micro-scenes, characterized in that, include: An external data interface module is used to receive environmental data of the 3D mesh cells to be stitched together; The feature extraction module is used to extract key feature points, gradient features, and texture structure from the adjacent surfaces of each 3D mesh cell; The matching and correction module is used to dynamically match the multimodal edge features extracted from adjacent grid cells, construct feature twins, calculate coordinate correction based on the matching results, and perform spatial coordinate correction on grid edge points. The constraint interpolation module is used to select the corresponding physical equation as the constraint condition according to the type of environmental data, and calculate the interpolation result of the edge point by combining the original data and confidence weight of the adjacent grid edge point through a hybrid interpolation operator. The spatiotemporal verification module is used to verify the spatial continuity and temporal consistency of the stitched edge data, identify outliers, and generate a verification report. The correction module is used to receive the abnormal point information identified by the spatiotemporal verification module and send the abnormal point information to the feature extraction module to re-extract the multimodal edge features at the abnormal point; The results output module is used to output the seamlessly stitched integrated environment data.

7. A seamless environmental data stitching system based on micro-scenes according to claim 6, characterized in that, The feature extraction module includes: The data preprocessing unit is used to perform super-resolution reconstruction on the edge data of low-resolution grid cells and to smooth the region where high-resolution grid cells transition to low-resolution grid cells. The 3D SIFT feature extractor is used to identify key edge feature points and record their three-dimensional coordinates and attribute values. The 3D gradient calculator is used to calculate the spatial rate of change of data at the edges and obtain gradient features. Gray-level co-occurrence matrix processor, used to extract texture structure of edge regions.

8. The apparatus according to claim 7, characterized in that, The matching correction module includes: The DTW matching algorithm processor is used to dynamically match the multimodal edge features of adjacent grid cells and construct feature twins. The coordinate deviation calculation unit is used to calculate the coordinate correction amount based on the matching results; Coordinate converter, used to correct the spatial coordinates of grid edge points; The matching result evaluation unit is used to evaluate the matching accuracy. When the error exceeds a preset threshold, it triggers the re-extraction of features. A credibility assessment model is used to evaluate the credibility of environmental data for each grid cell. An initial confidence allocation unit is used to allocate an initial confidence factor to the environmental data of each grid cell. The confidence factor is used for feature filtering and weight calculation of coordinate correction during feature matching.

9. A seamless environmental data stitching system based on micro-scenes according to claim 8, characterized in that, The constraint interpolation module includes: A multiphysics equation library is used to store the physical equations corresponding to different types of environmental data; The equation selection unit is used to select the corresponding physical equation from the multiphysics equation library based on the type of environmental data. Hybrid interpolation operator calculator, used to combine physical equations, raw data of adjacent grid edge points and confidence weights to calculate interpolation results of edge points; The weighted fusion calculator is used to calculate the final fusion value of edge points based on the confidence factor.

10. A seamless environmental data stitching system based on micro-scenes according to claim 9, characterized in that, The spatiotemporal verification module includes: The spatial mutation index calculation unit is used to calculate the mutation index of the edge points after splicing and to identify spatial anomalies. The timestamp verification unit is used to verify the timestamp difference between environmental data of adjacent grid cells. When the difference exceeds a preset time threshold, time interpolation is triggered. Anomaly marking unit is used to mark anomalies and transmit the anomaly information to the correction module.

Images