Three-dimensional terrain model creation method and system based on multi-source data

Abstract

The application is suitable for the technical field of terrain modeling, and provides a three-dimensional terrain model creation method and system based on multi-source data, including the following steps: collecting multi-source terrain data, including satellite remote sensing data, laser radar point cloud data, unmanned aerial vehicle aerial photography data and ground measurement data; performing consistency processing on the multi-source terrain data, performing quality grading on the data based on spatial resolution, collection time and coverage integrity, and establishing a data priority weight matrix; extracting terrain features from data of different quality levels using a multi-scale feature pyramid network; introducing an attention mechanism in the feature extraction process to dynamically adjust the weights of features of different scales; and constructing a dynamic fusion function based on the data priority weight matrix and the feature weights of different scales. The application can dynamically adjust the weights of features of different scales, so that the model can better capture the multi-scale features of the terrain and improve the accuracy of feature extraction and the fusion effect.

Description

Technical Field

[0001] This invention relates to the field of terrain modeling technology, specifically to a method and system for creating three-dimensional terrain models based on multi-source data. Background Technology

[0002] With the continuous development of geographic information systems, remote sensing technology, and computer graphics, 3D terrain models are increasingly widely and critically applied in various fields such as urban planning, disaster monitoring, virtual reality, and military simulation. Traditional methods for creating 3D terrain models often rely on a single data source, such as satellite remote sensing data or lidar point cloud data. However, single data sources have their own limitations. To overcome the shortcomings of single data sources, multi-source data fusion technology has gradually become a research hotspot. However, current multi-source data fusion methods lack effective data quality assessment and priority determination mechanisms when processing data of different qualities and scales, leading to potential problems such as feature loss, discontinuity, or uneven accuracy in the fused terrain model. Furthermore, the physical constraints of the terrain itself are often ignored during the construction of 3D terrain models, resulting in models that may not conform to the physical laws of the actual geographical environment, affecting the model's realism and usability. Therefore, a method and system for creating 3D terrain models based on multi-source data is needed to address these issues. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a method and system for creating three-dimensional terrain models based on multi-source data, so as to solve the problems existing in the above-mentioned background technology.

[0004] This invention is implemented as follows: a method for creating a three-dimensional terrain model based on multi-source data, the method comprising the following steps: Collect multi-source terrain data, including satellite remote sensing data, lidar point cloud data, UAV aerial photography data, and ground measurement data; Consistency processing is performed on multi-source terrain data, and the data quality is graded based on spatial resolution, acquisition time and coverage integrity. A data priority weight matrix is ​​established based on the quality grading results. A multi-scale feature pyramid network is used to extract terrain features from data of different quality levels; an attention mechanism is introduced during the feature extraction process to dynamically adjust the weights of features at different scales. A dynamic fusion function is constructed based on the data priority weight matrix and feature weights at different scales; A topographic physical constraint model is established, including slope continuity constraint, water flow direction constraint, and elevation change constraint. The physical constraints are added as regularization terms to the model's objective function, and the alternating direction multiplier method is used to solve the constrained optimization problem.

[0005] As a further aspect of the present invention, the step of classifying data quality based on spatial resolution, acquisition time, and coverage integrity, and establishing a data priority weight matrix based on the quality classification results, specifically includes: The spatial resolution score is determined by calculating the ratio of the actual ground sampling distance of the data source to the target resolution. The data collection time score is determined based on the time difference between the data source collection date and the model construction baseline date. The coverage integrity score is determined by calculating the ratio of the number of valid data pixels in the target area to the total number of pixels. The overall quality score for each data source is obtained based on spatial resolution score, acquisition time score, and coverage integrity score. Establish a data priority weight matrix, where rows represent different data sources and columns represent different terrain regions; for each terrain region, calculate the relative weight based on the comprehensive quality score of each data source within that region.

[0006] As a further aspect of the present invention, the step of extracting terrain features from data of different quality levels using a multi-scale feature pyramid network specifically includes: A three-level feature pyramid network is constructed, including a coarse-scale feature extraction layer, a meso-scale feature extraction layer, and a fine-scale feature extraction layer; Data of different quality levels are input into the feature pyramid network; for high-quality data, fine-scale layers are used to extract detailed features; for medium-quality data, medium-scale layers are used to extract main features; and for low-quality data, coarse-scale layers are used to extract overall features. An attention mechanism is introduced between the different levels of the feature pyramid; the fusion weights of features at each scale are dynamically adjusted based on the quality level of the input data and the complexity of the terrain. Features extracted from each scale are fused across levels, and multi-scale information is integrated by feature splicing.

[0007] As a further aspect of the present invention, the step of constructing the dynamic fusion function specifically includes: Based on the data priority weight matrix, basic fusion weights are assigned to each data source; weight coefficients are assigned to features at different scales according to the feature weights at different scales. A dynamic fusion function based on weighted average is constructed, and a quality constraint term is introduced into the function to adjust the function parameters in real time according to the fusion effect.

[0008] As a further aspect of the present invention, the step of solving the constrained optimization problem using the alternating direction multiplier method specifically includes: The optimization problem is decomposed into multiple sub-problems, and the data fitting term and each physical constraint term are handled separately. By introducing auxiliary variables and Lagrange multipliers, the constrained optimization problem is transformed into an unconstrained optimization problem; Alternately update the original and dual variables to gradually approach the optimal solution; set a convergence criterion and stop the calculation when the change in the objective function is less than a threshold or the maximum number of iterations is reached.

[0009] Another object of the present invention is to provide a three-dimensional terrain model creation system based on multi-source data, the system comprising: The multi-source data acquisition module is used to collect multi-source terrain data, including satellite remote sensing data, lidar point cloud data, UAV aerial photography data, and ground measurement data. The priority weight matrix module is used to perform consistency processing on multi-source terrain data, classify the data quality based on spatial resolution, acquisition time and coverage integrity, and establish a data priority weight matrix based on the quality classification results. The terrain feature extraction module is used to extract terrain features from data of different quality levels using a multi-scale feature pyramid network; an attention mechanism is introduced during the feature extraction process to dynamically adjust the weights of features at different scales. The dynamic fusion function module is used to construct dynamic fusion functions based on data priority weight matrix and feature weights at different scales; The physical constraint optimization module is used to establish a terrain physical constraint model, including slope continuity constraints, water flow direction constraints, and elevation change constraints. The physical constraints are added as regularization terms to the model optimization objective function, and the alternating direction multiplier method is used to solve the constrained optimization problem.

[0010] As a further aspect of the present invention, the priority weight matrix module includes: The spatial resolution scoring unit is used to calculate the ratio of the actual ground sampling distance of the data source to the target resolution to determine the spatial resolution score; The data collection time scoring unit is used to determine the data collection time score based on the time difference between the data source collection date and the model construction baseline date. The coverage integrity scoring unit is used to calculate the ratio of the number of valid data pixels in the target area to the total number of pixels to determine the coverage integrity score; The comprehensive quality score calculation unit is used to obtain the comprehensive quality score for each data source based on spatial resolution score, acquisition time score, and coverage integrity score; The weight matrix construction unit is used to build a data priority weight matrix. The rows of the matrix represent different data sources, and the columns represent different terrain regions. For each terrain region, the relative weight is calculated based on the comprehensive quality score of each data source in that region.

[0011] As a further embodiment of the present invention, the terrain feature extraction module includes: The pyramid network building unit is used to construct a three-level feature pyramid network, including a coarse-scale feature extraction layer, a meso-scale feature extraction layer, and a fine-scale feature extraction layer. The multi-scale feature extraction unit is used to input data of different quality levels into the feature pyramid network. For high-quality data, fine-scale layers are used to extract detailed features; for medium-quality data, medium-scale layers are used to extract main features; and for low-quality data, coarse-scale layers are used to extract overall features. The fusion weight adjustment unit is used to introduce an attention mechanism between different levels of the feature pyramid; it dynamically adjusts the fusion weights of features at each scale based on the quality level and terrain complexity of the input data. The multi-scale feature fusion unit is used to fuse features extracted from various scales across levels, integrating multi-scale information by feature splicing.

[0012] As a further aspect of the present invention, the dynamic fusion function module includes: The weight allocation unit is used to assign basic fusion weights to each data source based on the data priority weight matrix; and to assign weight coefficients to features at different scales according to the feature weights at different scales. The fusion function construction unit is used to construct a dynamic fusion function based on weighted average. It introduces quality constraint terms into the function and adjusts the function parameters in real time according to the fusion effect.

[0013] As a further aspect of the present invention, the physical constraint optimization module includes: The optimization problem decomposition unit is used to decompose the optimization problem into multiple sub-problems, and to handle the data fitting term and each physical constraint term respectively. The optimization problem transformation unit is used to introduce auxiliary variables and Lagrange multipliers to transform a constrained optimization problem into an unconstrained optimization problem. The optimal solution determination unit is used to alternately update the original and dual variables to gradually approach the optimal solution; a convergence criterion is set, and the calculation stops when the change in the objective function is less than a threshold or the maximum number of iterations is reached.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention collects multi-source terrain data, performs consistency processing and quality grading, and establishes a data priority weight matrix. This fully leverages the advantages of different data sources and dynamically adjusts the weights of data in model construction based on data quality, generating a more accurate and complete terrain model. A multi-scale feature pyramid network combined with an attention mechanism is employed to extract terrain features from data of different quality levels and dynamically adjust the weights of features at different scales. This allows the model to better capture multi-scale terrain features, improving the accuracy of feature extraction and fusion effects. A terrain physical constraint model is established, and physical constraints are added as regularization terms to the model's optimization objective function. The alternating direction multiplier method is used to solve the constrained optimization problem, ensuring that the generated 3D terrain model conforms to the physical laws of the actual geographical environment, making it more suitable for various practical application scenarios. Attached Figure Description

[0015] Figure 1 A flowchart of a method for creating a 3D terrain model based on multi-source data.

[0016] Figure 2 A flowchart for establishing a data priority weight matrix in a method for creating a 3D terrain model based on multi-source data.

[0017] Figure 3 A flowchart illustrating the method for extracting terrain features in a multi-source data-based 3D terrain model creation process.

[0018] Figure 4 This is a flowchart illustrating the construction of a dynamic fusion function in a method for creating 3D terrain models based on multi-source data.

[0019] Figure 5 This is a flowchart illustrating the use of the alternating direction multiplier method to solve constrained optimization problems in a method for creating 3D terrain models based on multi-source data.

[0020] Figure 6 A schematic diagram of the structure of a system for creating 3D terrain models based on multi-source data. Detailed Implementation

[0021] To make the objectives, technical solutions, 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 merely illustrative and not intended to limit the invention.

[0022] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0023] like Figure 1 As shown in the figure, this embodiment of the invention provides a method for creating a three-dimensional terrain model based on multi-source data, the method comprising the following steps: The S100 collects multi-source terrain data, including satellite remote sensing data, lidar point cloud data, UAV aerial photography data, and ground measurement data. S200 performs consistency processing on multi-source terrain data, classifies the data quality based on spatial resolution, acquisition time and coverage integrity, and establishes a data priority weight matrix based on the quality classification results. S300 uses a multi-scale feature pyramid network to extract terrain features from data of different quality levels; an attention mechanism is introduced during the feature extraction process to dynamically adjust the weights of features at different scales. S400 constructs a dynamic fusion function based on the data priority weight matrix and feature weights at different scales; S500 establishes a terrain physical constraint model, including slope continuity constraints, water flow direction constraints, and elevation change constraints; the physical constraints are added as regularization terms to the model optimization objective function, and the alternating direction multiplier method is used to solve the constrained optimization problem.

[0024] In this embodiment of the invention, the collected multi-source terrain data includes satellite remote sensing data, lidar point cloud data, UAV aerial photography data, and ground measurement data. After obtaining the multi-source terrain data, the collected data needs to be formatted and coordinate system unified. The data is then graded according to spatial resolution, acquisition time, and coverage integrity, and a data priority weight matrix is ​​established based on the quality grading results. This fully utilizes the advantages of different data sources, facilitating dynamic adjustment of data weights in model construction based on data quality, thereby improving data utilization efficiency and generating a more accurate and complete terrain model. Then, a multi-scale feature pyramid network is used to extract terrain features from data of different quality levels. It should be noted that an attention mechanism is introduced during feature extraction to dynamically adjust the weights of features at different scales, enabling the model to better capture multi-scale terrain features, improving the accuracy of feature extraction and fusion effect, and generating a more detailed and feature-rich 3D terrain model. Next, a dynamic fusion function is constructed based on the data priority weight matrix and the weights of features at different scales. In addition, a terrain physical constraint model is established, including slope continuity constraints, water flow direction constraints, and elevation change constraints. Slope continuity constraints are used to establish continuity constraints based on terrain gradients, ensuring that slope changes between adjacent terrain units are natural and gradual, avoiding unreasonable steep changes. Water flow direction constraints are constructed based on hydrogeological principles to ensure that the terrain surface conforms to natural water flow patterns, with slope aspect consistent with water flow direction. Elevation abrupt change constraints are used to identify and protect reasonable elevation abrupt changes (such as cliffs and embankments) while suppressing unreasonable elevation jumps, maintaining the physical rationality of the terrain morphology. Next, these three physical constraints are added as regularization terms to the model's optimization objective function, dynamically adjusting the constraint strength according to terrain complexity, and using the alternating direction multiplier method to solve the constrained optimization problem. This further ensures that the generated 3D terrain model conforms to the physical laws of the actual geographical environment, such as slope continuity, reasonable water flow direction, and elevation abrupt changes that reflect reality, greatly improving the model's realism and reliability, making it more suitable for various practical application scenarios.

[0025] like Figure 2 As shown in the preferred embodiment of the present invention, the step of classifying data quality based on spatial resolution, acquisition time, and coverage integrity, and establishing a data priority weight matrix based on the quality classification results, specifically includes: S201, Calculate the ratio of the actual ground sampling distance of the data source to the target resolution to determine the spatial resolution score; S202, determine the collection time score based on the time difference between the data source collection date and the model construction baseline date; S203, Calculate the ratio of the number of valid data pixels in the target area to the total number of pixels to determine the coverage integrity score; S204, the comprehensive quality score of each data source is obtained based on spatial resolution score, acquisition time score and coverage integrity score; S205. Establish a data priority weight matrix, where rows represent different data sources and columns represent different terrain regions. For each terrain region, calculate the relative weight based on the comprehensive quality score of each data source within that region.

[0026] In this embodiment of the invention, after the spatial resolution score, acquisition time score, and coverage integrity score are determined, a comprehensive quality score is obtained through weighted summation. Based on the comprehensive quality score, the quality indicators are divided into three levels: high, medium, and low. Then, a data priority weight matrix can be established, where rows represent different data sources and columns represent different terrain regions. For each terrain region, the relative weight is calculated based on the comprehensive quality scores of each data source within that region. The weight calculation uses the analytic hierarchy process (AHP), comprehensively considering the relative importance of the three quality indicators. The final generated weight matrix is ​​used to guide the subsequent data fusion process.

[0027] like Figure 3 As shown, in a preferred embodiment of the present invention, the step of extracting terrain features from data of different quality levels using a multi-scale feature pyramid network specifically includes: S301, construct a three-level feature pyramid network, including a coarse-scale feature extraction layer, a medium-scale feature extraction layer, and a fine-scale feature extraction layer; S302: Input data of different quality levels into the feature pyramid network respectively; for high-quality data, use fine-scale layers to extract detailed features; for medium-quality data, use medium-scale layers to extract main features; for low-quality data, use coarse-scale layers to extract overall features. S303 introduces an attention mechanism between different levels of the feature pyramid; it dynamically adjusts the fusion weights of features at each scale based on the quality level of the input data and the complexity of the terrain. S304 integrates features extracted from various scales across different levels by using feature splicing to combine multi-scale information.

[0028] In this embodiment of the invention, a three-level feature pyramid network is constructed. The coarse-scale layer uses large-size convolutional kernels to extract macroscopic topographic features, such as mountain range orientation and major watershed distribution. The meso-scale layer uses medium-size convolutional kernels to extract mesoscopic topographic features, such as slope morphology and valley structure. The fine-scale layer uses small-size convolutional kernels to extract microscopic topographic features, such as minor undulations and micro-landforms. Data of different quality levels are then input into the feature pyramid network. For high-quality data, the fine-scale layer is used to extract detailed features; for medium-quality data, the meso-scale layer is used to extract main features; and for low-quality data, only the coarse-scale layer is used to extract general features. An attention mechanism is introduced between the different levels of the feature pyramid to achieve adaptive feature selection and enhancement through attention weights. Finally, the features extracted at each scale are fused across levels, and multi-scale information is integrated using feature concatenation to output a multi-layered topographic feature map with rich semantic information.

[0029] like Figure 4 As shown, in a preferred embodiment of the present invention, the step of constructing the dynamic fusion function specifically includes: S401 assigns basic fusion weights to each data source based on the data priority weight matrix; and assigns weight coefficients to features at different scales according to the feature weights at different scales. S402, construct a dynamic fusion function based on weighted average, introduce quality constraint terms into the function, and adjust the function parameters in real time according to the fusion effect.

[0030] In this embodiment of the invention, based on the previously determined data priority weight matrix and feature weights at different scales, basic fusion weights are assigned to each data source, and weight coefficients are assigned to features at each scale. Then, a dynamic fusion function is constructed, incorporating quality constraint terms to ensure that low-quality data does not negatively impact the fusion result. The function parameters are adjusted in real time based on the fusion effect, automatically increasing the weight of high-quality data in areas of drastic feature changes and appropriately reducing weight differences in flat areas to improve computational efficiency; through iterative optimization, the fusion function is made adaptable to different terrain conditions.

[0031] like Figure 5 As shown, in a preferred embodiment of the present invention, the step of solving the constrained optimization problem using the alternating direction multiplier method specifically includes: S501 decomposes the optimization problem into multiple sub-problems, and handles the data fitting term and each physical constraint term separately. S502 introduces auxiliary variables and Lagrange multipliers to transform the constrained optimization problem into an unconstrained optimization problem; S503 alternately updates the original and dual variables to gradually approach the optimal solution; a convergence criterion is set, and the calculation stops when the change in the objective function is less than a threshold or the maximum number of iterations is reached.

[0032] In this embodiment of the invention, the optimization problem is decomposed into multiple sub-problems, such as a data fitting problem, a slope constraint problem, a flow constraint problem, and an elevation change constraint problem. Then, by introducing auxiliary variables and Lagrange multipliers, the original problem with strict constraints is transformed into an equivalent unconstrained optimization problem. The Lagrange multipliers act as a "penalty" for violations of constraints, thereby guiding the solution to satisfy the constraints during the solution process. The variables in the transformed problem (including the original model variables, the introduced auxiliary variables, and the dual variables / Lagrange multipliers) are updated alternately. That is, while keeping other variables fixed, one subset of variables is optimized sequentially. Through this iterative alternation, the contradiction between data fitting and physical constraints is gradually reconciled, causing the solution to gradually approach the optimal solution that satisfies all conditions.

[0033] like Figure 6 As shown, this embodiment of the invention also provides a three-dimensional terrain model creation system based on multi-source data, the system comprising: The multi-source data acquisition module 100 is used to acquire multi-source terrain data, including satellite remote sensing data, lidar point cloud data, UAV aerial photography data and ground measurement data; The priority weight matrix module 200 is used to perform consistency processing on multi-source terrain data, classify the data quality based on spatial resolution, acquisition time and coverage integrity, and establish a data priority weight matrix based on the quality classification results. The terrain feature extraction module 300 is used to extract terrain features from data of different quality levels using a multi-scale feature pyramid network; an attention mechanism is introduced during the feature extraction process to dynamically adjust the weights of features at different scales. The dynamic fusion function module 400 is used to construct a dynamic fusion function based on the data priority weight matrix and feature weights at different scales. The physical constraint optimization module 500 is used to establish a terrain physical constraint model, including slope continuity constraints, water flow direction constraints, and elevation change constraints. The physical constraints are added as regularization terms to the model optimization objective function, and the alternating direction multiplier method is used to solve the constrained optimization problem.

[0034] In a preferred embodiment of the present invention, the priority weight matrix module 200 includes: The spatial resolution scoring unit is used to calculate the ratio of the actual ground sampling distance of the data source to the target resolution to determine the spatial resolution score; The data collection time scoring unit is used to determine the data collection time score based on the time difference between the data source collection date and the model construction baseline date. The coverage integrity scoring unit is used to calculate the ratio of the number of valid data pixels in the target area to the total number of pixels to determine the coverage integrity score; The comprehensive quality score calculation unit is used to obtain the comprehensive quality score for each data source based on spatial resolution score, acquisition time score, and coverage integrity score; The weight matrix construction unit is used to build a data priority weight matrix. The rows of the matrix represent different data sources, and the columns represent different terrain regions. For each terrain region, the relative weight is calculated based on the comprehensive quality score of each data source in that region.

[0035] In a preferred embodiment of the present invention, the terrain feature extraction module 300 includes: The pyramid network building unit is used to construct a three-level feature pyramid network, including a coarse-scale feature extraction layer, a meso-scale feature extraction layer, and a fine-scale feature extraction layer. The multi-scale feature extraction unit is used to input data of different quality levels into the feature pyramid network. For high-quality data, fine-scale layers are used to extract detailed features; for medium-quality data, medium-scale layers are used to extract main features; and for low-quality data, coarse-scale layers are used to extract overall features. The fusion weight adjustment unit is used to introduce an attention mechanism between different levels of the feature pyramid; it dynamically adjusts the fusion weights of features at each scale based on the quality level and terrain complexity of the input data. The multi-scale feature fusion unit is used to fuse features extracted from various scales across levels, integrating multi-scale information by feature splicing.

[0036] In a preferred embodiment of the present invention, the dynamic fusion function module 400 includes: The weight allocation unit is used to assign basic fusion weights to each data source based on the data priority weight matrix; and to assign weight coefficients to features at different scales according to the feature weights at different scales. The fusion function construction unit is used to construct a dynamic fusion function based on weighted average. It introduces quality constraint terms into the function and adjusts the function parameters in real time according to the fusion effect.

[0037] In a preferred embodiment of the present invention, the physical constraint optimization module 500 includes: The optimization problem decomposition unit is used to decompose the optimization problem into multiple sub-problems, and to handle the data fitting term and each physical constraint term respectively. The optimization problem transformation unit is used to introduce auxiliary variables and Lagrange multipliers to transform a constrained optimization problem into an unconstrained optimization problem. The optimal solution determination unit is used to alternately update the original and dual variables to gradually approach the optimal solution; a convergence criterion is set, and the calculation stops when the change in the objective function is less than a threshold or the maximum number of iterations is reached.

[0038] The above description only details the preferred embodiments of the present invention and is not intended to limit the present 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.

[0039] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0040] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0041] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. A method for creating a 3D terrain model based on multi-source data, characterized in that, The method includes the following steps: Collect multi-source terrain data, including satellite remote sensing data, lidar point cloud data, UAV aerial photography data, and ground measurement data; Consistency processing is performed on multi-source terrain data, and the data quality is graded based on spatial resolution, acquisition time and coverage integrity. A data priority weight matrix is ​​established based on the quality grading results. A multi-scale feature pyramid network is used to extract terrain features from data of different quality levels; an attention mechanism is introduced during the feature extraction process to dynamically adjust the weights of features at different scales. A dynamic fusion function is constructed based on the data priority weight matrix and feature weights at different scales; A topographic physical constraint model is established, including slope continuity constraint, water flow direction constraint, and elevation change constraint. The physical constraints are added as regularization terms to the model's objective function, and the alternating direction multiplier method is used to solve the constrained optimization problem.

2. The method for creating a three-dimensional terrain model based on multi-source data according to claim 1, characterized in that, The steps of classifying data quality based on spatial resolution, acquisition time, and coverage integrity, and establishing a data priority weight matrix based on the quality classification results, specifically include: The spatial resolution score is determined by calculating the ratio of the actual ground sampling distance of the data source to the target resolution. The data collection time score is determined based on the time difference between the data source collection date and the model construction baseline date. The coverage integrity score is determined by calculating the ratio of the number of valid data pixels in the target area to the total number of pixels. The overall quality score for each data source is obtained based on spatial resolution score, acquisition time score, and coverage integrity score. Establish a data priority weight matrix, where rows represent different data sources and columns represent different terrain regions; for each terrain region, calculate the relative weight based on the comprehensive quality score of each data source within that region.

3. The method for creating a three-dimensional terrain model based on multi-source data according to claim 1, characterized in that, The steps for extracting terrain features from data of different quality levels using a multi-scale feature pyramid network include: A three-level feature pyramid network is constructed, including a coarse-scale feature extraction layer, a meso-scale feature extraction layer, and a fine-scale feature extraction layer; Data of different quality levels are input into the feature pyramid network; for high-quality data, fine-scale layers are used to extract detailed features; for medium-quality data, medium-scale layers are used to extract main features; and for low-quality data, coarse-scale layers are used to extract overall features. An attention mechanism is introduced between the different levels of the feature pyramid; the fusion weights of features at each scale are dynamically adjusted based on the quality level of the input data and the complexity of the terrain. Features extracted from each scale are fused across levels, and multi-scale information is integrated by feature splicing.

4. The method for creating a three-dimensional terrain model based on multi-source data according to claim 1, characterized in that, The steps for constructing the dynamic fusion function specifically include: Based on the data priority weight matrix, basic fusion weights are assigned to each data source; weight coefficients are assigned to features at different scales according to the feature weights at different scales. A dynamic fusion function based on weighted average is constructed, and a quality constraint term is introduced into the function to adjust the function parameters in real time according to the fusion effect.

5. The method for creating a three-dimensional terrain model based on multi-source data according to claim 1, characterized in that, The steps for solving constrained optimization problems using the alternating direction multiplier method specifically include: The optimization problem is decomposed into multiple sub-problems, and the data fitting term and each physical constraint term are handled separately. By introducing auxiliary variables and Lagrange multipliers, the constrained optimization problem is transformed into an unconstrained optimization problem; Alternately update the original and dual variables to gradually approach the optimal solution; set a convergence criterion and stop the calculation when the change in the objective function is less than a threshold or the maximum number of iterations is reached.

6. A three-dimensional terrain model creation system based on multi-source data, characterized in that, The system includes: The multi-source data acquisition module is used to collect multi-source terrain data, including satellite remote sensing data, lidar point cloud data, UAV aerial photography data, and ground measurement data. The priority weight matrix module is used to perform consistency processing on multi-source terrain data, classify the data quality based on spatial resolution, acquisition time and coverage integrity, and establish a data priority weight matrix based on the quality classification results. The terrain feature extraction module is used to extract terrain features from data of different quality levels using a multi-scale feature pyramid network; an attention mechanism is introduced during the feature extraction process to dynamically adjust the weights of features at different scales. The dynamic fusion function module is used to construct dynamic fusion functions based on data priority weight matrix and feature weights at different scales; The physical constraint optimization module is used to establish a terrain physical constraint model, including slope continuity constraints, water flow direction constraints, and elevation change constraints. The physical constraints are added as regularization terms to the model optimization objective function, and the alternating direction multiplier method is used to solve the constrained optimization problem.

7. The three-dimensional terrain model creation system based on multi-source data according to claim 6, characterized in that, The priority weight matrix module includes: The spatial resolution scoring unit is used to calculate the ratio of the actual ground sampling distance of the data source to the target resolution to determine the spatial resolution score; The data collection time scoring unit is used to determine the data collection time score based on the time difference between the data source collection date and the model construction baseline date. The coverage integrity scoring unit is used to calculate the ratio of the number of valid data pixels in the target area to the total number of pixels to determine the coverage integrity score; The comprehensive quality score calculation unit is used to obtain the comprehensive quality score for each data source based on spatial resolution score, acquisition time score, and coverage integrity score; The weight matrix construction unit is used to build a data priority weight matrix. The rows of the matrix represent different data sources, and the columns represent different terrain regions. For each terrain region, the relative weight is calculated based on the comprehensive quality score of each data source in that region.

8. The three-dimensional terrain model creation system based on multi-source data according to claim 6, characterized in that, The terrain feature extraction module includes: The pyramid network building unit is used to construct a three-level feature pyramid network, including a coarse-scale feature extraction layer, a meso-scale feature extraction layer, and a fine-scale feature extraction layer. The multi-scale feature extraction unit is used to input data of different quality levels into the feature pyramid network. For high-quality data, fine-scale layers are used to extract detailed features; for medium-quality data, medium-scale layers are used to extract main features; and for low-quality data, coarse-scale layers are used to extract overall features. The fusion weight adjustment unit is used to introduce an attention mechanism between different levels of the feature pyramid; it dynamically adjusts the fusion weights of features at each scale based on the quality level and terrain complexity of the input data. The multi-scale feature fusion unit is used to fuse features extracted from various scales across levels, integrating multi-scale information by feature splicing.

9. The three-dimensional terrain model creation system based on multi-source data according to claim 6, characterized in that, The dynamic fusion function module includes: The weight allocation unit is used to assign basic fusion weights to each data source based on the data priority weight matrix; and to assign weight coefficients to features at different scales according to the feature weights at different scales. The fusion function construction unit is used to construct a dynamic fusion function based on weighted average. It introduces quality constraint terms into the function and adjusts the function parameters in real time according to the fusion effect.

10. The three-dimensional terrain model creation system based on multi-source data according to claim 6, characterized in that, The physical constraint optimization module includes: The optimization problem decomposition unit is used to decompose the optimization problem into multiple sub-problems, and to handle the data fitting term and each physical constraint term respectively. The optimization problem transformation unit is used to introduce auxiliary variables and Lagrange multipliers to transform a constrained optimization problem into an unconstrained optimization problem. The optimal solution determination unit is used to alternately update the original and dual variables to gradually approach the optimal solution; a convergence criterion is set, and the calculation stops when the change in the objective function is less than a threshold or the maximum number of iterations is reached.

Images