A multi-angle remote sensing image screening and collecting method for satellite three-dimensional modeling

Abstract

The application provides a multi-angle remote sensing image screening and collection method for satellite three-dimensional modeling, and relates to the technical field of remote sensing image processing. The method comprises the following steps: obtaining task parameters and a digital elevation model of a target area, calculating terrain fluctuation characteristic parameters according to the digital elevation model, and dividing the target area into multiple grid units according to the terrain fluctuation characteristic parameters; determining a constraint parameter set of each grid unit according to the terrain fluctuation characteristic parameters, performing multi-dimensional screening from an image database according to the task parameters for each grid unit, and generating a candidate image set; according to the constraint parameter set, including images meeting the conditions into an effective image pair set; according to the scoring weight coefficient of each grid unit, selecting an image pair with a score greater than a preset score and a highest comprehensive score as a matching result, generating a corresponding unique identification code, and outputting a stereo image pair data set. The application has the technical effect of improving the matching of the screened stereo image pair and the actual terrain characteristics.

Description

Technical Field

[0001] This application relates to the field of remote sensing image processing technology, specifically to a method, system, device, and medium for multi-angle remote sensing image screening and acquisition for satellite 3D modeling. Background Technology

[0002] With the rapid development of satellite remote sensing technology and the increasing demand for 3D geographic information, stereo modeling based on multi-angle remote sensing imagery has become an important technical means to acquire high-precision digital elevation models and 3D geospatial information. In practical applications, when performing 3D modeling for different target areas, it is necessary to select the most suitable stereo image pairs from a massive satellite image database to ensure modeling accuracy and efficiency. However, due to the significant differences in the geometric configuration requirements of stereo image pairs in different terrain regions, and the varying sensitivities and tolerances of different terrain conditions such as plains, hills, and mountains to technical indicators such as base-to-height ratio, observation angle, and image quality, the selection and matching of stereo image pairs becomes a complex multi-constraint optimization problem.

[0003] Currently, a unified screening standard and a fixed evaluation system are typically used to address the selection of stereo image pairs. Candidate image pairs are screened by setting globally consistent thresholds for base-to-height ratio, angular symmetry requirements, and quality standards. Then, the image pairs are evaluated and ranked based on a single dimension or simple weighting. While this method is effective for areas with relatively uniform terrain, the unified screening standard cannot adapt to the diverse needs of different terrain conditions, and cannot provide precise screening for different terrain features, resulting in poor matching between the selected stereo image pairs and the actual terrain characteristics. Summary of the Invention

[0004] This application provides a method, system, device, and medium for multi-angle remote sensing image screening and acquisition for satellite 3D modeling, which improves the matching of the screened stereo image pairs with actual terrain features.

[0005] Firstly, this application provides a method for screening and acquiring multi-angle remote sensing images for satellite 3D modeling. The method includes: acquiring task parameters and a digital elevation model (DEM) of a target area; calculating terrain undulation characteristic parameters of the target area based on the DEM; and dividing the target area into multiple grid cells based on the terrain undulation characteristic parameters; wherein different terrain undulation degrees correspond to different grid sizes; determining a set of constraint parameters for each grid cell based on the terrain undulation characteristic parameters, the set of constraint parameters including: a base-to-height ratio threshold range, an angle symmetry threshold, and a scoring weight coefficient; and for each grid cell, based on... According to the task parameters, multi-dimensional filtering is sequentially performed from the image database to generate a candidate image set; according to the constraint parameter set, it is determined whether each image pair in the candidate image set meets the stereo image pair condition, and the image pairs that meet the stereo image pair condition are included in the effective image pair set; according to the scoring weight coefficient of each grid unit, the comprehensive score of each image pair in the effective image pair set of each grid unit is calculated, and the image pair with the highest comprehensive score that is greater than the preset score in each grid unit is selected as the matching result, a unique identifier code corresponding to the matching result is generated, and the stereo image pair dataset is output according to the unique identifier code.

[0006] By adopting the above technical solution, and calculating terrain undulation characteristic parameters based on a digital elevation model and then adaptively dividing the grid accordingly, differentiated grid sizes can be set for different terrain conditions, achieving terrain feature-guided regional subdivision. By defining a set of constraint parameters for each grid cell, including a base-to-height ratio threshold range, an angular symmetry threshold, and a scoring weight coefficient, a terrain adaptability screening standard system is established, overcoming the technical deficiency of a uniform screening standard that cannot adapt to terrain differences. Based on this, a candidate image set is generated through multi-dimensional screening, and the stereo image pair conditions are judged based on the constraint parameter set, ensuring the matching of the screening results with terrain features. Finally, the comprehensive score of each image pair is calculated using the scoring weight coefficient, and the optimal matching result is selected, achieving precise image pair selection and improving the matching of the selected stereo image pairs with actual terrain features.

[0007] Secondly, this application provides a multi-angle remote sensing image screening and acquisition system for satellite 3D modeling, the system comprising: an acquisition module, a determination module, a generation module, a judgment module, and a calculation module; wherein, The acquisition module is used to acquire the task parameters and digital elevation model of the target area, calculate the terrain undulation feature parameters of the target area based on the digital elevation model, and divide the target area into multiple grid cells based on the terrain undulation feature parameters; wherein different terrain undulation degrees correspond to different grid sizes; the determination module is used to determine the constraint parameter set of each grid cell based on the terrain undulation feature parameters, the constraint parameter set including: base-to-height ratio threshold range, angular symmetry threshold, and scoring weight coefficient; the generation module is used to, for each grid cell, sequentially retrieve data from the image database based on the task parameters. A multi-dimensional filtering process is performed to generate a candidate image set. The judgment module is used to determine whether each image pair in the candidate image set meets the stereo image pair condition based on the constraint parameter set, and to include the image pairs that meet the stereo image pair condition into the effective image pair set. The calculation module is used to calculate the comprehensive score of each image pair in the effective image pair set of each grid unit based on the scoring weight coefficient of each grid unit, select the image pair in each grid unit that is greater than a preset score and has the highest comprehensive score as the matching result, generate a unique identifier code corresponding to the matching result, and output the stereo image pair dataset based on the unique identifier code.

[0008] Thirdly, this application provides an electronic device that adopts the following technical solution: it includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes a computer program such as any of the above-mentioned methods for multi-angle remote sensing image screening and acquisition for satellite 3D modeling.

[0009] Fourthly, this application provides a computer-readable storage medium that employs the following technical solution: storing a computer program capable of being loaded by a processor and executing any of the above-mentioned methods for multi-angle remote sensing image screening and acquisition for satellite 3D modeling.

[0010] In summary, this application includes at least one of the following beneficial technical effects: By calculating terrain undulation characteristic parameters based on a digital elevation model and then adaptively dividing the grid accordingly, differentiated grid sizes can be set for different terrain conditions, achieving terrain feature-guided regional subdivision. A terrain adaptability screening standard system was established by defining a set of constraint parameters for each grid cell, including a base-to-height ratio threshold range, an angular symmetry threshold, and a scoring weight coefficient, overcoming the technical limitation of a uniform screening standard that cannot adapt to terrain differences. Based on this, a candidate image set is generated through multi-dimensional screening, and stereo image pair conditions are judged based on the constraint parameter set, ensuring the matching of the screening results with terrain features. Finally, the comprehensive score of each image pair is calculated using the scoring weight coefficient, and the optimal matching result is selected, achieving precise image pair selection and improving the matching of the selected stereo image pairs with actual terrain features. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating a method for screening and acquiring multi-angle remote sensing images for satellite 3D modeling, as provided in an embodiment of this application. Figure 2 This is a schematic diagram of the logical flow of a data transmission method provided in an embodiment of this application; Figure 3 This is a schematic diagram of a scenario for multi-angle remote sensing image selection and acquisition based on satellite 3D modeling, provided in an embodiment of this application. Figure 4 This is a schematic diagram of the structure of a multi-angle remote sensing image screening and acquisition system for satellite 3D modeling provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0012] Explanation of reference numerals in the attached figures: 1000, electronic device; 1001, processor; 1002, communication bus; 1003, user interface; 1004, network interface; 1005, memory. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification 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.

[0014] In the description of the embodiments in this application, words such as "illustrative," "for example," or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "illustrative," "for example," or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of words such as "illustrative," "for example," or "for example" is intended to present the relevant concepts in a specific manner.

[0015] Figure 1 This is a flowchart illustrating a method for screening and acquiring multi-angle remote sensing images for satellite 3D modeling, provided in an embodiment of this application. Figure 1 As shown, the method includes S101-S105: S101: Obtain the task parameters and digital elevation model of the target area, calculate the terrain undulation characteristic parameters of the target area based on the digital elevation model, and divide the target area into multiple grid units based on the terrain undulation characteristic parameters; wherein, different terrain undulation degrees correspond to different grid sizes.

[0016] The system first acquires the task parameters for the target area, including core parameters such as the latitude and longitude boundaries of the monitoring area, the base-to-height ratio requirement, the lateral tilt angle range, the stereo mode, the temporal threshold, and the resolution threshold. These parameters constitute the baseline constraints for subsequent image selection and acquisition. Simultaneously, it acquires digital elevation model (DEM) data covering the target area. DEM is raster data that records surface elevation information, typically with a spatial resolution of 30 meters, providing basic elevation data for terrain analysis. Traditional fixed-grid methods have significant drawbacks when dealing with complex and varied terrain. Using too small a grid in flat areas wastes computational resources, while using too large a grid in complex terrain leads to decreased stereo matching accuracy. Therefore, an adaptive grid generation mechanism is needed to address this issue.

[0017] The system is based on a digital elevation model and uses the fitted surface method to calculate the slope value of each pixel within the target area. The slope calculation formula is as follows: ,in These represent the elevation change rates in the x and y directions, respectively. A slope raster map covering the entire target area is generated using this formula. Next, the target area is coarsely divided according to a preset initial grid size, typically 10km × 10km, to ensure sufficient coverage of terrain samples without over-subdivision. Statistical analysis is then performed on the slope values ​​of all pixels within each initial grid, calculating the slope standard deviation. , where slopei represents the slope value of the i-th pixel, slope_avg represents the average slope within the grid, n represents the number of effective pixels within the grid, and the slope standard deviation reflects the degree of change in terrain undulation within the grid, which is a key indicator for quantifying terrain complexity.

[0018] The terrain undulation characteristic parameter is the calculated slope standard deviation, which effectively distinguishes the complexity of different terrain types. When the slope standard deviation is ≤5°, it indicates that the terrain is relatively flat with little surface undulation, requiring relatively low visual accuracy for 3D matching. When 5° < slope standard deviation < 12°, it indicates that the terrain is of medium complexity, with some undulation but relatively regular changes. When the slope standard deviation is ≥12°, it indicates that the terrain is of high complexity, with dramatic undulations, requiring higher geometric accuracy for 3D matching. Based on different slope standard deviation ranges, the system automatically determines the corresponding grid size: low-complexity terrain corresponds to a 5km × 5km grid size, medium-complexity terrain to a 3km × 3km grid size, and high-complexity terrain to a 1km × 1km grid size, achieving dynamic matching between grid size and terrain complexity.

[0019] Finally, the target area is re-divided according to the determined grid size, generating a set of non-uniform but fully covered grid cells. Each grid cell has a unique grid identifier and corresponding terrain undulation feature parameter values. This adaptive grid partitioning mechanism demonstrates significant effectiveness. In flat areas, using a larger grid reduces unnecessary computational overhead and improves processing efficiency, while in complex terrain areas, using a smaller grid ensures the accuracy requirements of stereo matching and avoids matching errors caused by drastic terrain changes. Compared to traditional fixed-grid methods, this method achieves a dynamic balance between computational efficiency and matching accuracy while maintaining full coverage, laying a solid foundation for subsequent grid-based differentiated stereo image processing.

[0020] Based on the above embodiments, as an optional implementation method S101, calculating the terrain undulation characteristic parameters of the target area according to the digital elevation model, and dividing the target area into multiple grid units according to the terrain undulation characteristic parameters specifically includes S11-S15: S11, Calculate the slope raster map of the target area based on the digital elevation model.

[0021] The system calculates a slope raster map of the target area based on the input Digital Elevation Model (DEM). The DEM is a raster dataset that records surface elevation information, with each cell containing the altitude value corresponding to its geographic location. The slope raster map is a slope distribution map generated by calculating the elevation gradient between each cell and its neighboring cells in the DEM, reflecting the degree of inclination and undulation of the land surface. The system uses the Horn algorithm to calculate slope values. This algorithm calculates the slope of the central cell using a 3×3 neighborhood window, exhibiting good noise resistance and computational stability. Slope values ​​are expressed in degrees, ranging from 0 degrees (completely flat) to 90 degrees (vertical slope), forming a slope raster map consistent with the spatial resolution of the DEM.

[0022] S12, the target area is divided into multiple initial grids according to the preset initial grid size, and the slope standard deviation of the initial grids is calculated according to the slope raster map.

[0023] The system first divides the target area evenly according to a preset initial grid size, forming multiple initial grids. The initial grid size is typically set to a small value, such as 1 km × 1 km, to ensure sufficient spatial resolution for subsequent terrain feature analysis. The system then calculates the standard deviation of the slope values ​​of all pixels within each initial grid based on the slope raster map, serving as a quantitative indicator of the terrain complexity of that grid. The slope standard deviation reflects the degree of variation in terrain undulation within the grid; a larger standard deviation indicates more complex and varied terrain, while a smaller standard deviation indicates flatter and more uniform terrain. During the calculation process, the system excludes invalid data and outliers to ensure the reliability of the statistical results.

[0024] S13, calculate the topographic relief characteristic parameters of the target area based on the slope standard deviation.

[0025] The system calculates the topographic relief characteristic parameters of the target area based on the slope standard deviation distribution characteristics of all initial grids. These parameters are comprehensive indicators describing the spatial distribution of topographic complexity across the entire target area, including statistical features such as the mean, standard deviation, and quantiles of the slope standard deviation. The system determines the grading thresholds for the slope standard deviation through statistical analysis, classifying topographic complexity into three levels: low, medium, and high. Low complexity corresponds to flat areas, medium complexity to hilly areas, and high complexity to mountainous areas. The grading thresholds are determined based on the 33% and 67% quantiles of the slope standard deviation distribution to ensure a relatively balanced area distribution across each level.

[0026] S14. Based on the numerical range of the terrain undulation feature parameters, determine the corresponding grid size, and divide the target area into multiple grid units according to the grid size.

[0027] The system determines the corresponding grid size configuration strategy based on the numerical range of terrain undulation characteristic parameters. For low-complexity terrain areas, the system adopts a larger grid size, such as 5 km × 5 km, because the stereo image screening requirements in flat areas are relatively simple, and a larger grid size can improve processing efficiency. For medium-complexity terrain areas, the system adopts a medium grid size, such as 3 km × 3 km, to balance processing accuracy and computational efficiency. For high-complexity terrain areas, the system adopts a smaller grid size, such as 1 km × 1 km, to ensure that the needs for refined management under complex terrain conditions are met.

[0028] S102. Based on the terrain undulation characteristic parameters, determine the constraint parameter set for each grid cell. The constraint parameter set includes: base-to-height ratio threshold range, angle symmetry threshold, and scoring weight coefficient.

[0029] The system needs to establish differentiated constraint parameter sets based on the terrain undulation characteristics of each grid cell. This is because terrains of varying complexity have significantly different accuracy requirements and fault tolerance for stereo image matching. A uniform parameter configuration cannot simultaneously meet the efficiency needs of flat areas and the accuracy requirements of complex terrain. Terrain complexity directly affects the stability and accuracy of stereo matching. Complex terrain requires stricter geometric constraints to ensure matching quality, while flat areas can appropriately relax constraints to improve the matching success rate. Therefore, a differentiated parameter configuration mechanism with a "one grid, one policy" approach must be established.

[0030] The system first classifies the terrain based on the topographic relief characteristic parameters of each grid cell. Grid cells with a slope standard deviation ≤ 5° are classified as low-complexity terrain types, those with a slope standard deviation between 5° and 12° are classified as medium-complexity terrain types, and those with a slope standard deviation ≥ 12° are classified as high-complexity terrain types. This classification method, based on the objective characteristics of terrain undulation, can effectively distinguish the technical difficulty and accuracy requirements of stereoscopic imaging under different terrain environments.

[0031] For different terrain types, the system configures corresponding baseline-to-elevation ratio (BAR) threshold ranges. The BAR refers to the ratio of baseline length to flight altitude, directly affecting the elevation measurement accuracy and matching stability of stereo image pairs. For low-complexity terrain, where the surface is relatively flat and elevation changes are small, a BAR threshold range of 0.5-0.7 is configured. This smaller BAR range ensures sufficient stereo parallax while avoiding excessive geometric distortion. For medium-complexity terrain, with some terrain undulations, a larger stereo parallax is needed to accurately represent terrain changes; a BAR threshold range of 0.6-0.8 is configured. For high-complexity terrain, with dramatic terrain undulations, the maximum stereo parallax is needed to capture subtle terrain changes; a BAR threshold range of 0.7-1.0 is configured to ensure sufficient elevation sensitivity even under complex terrain conditions.

[0032] Simultaneously, a corresponding angular symmetry threshold is configured. The angular symmetry threshold refers to the allowable range of the difference in the absolute value of the incident angles between the two images in a stereo pair, used to control the geometric symmetry of the stereo pair. For low-complexity terrain, due to relatively regular terrain changes, stereo matching is less sensitive to geometric symmetry; therefore, an angular symmetry threshold of 5° is configured to allow for larger angular deviations to improve the matching success rate. For medium-complexity terrain, terrain changes increase the complexity of matching, requiring better geometric symmetry to ensure matching accuracy; therefore, an angular symmetry threshold of 4° is configured. For high-complexity terrain, drastic terrain changes place extremely high demands on geometric accuracy; therefore, an angular symmetry threshold of 3° is configured to ensure matching reliability under complex terrain through strict symmetry constraints.

[0033] The system also needs to be configured with corresponding scoring weight coefficients, which are used to balance the importance of different evaluation dimensions in the multi-dimensional comprehensive scoring. The scoring dimensions include four aspects: baseline-to-height ratio score, angular symmetry score, image quality score, and temporal phase difference score. For highly complex terrain, geometric accuracy is a key factor for successful matching; therefore, the weights of the baseline-to-height ratio score and angular symmetry score are each set to 0.3 to emphasize the importance of geometric fit, while the weights of the image quality score and temporal phase difference score are each set to 0.2. For low-complexity terrain, due to relatively relaxed geometric constraints, data quality and temporal consistency become more important considerations; therefore, the weights of the image quality score and temporal phase difference score are each increased to 0.3, while the weights of the baseline-to-height ratio score and angular symmetry score are adjusted to 0.2. For moderately complex terrain, a balanced weight configuration is adopted, with each dimension's weight set to 0.25 to ensure a balanced consideration of all indicators.

[0034] By combining the base-to-height ratio threshold range, angular symmetry threshold, and scoring weight coefficient to form a constraint parameter set, the system automatically determines the constraint parameter set corresponding to each grid cell based on its terrain type. This differentiated configuration mechanism significantly improves the overall performance of stereo image matching. In flat areas, appropriately relaxing constraints increases matching efficiency and success rate, while strict constraints ensure matching accuracy and reliability in complex terrain areas. Compared to traditional unified parameter configuration methods, this differentiated constraint mechanism achieves regional optimization of matching efficiency and accuracy, providing a scientific and accurate parameter basis for subsequent image selection and stereo pair determination, effectively solving the parameter adaptation problem faced in stereo image processing under complex terrain environments.

[0035] Based on the above embodiments, as an optional implementation, in S102, determining the constraint parameter set of each grid cell according to the terrain undulation characteristic parameters specifically includes S21-S23: S21. Based on the terrain undulation characteristic parameters of each grid cell, the grid cells are divided into multiple terrain types.

[0036] Based on the terrain undulation characteristic parameters calculated in step S13, the system classifies all grid cells into different terrain types. The terrain type is a classification system based on terrain complexity indicators such as slope standard deviation, used to distinguish different landform features and three-dimensional measurement needs. The system adopts a three-level classification standard, dividing grid cells into three basic types: flat terrain, hilly terrain, and mountainous terrain. Flat terrain corresponds to grid cells with a slope standard deviation less than the low complexity threshold, mainly including plains and plateaus; hilly terrain corresponds to grid cells with a slope standard deviation between the low and high complexity thresholds, mainly including gentle hills and low mountains; mountainous terrain corresponds to grid cells with a slope standard deviation greater than the high complexity threshold, mainly including medium-high mountains and canyons.

[0037] The system also considers the spatial adjacency of grid cells during terrain type classification, avoiding excessive fragmentation of terrain types through spatial smoothing and ensuring that grid cells within the same geomorphic unit have a relatively consistent terrain type affiliation. This approach improves the spatial continuity and practicality of terrain classification.

[0038] S22, for each terrain type, configure the corresponding base-to-height ratio threshold range, the corresponding angle symmetry threshold, and the corresponding scoring weight coefficient, and use the base-to-height ratio threshold range, the angle symmetry threshold, and the scoring weight coefficient as a set of constraint parameters.

[0039] The system configures corresponding combinations of technical parameters for each terrain type, forming a differentiated set of constraint parameters. The set of constraint parameters includes three core components: the base-to-height ratio threshold range, the angular symmetry threshold, and the scoring weight coefficient. These parameters together constitute the technical standard system for stereo image selection under this terrain type.

[0040] The configuration of the basal-to-elevation ratio threshold range is based on the technical requirements for elevation measurement accuracy under different terrain conditions. For flat terrain, the system is configured with a relatively lenient basal-to-elevation ratio threshold range, such as 0.3-0.8, because the elevation changes in flat areas are small, and a moderate basal-to-elevation ratio can meet the measurement accuracy requirements. An excessively large basal-to-elevation ratio may introduce unnecessary geometric deformation. For hilly terrain, the system is configured with a moderate basal-to-elevation ratio threshold range, such as 0.5-1.0, to balance the requirements of elevation sensitivity and geometric stability. For mountainous terrain, the system is configured with a stricter basal-to-elevation ratio threshold range, such as 0.7-1.2, to ensure sufficient elevation resolution under complex terrain conditions.

[0041] The angular symmetry threshold was configured to consider the stability requirements of stereo geometry under different terrain conditions. Flat terrain, due to its gentle surface changes, has a relatively relaxed requirement for angular symmetry, with a threshold set at 15 degrees to allow for a certain degree of angular asymmetry. Hilly terrain has an angular symmetry threshold set at 10 degrees to ensure moderate geometric balance. Mountainous terrain, due to its complex slope aspect changes, has the most stringent requirement for angular symmetry, with a threshold set at 8 degrees to ensure the geometric reliability of stereo observation.

[0042] The weighting coefficients reflect the varying importance of each evaluation dimension under different terrain conditions. For flat terrain, image quality and temporal difference have relatively high weights, set at 0.3 and 0.25 respectively, while basal-to-height ratio and angular symmetry have relatively low weights, set at 0.25 and 0.2 respectively, because flat areas place greater emphasis on data quality consistency. For hilly terrain, the weighting of the four dimensions is relatively balanced: basal-to-height ratio 0.3, angular symmetry 0.25, image quality 0.25, and temporal difference 0.2. For mountainous terrain, basal-to-height ratio and angular symmetry have significantly increased weights, set at 0.35 and 0.3 respectively, while image quality and temporal difference have correspondingly decreased weights, set at 0.2 and 0.15 respectively, highlighting the priority of geometric accuracy.

[0043] S23, determine the set of constraint parameters for each grid cell based on the terrain type to which each grid cell belongs.

[0044] Based on the terrain type classification determined in step S21, the system assigns a corresponding set of constraint parameters to each grid cell. The system establishes a mapping relationship between grid cells and constraint parameter sets, ensuring that each grid cell receives technical parameter configurations that match its terrain characteristics. During the mapping process, the system verifies the completeness and consistency of the parameter configurations, ensuring that all necessary technical parameters are correctly set and the sum of the weighting coefficients equals 1.0.

[0045] The system also records the configuration basis and version information of the constraint parameter set for each grid cell, facilitating subsequent parameter tuning and performance evaluation. When the screening effect of some grid cells is found to be unsatisfactory, their constraint parameter sets can be fine-tuned according to the specific situation to achieve continuous optimization of parameter configuration.

[0046] S103: For each grid cell, perform multi-dimensional filtering sequentially from the image database according to the task parameters to generate a candidate image set.

[0047] The system needs to perform multi-dimensional filtering from the image database for each grid cell to generate a candidate image set. This is because the image database stores a large number of remote sensing images acquired at different times, from different satellites, and under different imaging conditions. These images vary greatly in terms of temporal phase, resolution, coverage, and geometric features. Directly using all images for stereo matching would lead to a sharp increase in computational load and make it difficult to guarantee matching quality. The multi-dimensional filtering mechanism, by setting progressive filtering conditions, can systematically eliminate images that do not meet the task requirements, ensuring that high-quality, highly adaptable image data is used in subsequent stereo matching processes, thereby improving matching efficiency and success rate.

[0048] The image database is the core data resource of the system, storing complete metadata information for historical images from multiple satellite sources, including key parameters such as satellite model, imaging time, orbital altitude, side tilt angle, angle of incidence, resolution, cloud cover, and geographic coordinate range. The angle of incidence refers to the angle between the direction of light incident on the satellite and the normal to the Earth's surface during imaging; it is marked with a sign, with positive values ​​indicating east-view imaging and negative values ​​indicating west-view imaging. This parameter directly determines the observational perspective characteristics of the image. The system establishes a composite index structure based on multiple dimensions such as time, space, and resolution, supporting efficient multi-condition queries and data retrieval.

[0049] The system first retrieves all images whose spatial coverage overlaps with the current grid cell from the image database, forming an initial image set. Next, it performs a first round of temporal screening based on a temporal threshold specified in the task parameters. This threshold defines the allowed time window range for the task; by selecting images whose imaging time falls within this window, the system ensures the timeliness and seasonal consistency of the images. The purpose of temporal screening is to avoid the impact of changes in ground features due to excessively long time spans on stereo matching results, especially in vegetated areas where images from different seasons may exhibit significant spectral and textural differences. After temporal screening, a filtered image set is obtained, which meets the basic requirements of the task in the temporal dimension.

[0050] The system then performs a second round of resolution filtering on the temporally filtered images based on the resolution threshold specified in the task parameters. The resolution threshold represents the minimum spatial resolution requirement for the images, typically set to 0.5 meters or higher to ensure sufficient spatial detail for high-precision stereo measurement. This resolution filtering eliminates coarse-resolution images that do not meet the spatial resolution requirements, preventing a decrease in matching accuracy due to resolution differences. The resulting image set meets the task's accuracy requirements in terms of spatial resolution.

[0051] The system further performs a third round of coverage filtering on the resolution-filtered images based on the precise coverage range of the current grid cells. Coverage filtering calculates the spatial overlap rate between each image and the grid cell, retaining only images with an overlap rate of 60% or higher to ensure that the images provide sufficient spatial information for the grid cell. The overlap rate threshold is set based on the spatial continuity requirements of stereo matching; too low an overlap rate will lead to discontinuities in the matching area, affecting the integrity of the stereo model. After coverage filtering, a set of images is obtained, which meets the specific requirements of the grid cells in terms of spatial coverage.

[0052] The system then performs a fourth round of side-swing angle filtering on the images after coverage filtering, based on the side-swing angle range specified in the mission parameters. The side-swing angle is the deflection angle relative to the vertical downward-looking direction during satellite imaging, directly affecting the degree of geometric distortion and the magnitude of stereo parallax. Mission parameters typically limit the side-swing angle to the range of 10°-18°. Side-swing angles less than 10° result in insufficient stereo parallax, affecting the accuracy of elevation measurements, while side-swing angles greater than 18° introduce excessive geometric distortion, reducing image quality. Side-swing angle filtering ensures that the images have suitable geometric features, providing a good geometric foundation for subsequent stereo matching. The resulting image set after side-swing angle filtering is obtained.

[0053] Finally, the system calculates the incident angle of each image based on the imaging geometry parameters of the images after side-swing angle filtering, and performs a fifth round of incident angle filtering. The calculation of the incident angle is based on the geometric relationship between satellite orbit parameters, imaging time, and target location, and the incident angle of light is determined through the principles of spherical geometry. Incident angle filtering ensures that the filtered images have a suitable distribution of observation angles, creating conditions for forming effective stereo image pairs, according to the complementary viewing angle requirements of stereo imaging. Incident angle filtering considers the east-west side-view angle distribution, prioritizing the retention of image combinations that can form complementary viewing angles. After incident angle filtering, a set of images after incident angle filtering is obtained, which is the candidate image set for the current grid cell.

[0054] The implementation of this five-level progressive multi-dimensional screening mechanism has yielded significant results. By progressively narrowing the image search space, it precisely reduces the massive amount of images in the original image database into a high-quality candidate image set. Compared to traditional single-condition screening or simple combination screening, this method achieves comprehensive optimization across multiple dimensions, including time, space, geometry, and quality, ensuring that each image in the candidate image set meets the task requirements in each dimension. The progressive design in the screening process avoids redundant calculations, improving screening efficiency, while a reasonable screening order ensures the quality and suitability of the final candidate set. This candidate image set provides a high-quality data foundation for subsequent stereo image pair determination and comprehensive scoring, significantly improving the execution efficiency and success rate of the entire multi-angle remote sensing image screening and acquisition method.

[0055] Based on the above embodiments, as an optional implementation, in S103, for each grid cell, multi-dimensional filtering is sequentially performed from the image database according to the task parameters to generate a candidate image set, specifically including S31-S35: S31: Obtain images covering grid cells from the image database, perform temporal phase filtering based on the temporal phase threshold in the task parameters, and obtain the temporal phase filtered images.

[0056] The system first retrieves all image data whose spatial extent overlaps with the current grid cell from the image database. The determination of spatial coverage is based on spatial intersection analysis of the geographic boundary coordinates of the image and the boundary coordinates of the grid cell. The system employs efficient spatial indexing technology to quickly locate relevant images. After acquiring the covered images, the system performs temporal filtering based on preset temporal thresholds in the task parameters. The temporal threshold is a crucial parameter limiting the time range for image acquisition; it is typically set to a specific time window, such as the last three years, to ensure that the selected images have temporal relevance and timeliness.

[0057] During the temporal phase filtering process, the system calculates the time difference between the acquisition time of each image and the task baseline time, excluding outdated images that exceed the temporal phase threshold. This temporal phase control mechanism ensures the timeliness of subsequent stereo modeling results and avoids stereo matching errors caused by surface changes. For special tasks requiring multi-temporal phase analysis, the system also supports setting multiple temporal phase windows to acquire image data from different time periods.

[0058] S32, perform resolution filtering on the temporally filtered images according to the resolution threshold in the task parameters, and obtain the resolution-filtered images.

[0059] The system performs resolution filtering on the temporally selected images based on the resolution threshold specified in the task parameters. The resolution threshold defines the minimum spatial resolution required for the task and is typically determined based on the expected mapping scale and accuracy requirements. The system obtains the ground resolution parameters of each image by reading its metadata, excluding images with resolutions below the threshold to ensure that candidate images possess sufficient spatial detail representation. Resolution filtering considers not only the original resolution of the image but also the effective resolution changes caused by variations in the observation angle; for large-angle side-view images, corresponding resolution corrections are performed.

[0060] S33, based on the coverage range of the grid cells, perform coverage range filtering on the resolution-filtered images to obtain the coverage range-filtered images.

[0061] The system performs coverage filtering on the resolution-filtered images based on the precise coverage area of ​​the grid cells. Although preliminary spatial filtering was performed in step S31, this step conducts a more precise coverage analysis, calculating the actual coverage ratio of each image to the grid cell. The system excludes images with excessively low coverage ratios, typically requiring an image coverage ratio of at least 80% for each grid cell to ensure that candidate images can provide sufficient data support for that grid cell. Coverage filtering also considers the data quality issues at image edges, appropriately narrowing the effective coverage area of ​​the image and avoiding the use of areas with poor image edge quality.

[0062] S34, perform side-swing angle filtering on the images after coverage filtering according to the side-swing angle range in the task parameters, and obtain the images after side-swing angle filtering.

[0063] The system performs side-swing angle filtering on images after coverage screening based on the side-swing angle range specified in the mission parameters. The side-swing angle is the deflection angle of the satellite sensor relative to the vertical downward direction, directly affecting the observation geometry and stereoscopic effect of the image. The side-swing angle range is typically set as a symmetrical interval, such as -30 degrees to +30 degrees, to ensure the image has a suitable observation angle. The system obtains the side-swing angle information by reading the image's observation geometry parameters and excludes images with side-swing angles exceeding the allowable range.

[0064] A crucial aspect of lateral tilt angle screening is ensuring that the candidate image set includes images with different lateral tilt angle symbols, meaning both left-side and right-side views are present. This provides the foundation for subsequent stereo image pair formation. During the screening process, the system analyzes the distribution of images across different lateral tilt angle ranges to ensure a balanced angle distribution.

[0065] S35, calculate the incident angle based on the imaging geometry parameters of the image after side-swing angle screening, perform incident angle screening on the image after side-swing angle screening, obtain the image after incident angle screening, and use the image after incident angle screening as the candidate image set of grid unit.

[0066] The system calculates the incident angle based on the imaging geometry parameters of the image after side-swing angle filtering, and then performs incident angle filtering. The incident angle is the angle between the sunlight and the surface normal vector, reflecting the illumination conditions at the time of imaging. The calculation of the incident angle requires comprehensive consideration of multiple geometric parameters such as solar altitude angle, solar azimuth angle, surface slope, and aspect. The system uses a precise astronomical calculation model to determine the sun's position at the time of imaging, and combines this with a digital elevation model to calculate the surface geometry parameters of the grid cells, thus obtaining an accurate incident angle value.

[0067] The purpose of incident angle screening is to eliminate images with extremely unfavorable lighting conditions, such as shadows caused by excessively large incident angles or overexposure caused by excessively small incident angles. The system typically sets an allowable range for the incident angle, such as 20 to 70 degrees, within which images have better lighting conditions and image quality. After incident angle screening, the remaining images constitute the candidate image set for that grid cell, providing a data basis for subsequent stereo pair formation and quality assessment.

[0068] S104. Based on the constraint parameter set, determine whether each image pair in the candidate image set satisfies the stereo image pair condition, and include the image pairs that satisfy the stereo image pair condition into the effective image pair set.

[0069] The system needs to determine whether each image pair in the candidate image set meets the stereo pair conditions based on the constraint parameter set. This is because although the images in the candidate image set meet the basic requirements in individual dimensions, whether they can form an effective stereo pair after being combined in pairs still requires rigorous geometric compatibility verification. The stereo pair conditions are key constraints to ensure that two images can perform high-quality stereo matching and 3D reconstruction. Only image pairs that simultaneously meet the requirements of basal-to-height ratio, angular symmetry, and viewpoint complementarity can generate stable and reliable stereo parallax information. Relying solely on the individual quality of the images cannot guarantee the stereo imaging effect after pairing; a systematic determination of stereo pair conditions is necessary to screen out image combinations that truly possess stereo measurement capabilities.

[0070] Stereo image pair conditions refer to the set of geometric and spatiotemporal constraints that must be met for effective stereo measurement after combining two remote sensing images. These constraints mainly include three core elements: baseline-to-height ratio compliance, angular symmetry compliance, and opposite sign of the viewing angle. The baseline-to-height ratio compliance requirement ensures that the baseline-to-height ratio of the image pair falls within the baseline-to-height ratio threshold range defined in the constraint parameter set, ensuring sufficient elevation sensitivity while avoiding excessive geometric distortion. The angular symmetry compliance requirement ensures that the absolute difference in the incident angles of the two images does not exceed the angular symmetry threshold specified in the constraint parameter set, ensuring the geometric symmetry and matching stability of the stereo image pair. The opposite sign of the viewing angle requirement requires that the incident angles of the two images have opposite signs, i.e., one image is an east-view image and the other is a west-view image, forming a true stereo observation geometry.

[0071] The system first pairs the images in the candidate image set, generating all possible image pair combinations. For a candidate set containing n images, theoretically, C(n, 2) = n(n-1) / 2 distinct image pairs can be formed. The system employs an efficient pairing algorithm to avoid redundant calculations and invalid combinations, ensuring that every possible image pair is evaluated. Next, for each image pair, the system first verifies the opposite sign condition by checking whether the incident angles of the two images have opposite signs to determine if the image pair possesses basic stereo geometry. The sign information of the incident angles is already marked in the image metadata; positive values ​​indicate eastward-view imaging, and negative values ​​indicate westward-view imaging. Only image pairs with opposite signs can form valid stereo intersection geometry.

[0072] For image pairs that meet the condition of different viewing angles, the system further calculates their base-to-height ratio (ATR). The ART is calculated based on the imaging geometry parameters of the two images, including satellite orbital altitude, imaging position coordinates, and baseline length. The baseline length is the projected length of the spatial distance between the satellite positions during the two imaging events on the ground, and the ART is equal to the ratio of the baseline length to the average flight altitude. The system obtains the ART value for each image pair through precise geometric calculations and then compares it with the ART threshold range in the constraint parameter set of that grid cell. Only image pairs whose ART values ​​fall within the threshold range are considered to meet the ART criterion. An ART that is too small will lead to insufficient stereo parallax, affecting the accuracy of elevation measurement, while an ART that is too large will introduce excessive distortion, affecting the reliability of matching.

[0073] The system simultaneously calculates the angular symmetry index for each image pair, which is the difference in the absolute values ​​of the incident angles of the two images. Angular symmetry reflects the geometric balance of the stereo image pair; an ideal stereo image pair should have a geometric configuration with similar absolute values ​​of incident angles but opposite signs. The system compares the calculated angular symmetry values ​​with an angular symmetry threshold in the constraint parameter set. Only image pairs with an angular difference less than or equal to the threshold are considered to meet the angular symmetry requirement. Good angular symmetry ensures that the two images have similar geometric distortion characteristics during stereo matching, reducing matching errors caused by viewpoint asymmetry.

[0074] When an image pair simultaneously meets the conditions of opposite viewing angle, stoichiometric ratio, and angular symmetry, the system determines that the image pair meets the stereo pair criteria and includes it in the valid image pair set of the current grid cell. The valid image pair set refers to the set of all image pairs within the current grid cell that meet the stereo pair criteria; these image pairs possess the basic geometric conditions for high-quality stereo measurement. The system generates detailed pairing information records for each image pair in the valid image pair set, including key parameters such as the identifiers of the two images, stoichiometric ratio values, angular symmetry values, and incident angle information, providing data support for subsequent comprehensive scoring and optimization.

[0075] For image pairs that do not meet the stereo pair criteria, the system excludes them from the valid image pair set and records the specific reasons for exclusion, such as a base-to-height ratio exceeding the range, angular symmetry not meeting the standard, or identical viewing angles. This detailed screening record helps analyze the quality distribution and geometric characteristics of image resources, providing reference information for image database optimization and re-acquisition planning.

[0076] The implementation of the stereo image pair condition determination mechanism significantly improves the accuracy of image selection and the success rate of subsequent stereo processing. Through rigorous geometric constraint verification, the system ensures that each image pair in the valid image pair set possesses reliable stereo measurement capabilities, avoiding stereo processing failures caused by geometric mismatches. Compared to traditional simple pairing methods, this mechanism achieves accurate evaluation of image pair quality through comprehensive judgment of multiple conditions, providing a reliable data foundation for high-precision 3D modeling. The systematic condition determination process also improves processing efficiency by early exclusion of unqualified image pairs, reducing the burden of subsequent complex calculations and achieving a dual optimization of image selection efficiency and quality.

[0077] S105. Based on the scoring weight coefficient of each grid cell, calculate the comprehensive score of each image pair in the effective image pair set of each grid cell, select the image pair with the highest comprehensive score that is greater than the preset score in each grid cell as the matching result, generate a unique identifier code for the corresponding matching result, and output the stereo image pair dataset based on the unique identifier code.

[0078] The system needs to calculate the comprehensive score of each image pair in the effective image pair set based on the scoring weight coefficient of each grid cell and select the optimal matching result. This is because although all image pairs in the effective image pair set meet the basic stereo image pair conditions, they still have significant differences in geometric accuracy, image quality, and temporal consistency. A multi-dimensional comprehensive evaluation system is needed to quantify the overall suitability of each image pair. Stereo measurements exhibit different sensitivities to various indicators under different terrain conditions. A single evaluation standard cannot adapt to complex and diverse terrain environments. It is necessary to combine the terrain characteristics of the grid cells with differentiated weight configurations to ensure that the most suitable image pair can be selected under different terrain conditions, thereby maximizing the accuracy and reliability of stereo modeling.

[0079] The comprehensive score is a weighted composite index based on multiple evaluation dimensions, used to quantify the overall adaptability and expected performance of image pairs under the current grid cell conditions. The scoring system includes four core dimensions: basal-to-height ratio score reflects the elevation measurement sensitivity of the image pair; angular symmetry score reflects the geometric balance of the image pair; image quality score reflects the image sharpness and signal-to-noise ratio; and temporal phase difference score reflects the temporal consistency of the two images. The individual scores for each dimension are converted into standardized values ​​from 0 to 100 using a standardization algorithm, facilitating comparison and synthesis between different dimensions.

[0080] The system first calculates individual scores across four dimensions for each image pair in the effective image pair set of each grid cell. The basal-to-height ratio score is calculated based on the degree of matching between the actual basal-to-height ratio of the image pair and the basal-to-height ratio threshold range in the constraint parameter set of that grid cell. A perfect score of 100 is awarded when the basal-to-height ratio is at the center of the threshold range. The score gradually decreases as the deviation from the center increases. A Gaussian distribution model is used in the scoring function to ensure a smooth scoring transition. The angular symmetry score is calculated based on the ratio of the absolute difference in the incident angles of two images to the angular symmetry threshold. Perfectly symmetrical image pairs receive a perfect score. The score decreases linearly as the degree of asymmetry increases. A lower limit of 60 points is set to ensure basic usability.

[0081] Image quality scoring comprehensively considers factors such as signal-to-noise ratio (SNR), contrast, sharpness, and cloud cover, and is calculated using quality parameters recorded in the image metadata. SNR reflects the noise level of the image, contrast reflects the dynamic range, sharpness reflects the degree of spatial detail retention, and cloud cover reflects the effective observation area. The system uses a weighted average method to synthesize these quality indicators into a unified image quality score, with SNR and sharpness having higher weights, reflecting the dependence of stereo matching on image detail information. Temporal difference scoring is calculated based on the imaging time interval between two images; a smaller time interval indicates less change in ground features and higher reliability of stereo matching. The system sets the ideal temporal difference to be within 7 days; scores gradually decrease beyond this range, with the maximum permissible temporal difference being the temporal threshold specified in the task parameters.

[0082] After calculating the individual scores for the four dimensions, the system performs a weighted sum based on the score weight coefficients of the current grid cell to generate a comprehensive score for each image pair. The score weight coefficients were differentiated in step S102 according to the terrain type of the grid cell. For high-complexity terrain, geometric accuracy is emphasized, therefore the baseline-to-height ratio score and angular symmetry score have higher weights. For low-complexity terrain, data quality is emphasized, therefore the image quality score and temporal phase difference score have higher weights. The comprehensive score is calculated as follows: Comprehensive Score = Baseline-to-Height Ratio Score × Baseline-to-Height Ratio Weight + Angular Symmetry Score × Angular Symmetry Weight + Image Quality Score × Image Quality Weight + Temporal Phase Difference Score × Temporal Phase Difference Weight, ensuring that the sum of all weights equals 1.0.

[0083] After calculating the comprehensive score of all image pairs in the current grid cell, the system makes an optimal selection decision. First, image pairs with a comprehensive score greater than a preset score are selected as candidate matching results. The preset score is typically set at 75 points to ensure that the selected image pairs achieve a good level across all dimensions. The preset score is based on statistical analysis of a large amount of experimental data; the 75-point threshold effectively eliminates image pairs that do not meet the quality requirements while ensuring a sufficient number of candidates to support subsequent selection. Among the candidates that meet the preset score condition, the system selects the image pair with the highest comprehensive score as the final matching result for that grid cell, ensuring that the optimal configuration is selected while meeting basic quality requirements.

[0084] For each confirmed match, the system generates a corresponding unique identifier, which uses a structured encoding method to contain rich metadata. The unique identifier is formatted as "SPID_Grid Number_Image 1 Satellite Code_Image 1 Acquisition Date_Image 2 Satellite Code_Image 2 Acquisition Date_Basic-to-Height Ratio Category", where SPID represents the stereo image pair identifier prefix, the grid number identifies the grid cell to which the image pair belongs, the satellite code indicates the satellite platform from which the image was acquired, the acquisition date uses the YYYYMMDD format to represent the image imaging time, and the basic-to-height ratio category uses the letters L, M, and H to represent low, medium, and high basic-to-height ratio levels, respectively. This structured unique identifier not only uniquely identifies each stereo image pair but also facilitates subsequent data management, querying, and traceability.

[0085] The system constructs a complete stereo image pair dataset based on the matching results of all grid cells and their corresponding unique identifiers. This dataset contains an optimized set of stereo images covering the entire target area. Each pair has undergone rigorous multi-dimensional evaluation and optimization to ensure the overall quality and consistency of the dataset. The dataset is output in a structured format, containing detailed metadata information for each stereo image pair, such as grid cell location, image identifier, imaging parameters, scoring information, and geometric parameters, providing comprehensive data support for subsequent 3D modeling and measurement applications.

[0086] Based on the above embodiments, as an optional implementation, in S105, calculating the comprehensive score of each image pair in the effective image pair set of each grid unit according to the scoring weight coefficient of each grid unit specifically includes S51-S52: S51, for each image pair in the effective image pair set of each grid cell, calculate the datum-to-height ratio score, angular symmetry score, image quality score, and temporal phase difference score of the image pair.

[0087] The system calculates individual scores for each image pair in the effective image pair set of each grid cell, focusing on four core dimensions. The baseline-to-height ratio (BHR) score is calculated based on the stereo geometric configuration quality of the image pair. The system first calculates the BHR value (the ratio of baseline length to orbital height) based on the orbital parameters and observation geometry of the two images, then maps the BHR value to a scoring range of 0 to 1. The BHR score uses a trapezoidal function model, awarding a full score of 1.0 within the optimal BHR range, linearly decreasing within the acceptable range, and awarding 0 points to values ​​outside the acceptable range. The optimal BHR range is typically set between 0.6 and 1.0, within which stereo image pairs possess sufficient elevation sensitivity while maintaining good matching stability.

[0088] Angular symmetry scoring reflects the degree of symmetrical balance between the observation angles of two images. The system assesses symmetry quality by calculating the angular difference between the side-swing angles of the two images. An ideal stereo image pair should have side-swing angles of equal magnitude and opposite signs, forming a perfectly symmetrical observation geometry. The angular symmetry scoring also uses a trapezoidal function model, assigning the highest score when the angular difference is close to 0 degrees, and gradually decreasing the score as the angular difference increases. The system sets an angular difference of less than 5 degrees as excellent, 5-15 degrees as good, and more than 20 degrees as poor symmetry.

[0089] Image quality scoring comprehensively considers factors such as radiometric quality, geometric quality, and cloud cover. The system calculates a comprehensive evaluation value of image quality by reading quality identifier parameters from image metadata and combining them with quality level information obtained through statistical analysis. The calculation of image quality scoring also considers technical indicators such as signal-to-noise ratio, contrast, and sharpness, as well as the impact of cloud cover ratio on data usability. High-quality images exhibit excellent performance in texture sharpness, radiometric consistency, and atmospheric conditions, providing a reliable basis for stereo matching.

[0090] Temporal phase difference scoring measures the proximity of the acquisition times of two images, reflecting the potential impact of surface changes on the accuracy of stereo modeling. The system calculates the acquisition time interval between the two images in an image pair and converts the time difference into a normalized score. Temporal phase difference scoring uses a negative exponential decay model, assigning the highest score of 1.0 to image pairs acquired on the same day, with lower scores for larger time intervals. Typically, image pairs with time intervals within one month still receive higher scores, while image pairs with time intervals exceeding one year receive significantly lower scores, reflecting the degree of impact of seasonal changes and human construction on the surface.

[0091] S52, based on the scoring weight coefficients of the grid cells, the base-to-height ratio score, angular symmetry score, image quality score, and temporal phase difference score are weighted and summed to generate a comprehensive score for each image pair in the effective image pair set of each grid cell.

[0092] The system calculates a weighted sum of the individual scores across the four dimensions based on the scoring weight coefficients determined for each grid cell in step S22, generating a comprehensive score. The scoring weight coefficients reflect the relative importance of each evaluation dimension under different terrain conditions, and the weighted summation mechanism ensures that the comprehensive score accurately reflects the overall applicability of the imagery to specific terrain conditions.

[0093] The comprehensive score is calculated as follows: Comprehensive score = Base-to-height ratio score multiplied by base-to-height ratio weight + Angular symmetry score multiplied by angular symmetry weight + Image quality score multiplied by image quality weight + Temporal phase difference score multiplied by temporal phase difference weight. During the calculation process, the system verifies that the sum of the weighting coefficients is 1.0, ensuring that the comprehensive score remains between 0 and 1, facilitating comparison and ranking between different image pairs.

[0094] The weighted summation process fully reflects the terrain-oriented parameter configuration strategy. In flat terrain areas, the weights of image quality and temporal difference are relatively high, and the overall score tends to select image pairs with excellent quality and similar temporal phases, because the requirements for geometric accuracy are relatively relaxed in flat areas. In hilly terrain areas, the weights of the four dimensions are relatively balanced, and the overall score reflects a balanced consideration of performance in all aspects. In mountainous terrain areas, the weights of basal-to-height ratio and angular symmetry are significantly increased, and the overall score prioritizes the superiority of geometric configuration to ensure the accuracy of elevation measurements under complex terrain conditions.

[0095] The system also records detailed scoring for each image pair, including individual scores for each dimension and a final comprehensive score, facilitating subsequent result analysis and parameter optimization. When the selection effect of image pairs in certain grid units is found to be unsatisfactory, the weak points can be identified by analyzing the scoring details, and the weight configuration or individual score calculation method can be adjusted in a targeted manner.

[0096] Based on the above embodiments, as an optional implementation, after calculating the comprehensive score of each image pair in the effective image pair set of each grid cell according to the scoring weight coefficient of each grid cell, the implementation further includes steps S201-S207: S201, Obtain the image pairs in the effective image pair set whose comprehensive score is not greater than a preset score and extract the missing image features of the image pairs to be supplemented.

[0097] The system first identifies and acquires all grid cells that fail to meet the comprehensive score requirement and require supplementary acquisition. These grid cells are those whose comprehensive score for each image pair in their effective image pair set is no greater than a preset score. These grid cells represent weak links in the existing image resources and require new image acquisition to improve data quality. The system performs in-depth missing image analysis on each grid cell, extracting its missing image features. Missing image features refer to the image types and characteristic parameters lacking for forming high-quality stereo image pairs, including the missing observation angle range, the required incident angle sign, the ideal datum-to-height ratio configuration, the desired image quality level, and the appropriate imaging time window, among other key information.

[0098] The system analyzes the distribution characteristics of existing candidate images within the grid cell to identify the main reasons for substandard overall scores. If the existing images are primarily east-view images, the missing image features will indicate the need for west-view images to create complementary perspectives; if the incident angle distribution of the existing images is uneven, the ideal incident angle range will be indicated; if the quality of the existing images is generally low, the minimum required image quality level will be indicated. Missing image features also include the grid cell's geographic coordinate range, terrain complexity level, priority weight, and other information, providing a comprehensive description of requirements for subsequent data acquisition planning.

[0099] S202, Generate a list of supplementary acquisition requirements based on the missing image features of all grid cells to be supplemented.

[0100] The system summarizes and analyzes the missing image features of all grid cells to be acquired, generating a unified list of acquisition needs. This list is a collection of acquisition requirements formed after systematically organizing and optimizing all missing image features. It is categorized and organized according to dimensions such as geographical location, time requirements, and observation angle, facilitating subsequent acquisition planning and resource scheduling. When generating the acquisition need list, the system merges and optimizes the needs. For grid cells with similar geographical locations and missing features, the system merges them into a single acquisition need to improve acquisition efficiency. For grid cells with urgent or important needs, the system marks them with high priority in the need list to ensure they are met first.

[0101] The supplementary acquisition request list records detailed parameters for each request in a structured format, including target area coordinates, required observation angle, incident angle sign requirements, target base-to-height ratio, image quality level, time window limits, and priority level. The system also calculates the technical difficulty and resource consumption of each acquisition request, providing a quantitative basis for acquisition planning.

[0102] S203: Obtain the transit parameters of satellite resources within the mission time window, combine the supplementary acquisition requirement list and transit parameters, generate at least one acquisition plan for each grid cell to be supplemented, and arrange the acquisition plans in chronological order to form an acquisition sequence.

[0103] The system acquires the transit parameters of currently available satellite resources within the mission time window, providing a technical feasibility basis for fulfilling supplementary acquisition needs. These transit parameters include key data such as orbital predictions, transit schedules, observation capability parameters, and resource occupancy for each satellite. Combining the supplementary acquisition needs list and transit parameters, the system uses optimization algorithms to generate at least one executable acquisition plan for each grid cell to be acquired. The acquisition plan is a specific acquisition scheme for a particular grid cell, detailing technical elements such as the satellite platform, acquisition time, observation angle, and imaging parameters.

[0104] When generating acquisition plans, the system comprehensively considers multiple factors, including satellite resource availability, observation geometry adaptability, time window constraints, and cost-effectiveness optimization. For the same grid cell to be acquired, the system typically generates multiple alternative acquisition plans to address uncertainties such as satellite malfunctions, weather changes, and resource conflicts. After generation, the acquisition plans are arranged in chronological order to form an acquisition sequence, ensuring the orderly execution of acquisition tasks and the rational allocation of resources.

[0105] S204, perform image acquisition sequentially according to the acquisition sequence, and pair each newly acquired image with images in the candidate image set.

[0106] The system executes image acquisition tasks sequentially according to the acquisition sequence, obtaining new image data for the grid cells to be acquired via the satellite platform. Each time a new image is successfully acquired, the system immediately adds it to the image database and updates the corresponding metadata information, including parameters such as imaging time, observation angle, and image quality. The system then pairs the new image with existing images in the candidate image set for that grid cell, generating image pairs that include the new image, thus preparing data for stereo image pair condition determination and comprehensive scoring calculation.

[0107] The pairing process between new and existing images follows a systematic set of combination rules. New images are tested against all images in the candidate image set that meet the basic pairing criteria, ensuring that the pairing potential of the new images is fully explored. The system pays particular attention to the complementarity of the new and existing images in terms of observation angles, prioritizing pairings that can form good stereo geometry.

[0108] S205, determine whether the newly paired image pair meets the stereo image pair condition based on the constraint parameter set, and calculate the comprehensive score of the newly paired image pair based on the scoring weight coefficient.

[0109] The system performs stereo pair condition determination on all image pairs containing new images, verifying whether they meet the conditions of different visual signs, basal-to-height ratio, and angular symmetry. The determination process uses the same set of constraint parameters and determination criteria as the previous steps to ensure the consistency and comparability of the evaluation system. For newly paired image pairs that meet the stereo pair conditions, the system further calculates their comprehensive score based on the scoring weight coefficient of that grid cell. The score calculation covers four dimensions: basal-to-height ratio score, angular symmetry score, image quality score, and temporal phase difference score. The same scoring algorithm and weight configuration are used to ensure the fairness and accuracy of the scoring results.

[0110] S206, when the overall score of the newly paired image pair is greater than the preset score, confirm that the supplementary acquisition task of the corresponding grid unit to be supplemented is completed, and generate a unique identifier for the newly paired image pair.

[0111] When the overall score of the newly paired image pair exceeds a preset score, the system confirms that the re-acquisition task for the grid cell to be re-acquisitioned has been successfully completed. At this point, the grid cell has obtained a stereo image pair that meets the quality requirements and can support subsequent high-precision 3D modeling tasks. The system generates a unique identifier for this successful new paired image pair, using the same coding rules as the main workflow to ensure seamless integration into the overall stereo image pair dataset. After the re-acquisition task is completed, the system updates the status flag of the grid cell, removes it from the list of cells to be re-acquisitioned, and records detailed information about the re-acquisition process for quality traceability and effect evaluation.

[0112] S207: When the overall score of the newly paired image pair is not greater than the preset score, execute the next acquisition plan for the grid cell to be acquired until the acquisition task is completed or the acquisition sequence is completed.

[0113] If the overall score of the newly paired image pair is still less than the preset score, it indicates that the newly acquired images are insufficient to resolve the data quality issue of that grid cell, and subsequent acquisition plans need to be executed. The system will initiate the next acquisition plan for the grid cell to be supplemented, continuing to acquire more new image data and gradually improving the image resource configuration of the area through multiple acquisitions. This iterative acquisition process will continue until the grid cell successfully obtains a high-quality stereo image pair to complete the supplementary acquisition task, or all acquisition plans for that grid cell have been executed.

[0114] When the acquisition sequence is completed but the quality requirements are not met, the system will record that the grid cell is a particularly difficult area and analyze the technical reasons for the failure of the supplementary acquisition, such as overly complex terrain, harsh weather conditions, and limited satellite resources, to provide reference information for subsequent technical improvements and resource allocation optimization.

[0115] Based on the above embodiments, as an optional implementation, after outputting the stereo image pair dataset according to the unique identifier, the method further includes: When the image database receives a new image, it pairs and scores the new image with existing images. When the overall score of the newly paired image pair is higher than the overall score of the image pair used by the current grid cell, the recommended image pair for that grid cell is updated and the unique identifier is updated.

[0116] After outputting the stereo image pair dataset, the system needs to establish a dynamic update mechanism to cope with the continuous expansion of the image database. This is because the satellite image database is a dynamically growing data resource; new image data is constantly being acquired and added. These newly added images may have better observation conditions, higher data quality, or more suitable geometric configurations, providing better data support for existing stereo modeling tasks. If the system lacks dynamic update capabilities, it cannot fully utilize the advantages of the newly added data, causing the recommended stereo image pairs to remain at the historical best level, failing to achieve continuous optimization of data resource allocation. Establishing an automatic update mechanism based on comprehensive score comparison can ensure that the recommended stereo image pairs always remain in the latest and optimal state.

[0117] When the image database receives new imagery, the system first performs comprehensive preprocessing and quality checks to ensure the integrity of the new imagery's metadata and the data quality meet system requirements. The system then determines the grid cells that the new imagery may affect based on its spatial coverage, and quickly locates all grid cells with coverage relationships to the new imagery using spatial indexing technology. For each relevant grid cell, the system performs pairing analysis between the new imagery and all existing images in the candidate image set for that grid cell, forming a series of potential stereo image pairs containing the new imagery.

[0118] The new image pairing process follows the same technical standards and constraints as the initial screening. The system first verifies the validity of newly paired image pairs based on the constraint parameter set of the grid cell, ensuring that key geometric parameters such as basal-to-elevation ratio and angular symmetry meet the terrain type requirements of that grid cell. Only newly paired image pairs that pass the validity verification will proceed to the subsequent scoring calculation process, thus avoiding invalid pairings consuming computational resources.

[0119] For newly paired image pairs that pass the validity verification, the system calculates a comprehensive score according to the scoring mechanism established in steps S51 and S52. The system calculates the datum-to-height ratio score, angular symmetry score, image quality score, and temporal phase difference score for the new image pair, and then performs a weighted sum based on the scoring weight coefficient of the grid cell to obtain the comprehensive score value for the new image pair. The scoring calculation process uses the same algorithm and parameters as existing image pairs to ensure the comparability and fairness of the scoring results.

[0120] The system then compares the overall score of the newly paired image pair with the overall score of the currently recommended image pair used in that grid cell. When the overall score of the newly paired image pair is higher than that of the currently recommended image pair, it indicates that the new data combination has a significant advantage in overall performance, meeting the update trigger condition. The system further verifies whether the score improvement reaches a preset update threshold, typically set at a score difference of 0.05, to avoid frequent updates due to small score fluctuations.

[0121] When a new image pair is confirmed to have a significant performance advantage, the system updates the recommended image pair. The update process includes replacing the recommended image pair record for that grid cell with the new image pair information, updating relevant scoring data and quality indicators, and retaining the update history for traceability and analysis. The system also recalculates the unique identifier for that grid cell. This unique identifier is a hash value generated based on the key attribute information of the recommended image pair and is used to uniquely identify a specific stereo image pair configuration.

[0122] The updating of unique identifiers ensures version management and consistency control of the stereo image pair dataset. The new unique identifier reflects the updated image pair configuration, enabling downstream application systems to promptly detect data changes and take appropriate action. When updating the unique identifier, the system also updates the timestamp and version number information, establishing a complete change tracking mechanism.

[0123] To ensure the stability and reliability of the update process, the system has also established a comprehensive exception handling and rollback mechanism. When data anomalies or system errors occur during the update process, the system can automatically roll back to the stable state before the update, avoiding data corruption or service interruption caused by update failure. The system also records detailed update operations, including key information such as update trigger conditions, scoring comparison results, and update execution process, facilitating problem diagnosis and performance optimization.

[0124] The system regularly generates updated statistical reports, summarizing key indicators such as update frequency, score improvement, and data quality trends for each grid unit, providing data support for system operation and maintenance and parameter optimization. By analyzing the update statistics, the spatiotemporal patterns of data acquisition, quality improvement trends, and system performance bottlenecks can be identified, guiding the formulation of image acquisition plans and the optimization of system architecture.

[0125] Figure 2 This is a general technical flow chart provided in an embodiment of this application, such as... Figure 2 As shown in the flowchart, this system illustrates the complete workflow of a task-requirement-based intelligent stereo image pair matching and re-acquisition planning system. The system first transforms the user's stereo mapping task requirements into actionable technical parameters through a task requirement analysis and core parameter quantification module. These parameters include key indicators such as accuracy requirements, coverage, and time constraints, providing quantitative standards for subsequent image selection and matching. The system then selects and matches historical image samples according to the core parameter requirements. This process involves filtering candidate images from a massive image database based on multiple dimensions, including spatial coverage, temporal requirements, resolution standards, and observation geometry. Finally, it uses stereo geometric analysis to form potential image pair combinations.

[0126] The key decision-making node is the matching result determination stage. The system comprehensively evaluates the stereo coverage integrity, geometric configuration quality, and data sufficiency of existing image resources to determine whether the currently available imagery meets the mission requirements. If the existing image resources meet the mission's technical requirements in terms of coverage, stereo geometry configuration, and image quality, the system will directly enter the final output stage, outputting the optimal set of stereo image pairs based on the optimized matching results and supplementary acquisition plan. In this case, efficient resource utilization and rapid response are achieved.

[0127] When the matching result is determined to be unsatisfactory, it indicates that the existing image resources have problems such as coverage gaps, unsatisfactory geometric configurations, insufficient data quality, or poor timeliness. The system will then activate the matching gap feedback and multi-source satellite collaborative supplementary acquisition planning and execution module. This module analyzes the specific data gap types and spatial distributions to formulate targeted supplementary acquisition strategies. This may involve scheduling different satellite platforms for collaborative observation, optimizing observation geometry configurations, and arranging multi-temporal repeated observations to ensure that a complete dataset that meets the mission requirements is obtained. Whether directly utilizing existing resources or obtaining new data through supplementary acquisition, the system will ultimately output an optimized set of stereo image pairs, providing a high-quality data foundation for downstream stereo modeling and mapping production. The entire process embodies the resource allocation concept of "using existing resources first, supplementing as needed, and unified optimization," ensuring both data quality and improving resource utilization efficiency.

[0128] Figure 3 This is a schematic diagram of a scene for multi-angle remote sensing image selection and acquisition based on satellite 3D modeling, as provided in an embodiment of this application. Figure 3 As shown, this diagram illustrates the geometric principle of dual-satellite stereo imaging, depicting the spatial geometric relationship of two satellites observing the same ground target point in stereo. The two blue satellite icons represent the left and right observation satellites, located at different orbital positions or observation times, imaging the same area from different perspectives. The central cube represents the ground target area, with black dots indicating specific ground features or observation targets. Multiple straight lines drawn from the two satellites to the ground target point constitute the geometric ray beams of stereo observation; these rays represent the imaging optical paths and line-of-sight directions of the different satellite sensors.

[0129] The geometric configuration between the two satellites constitutes the basic elements of stereo imaging, including key parameters such as baseline length, observation angle difference, and convergence angle. The observation rays from the left and right satellites intersect at the ground target point in space, forming a basic triangle in stereo geometry. This geometric relationship is the foundation for achieving high-precision stereo measurement. The degree of convergence and angular distribution of the ray beams directly affect the accuracy and reliability of stereo measurement. A larger convergence angle provides better elevation sensitivity, while appropriate angular symmetry helps reduce systematic errors.

[0130] The core advantage of this dual-satellite stereo observation mode lies in its ability to simultaneously acquire multi-angle observation data of the same region, and to calculate the three-dimensional coordinates of ground points through geometric triangulation. The collaborative observation of the two satellites not only improves data acquisition efficiency but also enhances measurement accuracy and data reliability through multiple geometric constraints. The geometric configuration shown in the figure is the core element that needs to be optimized in the stereo image pair selection algorithm. The system needs to select the image pair with the optimal geometric configuration and best observation conditions from among many possible satellite combinations to ensure that stereo modeling and mapping applications can obtain high-quality three-dimensional geographic information products.

[0131] Based on the above method, this application also discloses a multi-angle remote sensing image screening and acquisition system for satellite 3D modeling, such as... Figure 4 As shown, Figure 4 This is a schematic diagram of a multi-angle remote sensing image screening and acquisition system for satellite 3D modeling provided in an embodiment of this application. The system includes: an acquisition module, a determination module, a generation module, a judgment module, and a calculation module; wherein, The system comprises the following modules: an acquisition module, which acquires the task parameters and digital elevation model (DEM) of the target area; a determination module, which calculates the terrain undulation characteristic parameters of the target area based on the DEM; and a generation module, which performs multi-dimensional filtering from the image database for each grid cell based on the task parameters to generate a candidate image set. A judgment module, which determines whether each image pair in the candidate image set meets the stereo image pair criteria based on the constraint parameter set, and includes image pairs that meet the criteria in the valid image pair set. A calculation module, which calculates the comprehensive score of each image pair in the valid image pair set of each grid cell based on the scoring weight coefficient of each grid cell, selects the image pair with the highest comprehensive score that is greater than a preset value from each grid cell as the matching result, generates a unique identifier for the matching result, and outputs the stereo image pair dataset based on the unique identifier.

[0132] It should be noted that the system provided in the above embodiments is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0133] Please see Figure 5This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 As shown, the electronic device 1000 may include: at least one processor 1001, at least one network interface 1004, a user interface 1003, a memory 1005, and at least one communication bus 1002.

[0134] The communication bus 1002 is used to realize the connection and communication between these components.

[0135] The user interface 1003 may include a display screen and a camera. Optionally, the user interface 1003 may also include a standard wired interface and a wireless interface.

[0136] The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0137] The processor 1001 may include one or more processing cores. The processor 1001 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 1005, and by calling data stored in the memory 1005. Optionally, the processor 1001 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 1001 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 1001 and may be implemented as a separate chip.

[0138] The memory 1005 may include random access memory (RAM) or read-only memory. Optionally, the memory 1005 may include a non-transitory computer-readable storage medium. The memory 1005 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 1005 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 1005 may also be at least one storage device located remotely from the aforementioned processor 1001. Figure 5 As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a multi-angle remote sensing image screening and acquisition method for satellite 3D modeling.

[0139] exist Figure 5 In the electronic device 1000 shown, the user interface 1003 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 1001 can be used to call the application program stored in the memory 1005, which is a method for screening and acquiring multi-angle remote sensing images for satellite three-dimensional modeling. When executed by one or more processors, the electronic device performs one or more of the methods described in the above embodiments.

[0140] An electronic device readable storage medium stores instructions that, when executed by one or more processors, cause the electronic device to perform one or more of the methods described in the above embodiments.

[0141] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0142] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0143] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some service interfaces; indirect couplings or communication connections between devices or units may be electrical or other forms.

[0144] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0145] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0146] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0147] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Other embodiments of this disclosure will be readily apparent to those skilled in the art upon consideration of the specification and practice of the disclosure herein. 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 described herein. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for screening and acquiring multi-angle remote sensing images for satellite 3D modeling, characterized in that, The method includes: The task parameters and digital elevation model of the target area are obtained. The task parameters include latitude and longitude boundaries, base-to-height ratio requirements, lateral tilt angle range, stereo mode, temporal threshold, and resolution threshold. The terrain undulation characteristic parameters of the target area are calculated based on the digital elevation model. The terrain undulation characteristic parameters are slope standard deviations, which characterize the complexity of different terrain types. The target area is divided into multiple grid cells based on the terrain undulation characteristic parameters. Different terrain undulation degrees correspond to different grid sizes. Based on the terrain undulation feature parameters, a set of constraint parameters for each grid cell is determined. The set of constraint parameters includes: base-to-height ratio threshold range, angle symmetry threshold, and scoring weight coefficient. For each of the grid cells, multi-dimensional filtering is performed sequentially from the image database according to the task parameters to generate a candidate image set; Based on the set of constraint parameters, determine whether each image pair in the candidate image set satisfies the stereo image pair condition, and include the image pairs that satisfy the stereo image pair condition into the effective image pair set. Based on the scoring weight coefficient of each grid cell, the comprehensive score of each image pair in the effective image pair set of each grid cell is calculated. The image pair with the highest comprehensive score that is greater than the preset score in each grid cell is selected as the matching result. A unique identifier code corresponding to the matching result is generated. The stereo image pair dataset is output based on the unique identifier code.

2. The method for multi-angle remote sensing image selection and acquisition for satellite 3D modeling according to claim 1, characterized in that, After calculating the comprehensive score of each image pair in the effective image pair set of each grid unit based on the scoring weight coefficient of each grid unit, the method further includes: obtaining the image pairs in the effective image pair set whose comprehensive scores are all not greater than a preset score and extracting the missing image features of the grid units to be acquired; generating a acquisition request list based on the missing image features of all the grid units to be acquired; obtaining the transit parameters of satellite resources within the mission time window, and generating at least one acquisition plan for each grid unit to be acquired by combining the acquisition request list and the transit parameters, and arranging the acquisition plans in chronological order to form an acquisition sequence. The process involves sequentially acquiring images according to the acquisition sequence, pairing each newly acquired image with images from the candidate image set, determining whether the newly paired image pair meets the stereo image pair conditions based on the constraint parameter set, and calculating the comprehensive score of the newly paired image pair based on the scoring weight coefficient. When the comprehensive score of the newly paired image pair is greater than the preset score, the completion of the supplementary acquisition task for the corresponding grid unit to be acquired is confirmed, and a unique identifier is generated for the newly paired image pair. When the comprehensive score of the newly paired image pair is not greater than the preset score, the next acquisition plan for the grid unit to be acquired is executed until the supplementary acquisition task is completed or the acquisition sequence is completed.

3. The method for multi-angle remote sensing image selection and acquisition for satellite 3D modeling according to claim 1, characterized in that, The step of calculating the terrain undulation feature parameters of the target area based on the digital elevation model and dividing the target area into multiple grid units based on the terrain undulation feature parameters includes: calculating a slope raster map of the target area based on the digital elevation model; dividing the target area into multiple initial grids according to a preset initial grid size; calculating the slope standard deviation of the initial grids based on the slope raster map; calculating the terrain undulation feature parameters of the target area based on the slope standard deviation; determining the corresponding grid size based on the numerical range of the terrain undulation feature parameters; and dividing the target area into multiple grid units according to the grid size.

4. The method for multi-angle remote sensing image selection and acquisition for satellite 3D modeling according to claim 1, characterized in that, The step of determining the constraint parameter set for each grid cell based on the terrain undulation feature parameters includes: dividing the grid cell into multiple terrain types based on the terrain undulation feature parameters of each grid cell; configuring a corresponding base-to-height ratio threshold range, a corresponding angle symmetry threshold, and a corresponding scoring weight coefficient for each terrain type, and using the base-to-height ratio threshold range, angle symmetry threshold, and scoring weight coefficient as the constraint parameter set; and determining the constraint parameter set for each grid cell based on the terrain type to which each grid cell belongs.

5. The method for multi-angle remote sensing image selection and acquisition for satellite 3D modeling according to claim 1, characterized in that, The step of sequentially performing multi-dimensional filtering from the image database for each grid cell according to the task parameters to generate a candidate image set includes: obtaining images covering the grid cell from the image database; performing temporal filtering according to a temporal threshold in the task parameters to obtain temporally filtered images; performing resolution filtering on the temporally filtered images according to a resolution threshold in the task parameters to obtain resolution filtered images; performing coverage range filtering on the resolution filtered images according to the coverage range of the grid cell to obtain coverage range filtered images; performing lateral angle filtering on the coverage range filtered images according to the lateral angle range in the task parameters to obtain lateral angle filtered images; calculating the incident angle based on the imaging geometry parameters of the lateral angle filtered images; performing incident angle filtering on the lateral angle filtered images to obtain incident angle filtered images; and using the incident angle filtered images as the candidate image set for the grid cell.

6. The method for multi-angle remote sensing image selection and acquisition for satellite 3D modeling according to claim 1, characterized in that, The step of calculating the comprehensive score of each image pair in the effective image pair set of each grid unit based on the scoring weight coefficient of each grid unit includes: calculating the base-to-height ratio score, angular symmetry score, image quality score, and temporal phase difference score for each image pair in the effective image pair set of each grid unit; and weighting and summing the base-to-height ratio score, angular symmetry score, image quality score, and temporal phase difference score according to the scoring weight coefficient of the grid unit to generate the comprehensive score of each image pair in the effective image pair set of each grid unit.

7. The method for multi-angle remote sensing image selection and acquisition for satellite 3D modeling according to claim 1, characterized in that, After outputting the stereo image pair dataset according to the unique identifier, the method further includes: when the image database receives a new image, pairing and scoring the new image with existing images; when the comprehensive score of the newly paired image pair is higher than the comprehensive score of the image pair currently used by the grid unit, updating the recommended image pair of the grid unit and updating the unique identifier.

8. A multi-angle remote sensing image selection and acquisition system for satellite 3D modeling, characterized in that, The system includes: an acquisition module, a determination module, a generation module, a judgment module, and a calculation module; wherein, The acquisition module is used to acquire task parameters and a digital elevation model of the target area. The task parameters include latitude and longitude boundaries, base-to-height ratio requirements, lateral tilt angle range, stereo mode, temporal threshold, and resolution threshold. Based on the digital elevation model, the module calculates the terrain undulation characteristic parameters of the target area, where the terrain undulation characteristic parameters are slope standard deviations, representing the complexity of different terrain types. The module then divides the target area into multiple grid cells based on the terrain undulation characteristic parameters, with different grid sizes corresponding to different degrees of terrain undulation. The determining module is used to determine the constraint parameter set of each grid cell based on the terrain undulation feature parameters. The constraint parameter set includes: base-to-height ratio threshold range, angle symmetry threshold, and scoring weight coefficient. The generation module is used to perform multi-dimensional filtering sequentially from the image database for each grid cell according to the task parameters to generate a candidate image set. The judgment module is used to determine whether each image pair in the candidate image set satisfies the stereo image pair condition based on the constraint parameter set, and to include the image pairs that satisfy the stereo image pair condition into the effective image pair set. The calculation module is used to calculate the comprehensive score of each image pair in the effective image pair set of each grid unit according to the scoring weight coefficient of each grid unit, select the image pair with the highest comprehensive score that is greater than a preset value in each grid unit as the matching result, generate a unique identifier code corresponding to the matching result, and output the stereo image pair dataset according to the unique identifier code.

9. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed as described in any one of claims 1-7.

Images