Satellite remote sensing image on-satellite geometric positioning method and system based on ground elevation data

Abstract

The invention belongs to the technical field of satellite remote sensing technology and remote sensing image processing, and particularly relates to a satellite remote sensing image on-satellite geometric positioning method and system based on ground elevation data. Precise elevation matching is realized by adopting SIFT feature matching and terrain adaptive Kriging interpolation, and multi-dimensional error correction is carried out in combination with GNSS RTK data and sliding window adjustment. The system comprises corresponding function modules and is realized by adopting an FPGA + DSP heterogeneous architecture. According to the invention, the problems of poor terrain adaptability, limited positioning precision and insufficient real-time performance in the prior art are solved, and high-precision real-time on-satellite geometric positioning of a complex terrain area is realized.

Description

Technical Field

[0001] This invention belongs to the field of satellite remote sensing technology and remote sensing image processing technology, specifically relating to a satellite remote sensing image on-board geometric positioning method and system based on ground elevation data. Background Technology

[0002] Onboard geometric positioning of satellite remote sensing images is a core technology for realizing real-time applications of remote sensing data, and its accuracy directly determines the application value of remote sensing images in emergency disaster monitoring, military reconnaissance, and resource surveys. With the development of aerospace technology, the computing power and storage capacity of onboard processing systems have been significantly improved. However, existing onboard geometric positioning methods mostly use single-resolution elevation data, which cannot balance positioning accuracy and processing efficiency.

[0003] The existing technology CN118859265A discloses a satellite remote sensing image on-board geometric positioning method and system based on ground elevation data. It uses a fixed downsampling rate to process elevation data and obtains elevation values ​​through an iterative approach from the highest altitude to sea level. However, it has the following limitations: First, the elevation data processing lacks terrain adaptability, and the fixed downsampling rate easily leads to the loss of elevation information in complex terrain areas. Second, the geometric positioning process does not consider the dynamic changes in the satellite's attitude and the radial error of its orbit during motion, resulting in limited positioning model accuracy. Third, the elevation iteration method is inefficient and cannot meet the real-time processing requirements of high-resolution remote sensing images. Fourth, it does not integrate multi-source auxiliary data, resulting in poor error correction capabilities.

[0004] Therefore, a satellite remote sensing image on-board geometric positioning method and system based on ground elevation data is provided to address the shortcomings of the aforementioned prior art, while taking into account adaptability, real-time performance, and high accuracy. Summary of the Invention

[0005] To address the aforementioned shortcomings in existing technologies, this invention provides a satellite remote sensing image on-board geometric positioning method based on ground elevation data. This method solves the problems in the background technology, such as the loss of elevation information in complex terrain areas due to fixed adoption rates, limited positioning model accuracy, low efficiency of elevation iteration methods, and poor error correction capabilities.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: A satellite remote sensing image on-board geometric positioning method based on ground elevation data includes the following steps: S1: Set the positioning accuracy threshold, use an adaptive block downsampling algorithm to process global ground elevation data, and obtain a multi-level resolution elevation dataset; simultaneously acquire multi-source satellite and payload data, and form a standardized input parameter set through fusion denoising, error compensation and on-orbit calibration. S2: Based on the standardized input parameter set, construct a dynamic collinear equation geometric model containing an elevation error correction term, a real-time attitude angle change rate, and a track radial error compensation factor; S3: Adaptive selection of resolution elevation data blocks achieves coordinate mapping through SIFT feature matching and quadtree indexing, and calculates accurate ground elevation values ​​by combining kriging interpolation of terrain slope factor. S4: Substitute the image plane coordinates and the ground elevation values ​​into the dynamic collinearity equation geometric model, solve the initial positioning value through variable step size Levenberg-Marquardt iteration, and obtain the preliminary positioning result by combining the terrain slope weight matrix for optimization. S5: Based on real-time dynamic positioning (GNSS RTK) data from the Global Navigation Satellite System, orbital parameters are corrected, outliers are removed using the Local Anomaly Factor (LOF) algorithm, and after sliding window adjustment optimization, high-precision on-board geometric positioning results with timestamps are output.

[0007] Furthermore, the specific process of the adaptive block downsampling algorithm in step S1 is as follows: Based on the on-board geometric positioning accuracy threshold, the elevation data is adaptively divided into blocks. The terrain complexity is determined by detecting changes in terrain slope within the blocks. For complex terrain with a slope greater than the preset slope threshold, a low-ratio downsampling is used (to retain sufficient elevation details). For gentle terrain with a slope less than or equal to the preset slope threshold, a high-ratio adaptive sampling rate is used (to reduce the amount of data processing). Overlapping sampling is used at the edges of the segmented units to avoid stitching errors, and the resulting multi-level resolution elevation dataset can be dynamically switched and invoked according to the terrain complexity of the imaging area.

[0008] Furthermore, the fusion denoising in step S1 employs a federated Kalman filter algorithm, specifically implemented as follows: The orbital parameters and attitude parameters are treated as two independent subsystems, and state equations and observation equations are established separately. The filtering results of the subsystems are fused by an adaptive weight allocation factor, which is calculated in real time based on the observation noise covariance of the two subsystems.

[0009] Furthermore, the calculation of the terrain slope shadow and the kriging interpolation adaptation process in step S3 are as follows: Based on the elevation values ​​of four adjacent elevation grid points around a specific image point, the terrain slope angle of that point is calculated using a plane fitting algorithm. Topographic slope factor ; The mutation function model of the Kriging interpolation algorithm is based on Adaptive adjustment, when When using a spherical variogram, when The Gaussian variogram is used to calculate the accurate ground elevation value.

[0010] Furthermore, the construction rule for the terrain slope weight matrix in step S4 is as follows: Set the terrain slope angle of the specific positioning image point Weight matrix elements , where k is the attenuation coefficient, which is dynamically called from a preset terrain parameter library; The weight matrix is ​​introduced into the localization result optimization process, and the initial localization value is optimized by weighted least squares to obtain the preliminary localization result.

[0011] Furthermore, the orbit parameter correction process based on GNSS RTK data in step S5 is as follows: Using the instantaneous satellite position calculated from GNSS RTK data as a reference, the deviation values ​​of orbital parameters in the standardized input parameter set are calculated. The orbital deviation correction for each specific image point is obtained through linear interpolation. This is then applied to the satellite sensor optical center coordinates in the dynamic collinearity equation geometric model. Make real-time corrections.

[0012] A satellite remote sensing image on-board geometric positioning system based on ground elevation data, and based on the aforementioned satellite remote sensing image on-board geometric positioning method based on ground elevation data, includes: The multi-source data fusion preprocessing module is used to set the on-board geometric positioning accuracy threshold, and uses an adaptive block downsampling algorithm to process global ground elevation data to obtain a multi-level resolution elevation dataset. It also uses federated Kalman waver rate, clock error correction, and on-orbit calibration to fuse and preprocess satellite orbital parameters, attitude parameters, and GNSS observation data, and outputs a standardized set of input parameters. The dynamic geometric model construction module is used to construct a dynamic collinear equation geometric model that includes an elevation error correction term and an orbital radial error compensation factor, based on a standardized set of input parameters and in combination with satellite motion characteristics and imaging mechanisms. The elevation intelligent matching interpolation module is used to adaptively select elevation data blocks based on remote sensing image imaging information, achieve rapid coordinate mapping through SIFT feature matching and quadtree indexing, and calculate accurate ground elevation values ​​using a Kriging interpolation algorithm that combines terrain slope factors. The iterative even-weighted positioning calculation module is used to substitute the image plane coordinates and the precise ground elevation values ​​into the dynamic collinearity equation geometric model, and obtain the preliminary positioning results through the variable step size Levenberg-Marquardt iterative algorithm and weight matrix weighted optimization. The multi-dimensional error correction output module is used to build a benchmark verification model based on GNSS RTK data, remove outliers through the local anomaly factor algorithm, optimize the positioning results by combining the sliding window adjustment algorithm, and output high-precision positioning data with timestamps. The on-board storage and scheduling module is used to store multi-level resolution elevation datasets, standardized input parameter sets, positioning intermediate results and final results, and supports dynamic scheduling of data read and write priorities based on the on-board processing resource usage.

[0013] Furthermore, the multi-source data fusion preprocessing module includes an elevation data processing unit, a parameter fusion processing unit, and a load parameter verification unit; The elevation data processing unit has a built-in adaptive block downsampling algorithm and block edge overlap processing logic; The parameter fusion processing unit integrates the federated Kalman filter algorithm and the GNSS error compensation model. The load parameter verification unit corrects the deviation of the internal orientation element in real time through the on-orbit calibration algorithm.

[0014] Furthermore, the elevation intelligent matching interpolation module includes a resolution adaptation unit, a coordinate mapping unit, and a fine interpolation unit; The resolution adaptation unit selects matching data blocks from a multi-level resolution elevation dataset based on the terrain complexity of the imaging area. The coordinate mapping unit achieves rapid localization through SIFT feature matching and quadtree indexing; The fine interpolation unit incorporates terrain slope factor calculation logic and an adaptive variogram-based Kriging interpolation algorithm.

[0015] 10. Further, the multidimensional error correction output model includes a GNSS reference correction unit, an outlier removal unit, and an adjustment optimization unit; The GNSS reference correction unit corrects orbital parameter deviations using GNSS RTK data. The anomaly removal unit uses a local anomaly factor algorithm to identify and locate outliers. The adjustment optimization unit achieves final optimization of the positioning results through a sliding window adjustment algorithm.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. An adaptive block downsampling algorithm is used to process elevation data. The downsampling rate is dynamically adjusted according to the terrain complexity, which not only ensures the integrity of elevation information in complex terrain areas, but also reduces the amount of data processing in flat terrain areas, thus solving the terrain adaptability problem caused by the fixed downsampling rate in existing technologies.

[0017] 2. A dynamic collinearity equation geometric model is constructed, which includes an elevation error correction term, an attitude dynamic change compensation factor, and an orbital radial error compensation factor. This model fully considers the dynamic errors during satellite motion and significantly improves positioning accuracy compared to the static model.

[0018] 3. The elevation matching technology, which integrates SIFT feature matching, quadtree indexing and terrain-adaptive kriging interpolation, significantly improves the accuracy and efficiency of elevation matching in complex terrain areas.

[0019] 4. By introducing GNSS RTK data to construct a benchmark verification model and combining a multi-dimensional error correction mechanism of local outlier elimination and sliding window adjustment, high-precision positioning can be achieved in areas without ground control points, meeting the needs of high-precision remote sensing applications.

[0020] 5. Each module adopts a parallel processing and dynamic resource scheduling strategy, which can control the on-board processing latency within the range that meets the real-time processing requirements, and adapt to the processing scenarios of high-resolution remote sensing images. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a satellite remote sensing image on-board geometric positioning method based on ground elevation data according to the present invention.

[0023] Figure 2 This is a system diagram of an on-board geometric positioning system for satellite remote sensing images based on ground elevation data, according to the present invention. Detailed Implementation

[0024] Please see Figure 1 This invention provides a satellite remote sensing image on-board geometric positioning method based on ground elevation data, comprising the following steps: S1: Set the positioning accuracy threshold, use an adaptive block downsampling algorithm to process global ground elevation data, and obtain a multi-level resolution elevation dataset; simultaneously acquire multi-source satellite and payload data, and form a standardized input parameter set through fusion denoising, error compensation and on-orbit calibration. Specifically, based on the positioning accuracy requirements of the satellite remote sensing mission, an on-board geometric positioning accuracy threshold is set (the value ranges from 1m to 5m, determined according to the requirements of the remote sensing mission). Based on this accuracy threshold, an optimized adaptive block downsampling algorithm is used to perform hierarchical processing on global ground elevation data (such as SRTM series elevation data). The core of the algorithm optimization lies in building a dual-objective adaptation mechanism of "accuracy-efficiency": 1) Adaptive block unit division: Divide the basic block unit into 1°×1° blocks according to latitude and longitude. By detecting the change of terrain slope within the unit (if there is a sudden change in slope within the unit, that is, the slope difference between adjacent grids exceeds the preset threshold), the sub-units are automatically subdivided to ensure that the data granularity of complex terrain areas can adapt to the positioning accuracy requirements; 2) Dynamic decision-making of downsampling rate: By establishing a mapping relationship between the positioning accuracy threshold and the terrain complexity, the downsampling rate of different areas is dynamically determined. That is, a low proportion of adaptive downsampling rate is used for complex terrain areas to retain elevation details, and a high proportion of adaptive downsampling rate is used for flat terrain areas to reduce the amount of data; 3) Block edge optimization processing: Through the strategy of overlapping sampling + smooth edge transition, the problem of discontinuity of elevation information or gradient abruptness at the block splicing is avoided. The final result is a multi-level resolution elevation dataset, which is stored in the on-board high-speed buffer unit and can be dynamically accessed in real time according to the terrain complexity of the imaging area.

[0025] Simultaneously acquire satellite orbital parameters (including orbital semi-major axis, eccentricity, and inclination), attitude parameters (including roll angle, pitch angle, and yaw angle), GNSS observation data (including pseudorange and carrier phase), azimuth elements of the remote sensing payload (including principal point coordinates and principal distance), and raw remote sensing images. Standardized input is constructed through a multi-source data preprocessing workflow. Its core function is to achieve noise reduction, completion, and consistency calibration of multi-source data. The steps are: 1) Data synchronization and alignment: Based on the high-precision on-board timestamp (1 mm accuracy), interpolation alignment of data with different acquisition rates is performed to ensure the consistency of orbital parameters, attitude parameters, and GNSS observation data in the time dimension. 2) Multi-source parameter fusion and denoising: The federated Kalman filter algorithm is used to fuse orbit and attitude parameters. The two are treated as independent subsystems to establish state equations and observation equations respectively. Combined with the GNSS error compensation model, error compensation such as clock error and ionospheric delay is completed. The filtering results of the subsystem are fused through an adaptive weight allocation factor (which is calculated in real time from the subsystem observation noise covariance); 3) Interior orientation element calibration: Based on the on-board preset calibration template, the on-orbit calibration and deviation correction are completed through the image matching algorithm. The final output includes a standardized input parameter set containing the elevation dataset index, standardized orbit / attitude / GNSS parameters, and corrected interior orientation elements.

[0026] S2: Based on the standardized input parameter set, construct a dynamic collinear equation geometric model containing an elevation error correction term, a real-time attitude angle change rate, and a track radial error compensation factor; Specifically, based on the standardized input parameter set, and combining the satellite platform's motion characteristics (attitude angle change rate, orbital radial velocity) with the imaging mechanism of the payload, an optimized dynamic collinearity equation geometric model is constructed. The core of the optimization lies in adapting to satellite dynamic errors and terrain errors. Specific optimization details include: 1) Attitude dynamic change compensation: Introducing the real-time attitude angle change rate (calculated by the differential of attitude parameters) and an adaptive attenuation factor (range 0.8~1.0), the attenuation factor value is dynamically adjusted by the attitude angle change frequency to reduce the impact of high-frequency dynamic errors on the direction cosine matrix; 2) Orbital radial error compensation: Based on the radial deviation between GNSS observation data and orbital parameters, a dynamic compensation factor is constructed. To avoid compensation fluctuations caused by single-point deviations, radial deviations are smoothed using a sliding window before being substituted into the calculation; 3) Terrain-adaptive elevation correction: To address elevation deviation issues in complex terrains, an elevation error correction term related to the terrain slope angle is designed. By correlating slope, targeted corrections are achieved for different terrains. The optimized dynamic collinearity equation geometric model, based on the traditional collinearity equation, adapts to the needs of on-board dynamic imaging and complex terrain positioning through multi-dimensional optimization.

[0027] The optimized expression for the dynamic collinearity equation geometric model is as follows: in, The image plane coordinates of the image point (with the principal image point as the origin); The corresponding three-dimensional coordinates of the ground points (WGS-48 coordinate system); These are the instantaneous coordinates of the optical center of the satellite sensor; For remote sensing systems, the main distance is [missing information]. , , , , , , , , Elements of the direction cosine matrix; This is the elevation error correction term. ,in This is a preliminary elevation estimate. The terrain slope angle for a specific image point (used to adapt to complex terrain); This is the radial error compensation factor for the track. ,in This represents the radial deviation of the orbit (obtained by comparing GNSS observation data with orbital parameters). This is the satellite's orbital radius (obtained from orbital parameters).

[0028] S3: Adaptive selection of resolution elevation data blocks achieves coordinate mapping through SIFT feature matching and quadtree indexing, and calculates accurate ground elevation values ​​by combining kriging interpolation of terrain slope factor. Specifically, based on the imaging field of view of the original remote sensing image (calculated from satellite orbit parameters and payload field of view angle) and the shooting time, elevation data blocks with matching resolution are adaptively selected from the multi-level resolution elevation dataset: if the imaging area is a complex terrain area, a high-resolution elevation data block is selected; if it is a flat terrain area, a low-resolution elevation data block is selected to balance processing efficiency and positioning accuracy.

[0029] Intelligent elevation matching is achieved using a combination of feature matching and index acceleration: 1) SIFT feature matching for rapid mapping: 500-1000 feature points are extracted from the original remote sensing image (each frame). These points are then matched against a pre-defined feature point database to obtain approximate ground coordinates. An affine transformation model is then constructed to perform initial coordinate mapping of the specific image points. 2) Quadtree indexing for accelerated positioning: A quadtree index is built from the elevation data blocks. After inputting the initially mapped ground coordinates, the index quickly locates the elevation search area within a pre-defined range. This combination of feature matching and index acceleration ensures both speed and accuracy in coordinate mapping, reducing the search range and computational load for subsequent fine interpolation.

[0030] An optimized terrain-adaptive kriging interpolation algorithm is used to calculate accurate ground elevation values. The core of the algorithm optimization is the dynamic adaptation of the variogram function: 1) Accurate calculation of terrain slope factor: Based on the elevation values ​​of adjacent elevation grid points within the search area, the elevation surface equation is solved by plane fitting. Calculate the normal vector of the image point to be located, and then obtain the terrain slope angle. Topographic slope factor ( (1) Core data related to the formula, used to characterize the flatness of the terrain); 2) Adaptive selection mechanism for the variogram function, with preset parameter templates for the spherical variogram function (adapted to nonlinear terrain) and the Gaussian variogram function (adapted to flat terrain), and using the slope factor. Trigger selection: when When the value is less than a preset threshold, the spherical mutation function is invoked; when... When the value is greater than or equal to the preset threshold, the Gaussian mutation function is called; 3) Smoothing optimization of interpolation results: the interpolated elevation values ​​are smoothed by a preset window to remove isolated outliers, and finally a high-precision, smooth ground elevation value Z of the image point to be located is obtained (Z is the core input data of the dynamic collinearity equation).

[0031] S4: Substitute the image plane coordinates and the ground elevation values ​​into the dynamic collinearity equation geometric model, solve the initial positioning value through variable step size Levenberg-Marquardt iteration, and obtain the preliminary positioning result by combining the terrain slope weight matrix for optimization. Specifically, the image plane coordinates (x, y) of the image point to be located and the precise ground elevation value Z are substituted into the optimized dynamic collinearity equation geometric model to construct a nonlinear equation system about the ground three-dimensional coordinates (X, Y). Emphasis is placed on the on-board adaptation and step-by-step connection of the iteration path: 1) Initial value optimization selection: based on a coarse mapping between satellite orbital altitude and image plane coordinates, combined with preliminary elevation information from multi-level elevation datasets, a more realistic initial iteration value is obtained; 2) Variable step size adaptive adjustment: the iteration step size is dynamically adjusted according to the residual change rate; 3) Convergence judgment optimization: a "residual threshold + iteration count dual judgment" is adopted. Finally, when the iteration meets the convergence condition, the initial value of the ground three-dimensional coordinates is obtained.

[0032] A weight matrix is ​​constructed based on slope information from ground elevation data: Let the terrain slope angle of the image point to be located be... Weight matrix elements Where k is the attenuation coefficient (dynamically adjusted according to the positioning accuracy threshold), dynamically called from a preset terrain parameter library. The weight matrix is ​​introduced into the positioning result optimization process, and weighted least squares optimization is performed on the initial values ​​of the ground 3D coordinates to obtain preliminary positioning results. (k is the core data used to construct the weight matrix, and it directly affects the optimization accuracy).

[0033] S5: Based on real-time dynamic positioning (GNSS RTK) data from the Global Navigation Satellite System, orbit parameters are corrected, outliers are removed using the Local Anomaly Factor (LOF) algorithm, and after sliding window adjustment optimization, high-precision on-board geometric positioning results with timestamps are output. Specifically, positioning accuracy is improved through a multi-source error correction process: 1) GNSS RTK reference correction: The GNSS RTK data calculation algorithm is called to obtain the instantaneous precise position of the satellite, a reference verification model is constructed, the deviation between the orbit parameter calculated position and the RTK reference position is calculated, and the orbit deviation correction amount for each image point to be positioned is obtained through linear interpolation, thus completing the satellite sensor optical center coordinates in the dynamic collinearity equation. The system employs the following steps: 1) Real-time correction to improve the accuracy of the dynamic collinearity equation geometric model; 2) LOF outlier removal: The LOF algorithm is used to construct the neighborhood relationship of image points, calculate the LOF value of each image point, and remove outlier positioning points with LOF values ​​greater than a preset threshold; 3) Sliding window adjustment optimization: The sliding window adjustment algorithm is used, with a fixed-size pixel window as the unit, and the effective positioning results within the window are used as observations to construct the error equation. The optimized coordinates of the image point at the center of the window are obtained by solving the least squares method, thereby reducing local random errors. The overall process of "benchmark correction - outlier removal - adjustment optimization" achieves multi-dimensional error cancellation at each level.

[0034] The Local Anomaly Factor (LOF) algorithm is used to identify and remove abnormal localization points: taking each image point to be located as the core, a preset number of neighboring image points are selected as the neighborhood, the local anomaly factor value of the core point is calculated, and if the LOF value is greater than the preset threshold, the point is determined to be an abnormal localization point and is removed.

[0035] The effective positioning results are finally optimized by combining the sliding window adjustment algorithm: a fixed-size sliding window is set, and all effective positioning results within the window are used as observations. An error equation is constructed, and the error equation is solved by the least squares method to obtain the optimized 3D ground coordinates of the window center image point. (This aligns with the objective of solving the dynamic collinearity equation). Finally, the optimized coordinates of all image points are associated with their corresponding cell positions and timestamps to form the final on-board geometric positioning results, which are stored in the on-board storage module. These results can be downloaded in real-time via the satellite-to-ground link or cached locally as needed for the mission.

[0036] Please see Figure 2 A satellite remote sensing image on-board geometric positioning system based on ground elevation data includes: The multi-source data fusion preprocessing module is used to set the on-board geometric positioning accuracy threshold, and uses an adaptive block downsampling algorithm to process global ground elevation data to obtain a multi-level resolution elevation dataset. It also uses federated Kalman waver rate, clock error correction, and on-orbit calibration to fuse and preprocess satellite orbital parameters, attitude parameters, and GNSS observation data, and outputs a standardized set of input parameters. The dynamic geometric model construction module is used to construct a dynamic collinear equation geometric model that includes an elevation error correction term and an orbital radial error compensation factor, based on a standardized set of input parameters and in combination with satellite motion characteristics and imaging mechanisms. The elevation intelligent matching interpolation module is used to adaptively select elevation data blocks based on remote sensing image imaging information, achieve rapid coordinate mapping through SIFT feature matching and quadtree indexing, and calculate accurate ground elevation values ​​using a Kriging interpolation algorithm that combines terrain slope factors. The iterative even-weighted positioning calculation module is used to substitute the image plane coordinates and the precise ground elevation values ​​into the dynamic collinearity equation geometric model, and obtain the preliminary positioning results through the variable step size Levenberg-Marquardt iterative algorithm and weight matrix weighted optimization. The multi-dimensional error correction output module is used to build a benchmark verification model based on GNSS RTK data, remove outliers through the local anomaly factor algorithm, optimize the positioning results by combining the sliding window adjustment algorithm, and output high-precision positioning data with timestamps. The on-board storage and scheduling module is used to store multi-level resolution elevation datasets, standardized input parameter sets, positioning intermediate results and final results, and supports dynamic scheduling of data read and write priorities based on the on-board processing resource usage.

[0037] The satellite remote sensing image onboard geometric positioning system based on ground elevation data is implemented using an FPGA+DSP irregular architecture. Each module interacts with data through a high-speed serial bus, adapting to the requirements of low power consumption and high reliability operation on the satellite.

[0038] The multi-source data fusion preprocessing module includes an elevation data processing unit, a parameter fusion processing unit, and a load parameter verification unit. The elevation data processing unit has a built-in adaptive block downsampling algorithm and block edge overlap processing logic; The parameter fusion processing unit integrates the federated Kalman filter algorithm and the GNSS error compensation model (supporting error compensation for ionospheric delay, tropospheric delay, clock error, etc.). The load parameter verification unit corrects the deviation of the interior orientation element in real time through the on-orbit calibration algorithm (calibration frequency once per hour, calibration accuracy 0.01 pixels).

[0039] Secondly, the elevation intelligent matching interpolation module includes a resolution adaptation unit, a coordinate mapping unit, and a fine interpolation unit; The resolution adaptation unit selects matching data blocks from a multi-level resolution elevation dataset based on the terrain complexity of the imaging area. The coordinate mapping unit achieves rapid localization through SIFT feature matching and quadtree indexing; The fine interpolation unit incorporates terrain slope factor calculation logic and an adaptive variogram-based Kriging interpolation algorithm.

[0040] Then, the GNSS reference correction unit corrects orbital parameter deviations using GNSS RTK data; The anomaly removal unit uses a local anomaly factor algorithm to identify and locate outliers. The adjustment optimization unit achieves final optimization of the positioning results through a sliding window adjustment algorithm.

[0041] This invention discloses an on-board geometric positioning method and system based on satellite remote sensing images using ground elevation data. It employs an adaptive block downsampling algorithm to perform hierarchical processing of global ground elevation data, and fuses multi-source data such as satellite orbit, attitude, and GNSS data for preprocessing. A dynamic collinearity equation geometric model is constructed, including elevation error correction terms, attitude dynamic change compensation factors, and orbit radial error compensation factors. Intelligent elevation matching is achieved using SIFT feature matching, quadtree indexing, and terrain-adaptive kriging interpolation. Preliminary positioning results are obtained through variable step-size iteration and weight optimization, and multi-dimensional error correction is performed by combining GNSS RTK data, outlier removal, and sliding window adjustment. The system includes a multi-source data fusion preprocessing module, a dynamic geometric model construction module, an intelligent elevation matching interpolation module, an iterative even-weighted positioning calculation module, a multi-dimensional error correction output module, and an on-board storage and scheduling module. This invention features strong terrain adaptability, high positioning accuracy, and good real-time performance, meeting the high-precision on-board geometric positioning requirements of various remote sensing satellites.

[0042] The above description discloses only one preferred embodiment of the present invention, and should not be construed as limiting the scope of the present invention. Those skilled in the art will understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A satellite-based geometric positioning method for satellite remote sensing images based on ground elevation data, characterized in that, Includes the following steps: S1: Set the positioning accuracy threshold, use an adaptive block downsampling algorithm to process global ground elevation data, and obtain a multi-level resolution elevation dataset; simultaneously acquire multi-source satellite and payload data, and form a standardized input parameter set through fusion denoising, error compensation and on-orbit calibration; S2: Based on the standardized input parameter set, construct a dynamic collinear equation geometric model containing an elevation error correction term, a real-time attitude angle change rate, and a track radial error compensation factor; S3: Adaptive selection of resolution elevation data blocks achieves coordinate mapping through SIFT feature matching and quadtree indexing, and calculates accurate ground elevation values ​​by combining kriging interpolation of terrain slope factor. S4: Substitute the image plane coordinates and the ground elevation values ​​into the dynamic collinearity equation geometric model, solve the initial positioning value through variable step size Levenberg-Marquardt iteration, and obtain the preliminary positioning result by combining the terrain slope weight matrix for optimization. S5: Based on real-time dynamic positioning (GNSS RTK) data from the Global Navigation Satellite System, orbital parameters are corrected, outliers are removed using the Local Anomaly Factor (LOF) algorithm, and after sliding window adjustment optimization, high-precision on-board geometric positioning results with timestamps are output.

2. The satellite remote sensing image on-board geometric positioning method based on ground elevation data as described in claim 1, characterized in that, The specific process of the adaptive block downsampling algorithm in step S1 is as follows: Based on the on-board geometric positioning accuracy threshold, the elevation data is adaptively divided into blocks. The terrain complexity is determined by detecting changes in terrain slope within the blocks. For complex terrain with a slope greater than the preset slope threshold, a low-ratio downsampling is used (to retain sufficient elevation details). For gentle terrain with a slope less than or equal to the preset slope threshold, a high-ratio adaptive sampling rate is used (to reduce the amount of data processing). Overlapping sampling is used at the edges of the segmented units to avoid stitching errors, and the resulting multi-level resolution elevation dataset can be dynamically switched and invoked according to the terrain complexity of the imaging area.

3. The satellite remote sensing image on-board geometric positioning method based on ground elevation data as described in claim 1, characterized in that, The fusion and denoising process in step S1 employs a federated Kalman filter algorithm, specifically implemented as follows: The orbital parameters and attitude parameters are treated as two independent subsystems, and state equations and observation equations are established separately. The filtering results of the subsystems are fused by an adaptive weight allocation factor, which is calculated in real time based on the observation noise covariance of the two subsystems.

4. The satellite remote sensing image on-board geometric positioning method based on ground elevation data as described in claim 1, characterized in that, The calculation of the terrain slope shadow and the Kriging interpolation adaptation process in step S3 are as follows: Based on the elevation values ​​of four adjacent elevation grid points around a specific image point, the terrain slope angle of that point is calculated using a plane fitting algorithm. Topographic slope factor ; The mutation function model of the Kriging interpolation algorithm is based on Adaptive adjustment, when When using a spherical variogram, when The Gaussian variogram is used to calculate the accurate ground elevation value.

5. The satellite remote sensing image on-board geometric positioning method based on ground elevation data as described in claim 4, characterized in that, The construction rule for the terrain slope weight matrix in step S4 is as follows: Set the terrain slope angle of the specific positioning image point Weight matrix elements , where k is the attenuation coefficient, which is dynamically called from a preset terrain parameter library; The weight matrix is ​​introduced into the localization result optimization process, and the initial localization value is optimized by weighted least squares to obtain the preliminary localization result.

6. The satellite remote sensing image on-board geometric positioning method based on ground elevation data as described in claim 4, characterized in that, The orbit parameter correction process based on GNSS RTK data in step S5 is as follows: Using the instantaneous satellite position calculated from GNSS RTK data as a reference, the deviation values ​​of orbital parameters in the standardized input parameter set are calculated. The orbital deviation correction for each specific image point is obtained through linear interpolation. This is then applied to the satellite sensor optical center coordinates in the dynamic collinearity equation geometric model. Make real-time corrections.

7. A satellite remote sensing image on-board geometric positioning system based on ground elevation data, comprising a satellite remote sensing image on-board geometric positioning method based on ground elevation data as described in any one of claims 1-6, characterized in that, include: The multi-source data fusion preprocessing module is used to set the on-board geometric positioning accuracy threshold, and uses an adaptive block downsampling algorithm to process global ground elevation data to obtain a multi-level resolution elevation dataset. It also uses federated Kalman waver rate, clock error correction, and on-orbit calibration to perform fusion preprocessing on satellite orbital parameters, attitude parameters, and GNSS observation data, and outputs a standardized set of input parameters. The dynamic geometric model construction module is used to construct a dynamic collinear equation geometric model that includes an elevation error correction term and an orbital radial error compensation factor, based on a standardized set of input parameters and in combination with satellite motion characteristics and imaging mechanisms. The elevation intelligent matching interpolation module is used to adaptively select elevation data blocks based on remote sensing image imaging information, achieve rapid coordinate mapping through SIFT feature matching and quadtree indexing, and calculate accurate ground elevation values ​​using a Kriging interpolation algorithm that combines terrain slope factors. The iterative even-weighted positioning calculation module is used to substitute the image plane coordinates and the precise ground elevation values ​​into the dynamic collinearity equation geometric model, and obtain the preliminary positioning results through the variable step size Levenberg-Marquardt iterative algorithm and weight matrix weighted optimization. The multi-dimensional error correction output module is used to build a benchmark verification model based on GNSS RTK data, remove outliers through the local anomaly factor algorithm, optimize the positioning results by combining the sliding window adjustment algorithm, and output high-precision positioning data with timestamps. The on-board storage and scheduling module is used to store multi-level resolution elevation datasets, standardized input parameter sets, positioning intermediate results and final results, and supports dynamic scheduling of data read and write priorities based on the on-board processing resource usage.

8. The satellite remote sensing image on-board geometric positioning system based on ground elevation data as described in claim 7, characterized in that, The multi-source data fusion preprocessing module includes an elevation data processing unit, a parameter fusion processing unit, and a load parameter verification unit. The elevation data processing unit has a built-in adaptive block downsampling algorithm and block edge overlap processing logic; The parameter fusion processing unit integrates the federated Kalman filter algorithm and the GNSS error compensation model. The load parameter verification unit corrects the deviation of the internal orientation element in real time through the on-orbit calibration algorithm.

9. The satellite remote sensing image on-board geometric positioning system based on ground elevation data as described in claim 7, characterized in that, The elevation intelligent matching interpolation module includes a resolution adaptation unit, a coordinate mapping unit, and a fine interpolation unit; The resolution adaptation unit selects matching data blocks from a multi-level resolution elevation dataset based on the terrain complexity of the imaging area. The coordinate mapping unit achieves rapid localization through SIFT feature matching and quadtree indexing; The fine interpolation unit incorporates terrain slope factor calculation logic and an adaptive variogram-based Kriging interpolation algorithm.

10. The satellite remote sensing image on-board geometric positioning system based on ground elevation data as described in claim 7, characterized in that, The multidimensional error correction output model includes a GNSS reference correction unit, an outlier removal unit, and adjustment optimization unit. The GNSS reference correction unit corrects orbital parameter deviations using GNSS RTK data. The anomaly removal unit uses a local anomaly factor algorithm to identify and locate outliers. The adjustment optimization unit achieves final optimization of the positioning results through a sliding window adjustment algorithm.

Images