Multi-frame super-resolution imaging method and device of optical remote sensing satellite and storage medium

Abstract

The application discloses a kind of multi-frame super-resolution imaging method, device and storage medium of optical remote sensing satellite, it is related to satellite remote sensing and image processing field, comprising: according to the observation task demand to earth, satellite orbit parameter and constellation configuration information, satellite attitude control instruction containing preset sub-pixel level displacement strategy is generated;Multi-frame ground target image collected by satellite according to satellite attitude control instruction is received, and auxiliary data synchronously returned by satellite;Based on auxiliary data, carry out sub-pixel level geometric registration to multi-frame ground target image, while executing photometric normalization processing to the image difference under different observation conditions;Multi-frame ground target image after registration normalization is fused and reconstructed to super-resolution, and optical remote sensing image is obtained.The application can improve the accuracy of multi-frame super-resolution imaging of optical remote sensing satellite.

Description

Technical Field

[0001] This application relates to the field of satellite remote sensing and image processing, and in particular to a multi-frame super-resolution imaging method, apparatus and storage medium for optical remote sensing satellites. Background Technology

[0002] In the fields of aerospace remote sensing and Earth observation, optical remote sensing satellites, with their high timeliness and high spatial resolution, have been widely used in key scenarios such as land mapping, disaster monitoring, resource exploration, and ecological assessment. Their imaging resolution directly determines the accuracy of identification and analysis of ground targets. However, the imaging resolution of optical remote sensing satellites has long been limited by the physical performance bottlenecks of the optical system, including optical aperture size, lens focal length, manufacturing precision of optical components, as well as factors such as the weight limit of the satellite payload, the structural strength of the entire system, and launch costs. Simply relying on hardware upgrades such as increasing the optical aperture and extending the focal length to improve spatial resolution would not only significantly increase the R&D, manufacturing, and launch costs of the satellite payload, but also introduce engineering challenges such as thermal deformation and optical distortion due to the increased complexity of the optical system. At the same time, it would place stringent requirements on the attitude stability of the satellite platform, making it difficult to meet the low-cost, lightweight deployment requirements of large-scale satellite constellations. Therefore, super-resolution reconstruction of remote sensing images through algorithmic means has become an important technological development direction for breaking through the physical limits of hardware and improving observation quality.

[0003] Currently, super-resolution reconstruction methods for remote sensing images are mainly divided into two categories. One category is deep learning-based reconstruction methods based on single-frame images. These methods rely on a large number of high- and low-resolution image sample pairs and use neural networks to learn the texture and structure mapping relationship between samples to restore the details of low-resolution images. The other category is super-resolution reconstruction methods based on multi-frame observations. The core logic of this method is to acquire multiple images of the same target area that have slight differences in space or pose. It uses inter-frame sub-pixel displacement information to complete the registration, interpolation, and fusion of multiple images, restoring higher-resolution ground feature details from real observation data. This method has a significant advantage in image texture authenticity.

[0004] In practical satellite remote sensing applications, both existing methods have significant limitations, with the core technical challenges of multi-frame super-resolution reconstruction methods being particularly prominent. Existing multi-frame super-resolution technologies typically rely on natural attitude jitter of the satellite platform, minor orbital deviations, or asynchronous observations from multiple satellites to obtain inter-frame displacements. These displacements exhibit significant randomness and uncontrollability, making precise quantification impossible and easily leading to the accumulation of registration errors across multiple frames, drastically reducing reconstruction accuracy. Furthermore, existing technologies lack a collaborative design between the satellite attitude control system and the super-resolution reconstruction algorithm, focusing solely on post-processing optimization of ground-based images. This results in the difficulty of consistently acquiring the precise sub-pixel displacement information theoretically required by the algorithm in actual on-orbit observations. Moreover, uncontrollable factors such as illumination variations, atmospheric disturbances, and ground object movement can cause photometric and geometric mismatches in multi-frame images, ultimately leading to artifacts and limited resolution improvement in the reconstructed images, failing to meet the practical needs of high-precision Earth observation. Summary of the Invention

[0005] In view of this, this application provides a method, apparatus and storage medium for multi-frame super-resolution imaging of optical remote sensing satellites, which can improve the accuracy of multi-frame super-resolution imaging of optical remote sensing satellites.

[0006] According to a first aspect of this application, a multi-frame super-resolution imaging method for an optical remote sensing satellite is provided, the method being applied to a ground system, comprising: Based on the requirements of Earth observation missions, satellite orbit parameters, and constellation configuration information, satellite attitude control commands containing preset sub-pixel level displacement strategies are generated. Receive multiple frames of ground target images acquired by the satellite according to the satellite attitude control commands, as well as auxiliary data synchronously transmitted back by the satellite; Based on the auxiliary data, sub-pixel-level geometric registration is performed on the multiple frames of ground target images, and photometric normalization is performed on the image differences under different observation conditions. Super-resolution fusion reconstruction is performed on the registered and normalized multi-frame ground target images to obtain optical remote sensing images.

[0007] According to a second aspect of this application, a multi-frame super-resolution imaging device for an optical remote sensing satellite is provided, the device being applied to a ground system, comprising: The generation module is used to generate satellite attitude control commands containing preset sub-pixel level displacement strategies based on the requirements of Earth observation missions, satellite orbit parameters and constellation configuration information. The receiving module is used to receive multiple frames of ground target images acquired by the satellite according to the satellite attitude control command, as well as auxiliary data synchronously transmitted back by the satellite; The registration module is used to perform sub-pixel-level geometric registration of the multi-frame ground target images based on the auxiliary data, and to perform photometric normalization processing for image differences under different observation conditions. The reconstruction module is used to perform super-resolution fusion reconstruction on registered and normalized multi-frame ground target images to obtain optical remote sensing images.

[0008] According to a third aspect of this application, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the aforementioned multi-frame super-resolution imaging method for optical remote sensing satellites.

[0009] According to a fourth aspect of this application, an electronic device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the aforementioned multi-frame super-resolution imaging method for optical remote sensing satellites.

[0010] By employing the aforementioned technical solutions, this application provides a multi-frame super-resolution imaging method, apparatus, and storage medium for optical remote sensing satellites. By generating attitude control commands containing preset sub-pixel-level displacement strategies, it deeply couples satellite attitude control with the requirements of the super-resolution algorithm. This eliminates reliance on random displacements such as natural satellite jitter and orbital deviations, providing precisely quantifiable sub-pixel displacement data for subsequent reconstruction. Simultaneously, it achieves high-resolution imaging without increasing the complexity of the optical system, contributing to lightweight and low-cost satellite payloads and reducing manufacturing, launch, and maintenance costs. Sub-pixel-level geometric registration and photometric normalization are performed based on auxiliary data transmitted from the satellite. This approach effectively corrects registration biases caused by terrain undulations, eliminates photometric and geometric mismatches in multi-frame images caused by factors such as illumination angle and atmospheric transparency, suppresses random noise, and significantly reduces the accumulation of registration errors. Super-resolution fusion reconstruction based on multi-frame real observation data after registration normalization avoids the illusionary drawbacks of single-frame deep learning super-resolution methods, ensuring the physical realism of ground textures. It also improves reconstruction accuracy by incorporating attitude control parameter constraints. Furthermore, this scheme is adaptable to multi-satellite collaborative observation scenarios, further ensuring observation consistency through temporal collaboration under constellation configuration, enhancing robustness and cross-scenario generalization capabilities in complex environments. Overall, this scheme achieves collaborative optimization of satellite on-orbit imaging and ground-based super-resolution algorithms, applicable to single-satellite platforms and scalable to multi-satellite constellations. The final output optical remote sensing image resolution is significantly improved compared to the original single-frame data, and surface positioning errors are controllable, meeting the practical application requirements of high-precision Earth observation.

[0011] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a multi-frame super-resolution imaging method for optical remote sensing satellites provided in an embodiment of this application is shown. Figure 2 A flowchart illustrating a multi-frame super-resolution imaging method for an optical remote sensing satellite according to another embodiment of this application is shown. Figure 3 This illustration shows an example of single-satellite multi-frame acquisition (including sub-pixel displacement, orthogonal stepping mode) provided in an embodiment of this application; Figure 4 This illustration shows an example of single-satellite multi-frame acquisition (including sub-pixel displacement, spiral scanning mode) provided in an embodiment of this application; Figure 5 This paper illustrates a schematic diagram of the principle process of multi-frame super-resolution imaging of an optical remote sensing satellite according to an embodiment of this application. Figure 6 A schematic diagram of the structure of a multi-frame super-resolution imaging device for an optical remote sensing satellite provided in an embodiment of this application is shown. Detailed Implementation

[0013] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0014] Currently, super-resolution reconstruction methods for remote sensing images are mainly divided into two categories. One category is deep learning-based reconstruction methods based on single-frame images. These methods rely on a large number of high- and low-resolution image sample pairs and use neural networks to learn the texture and structure mapping relationship between samples to restore the details of low-resolution images. The other category is super-resolution reconstruction methods based on multi-frame observations. The core logic of this method is to acquire multiple images of the same target area that have slight differences in space or pose. It uses inter-frame sub-pixel displacement information to complete the registration, interpolation, and fusion of multiple images, restoring higher-resolution ground feature details from real observation data. This method has a significant advantage in image texture authenticity.

[0015] In practical satellite remote sensing applications, both existing methods have significant limitations, with the core technical challenges of multi-frame super-resolution reconstruction methods being particularly prominent. Existing multi-frame super-resolution technologies typically rely on natural attitude jitter of the satellite platform, minor orbital deviations, or asynchronous observations from multiple satellites to obtain inter-frame displacements. These displacements exhibit significant randomness and uncontrollability, making precise quantification impossible and easily leading to the accumulation of registration errors across multiple frames, drastically reducing reconstruction accuracy. Furthermore, existing technologies lack a collaborative design between the satellite attitude control system and the super-resolution reconstruction algorithm, focusing solely on post-processing optimization of ground-based images. This results in the difficulty of consistently acquiring the precise sub-pixel displacement information theoretically required by the algorithm in actual on-orbit observations. Moreover, uncontrollable factors such as illumination variations, atmospheric disturbances, and ground object movement can cause photometric and geometric mismatches in multi-frame images, ultimately leading to artifacts and limited resolution improvement in the reconstructed images, failing to meet the practical needs of high-precision Earth observation.

[0016] To address the aforementioned technical problems, this invention provides a multi-frame super-resolution imaging method for optical remote sensing satellites. This method is applied to a ground system within a multi-frame super-resolution imaging system. The multi-frame super-resolution imaging system includes a satellite platform subsystem deployed on the satellite, an optical imaging payload, an attitude fine-tuning and displacement generation module, and a multi-satellite collaborative control module and a ground processing and reconstruction module deployed on the ground system. The satellite platform subsystem, serving as the core hardware carrier for active sub-pixel displacement, is responsible for receiving satellite attitude control commands from the ground system. It uses a reaction flywheel, magnetic torque converter, gyroscope, and star sensor to perform sub-pixel-level three-axis attitude adjustment and stabilization control, simultaneously acquiring and recording orbital parameters, attitude angle data, and high-precision timestamps, and compressing and buffering the relevant data for later transmission. The optical imaging payload, according to attitude control commands and a preset sub-pixel displacement strategy, controls the exposure sequence to acquire multiple frames of ground target images, ensuring at least three repeated imaging sessions during a single satellite transit and that adjacent frames generate sub-pixel-level projection displacement of 1 / 3 to 1 / 2 of the pixel projection size. It also adjusts exposure parameters to reduce the impact of illumination fluctuations on image radiometric consistency. The attitude fine-tuning and displacement generation module receives ground commands, analyzes sub-pixel displacement trajectory parameters, establishes geometric mapping relationships using a digital elevation model (DEM), calculates attitude adjustment amounts, and drives the satellite attitude control system to perform real-time fine-tuning, ensuring successful acquisition of multiple frames of ground target images. The system ensures that the target remains within the field of view and its displacement pattern is predictable and quantifiable. The multi-satellite collaborative control module calculates the satellite visibility window for the target area based on Earth observation mission requirements, satellite orbit parameters, and constellation configuration information. It then selects imaging periods that meet the lighting and viewing angle conditions, optimizes the multi-satellite collaborative observation sequence (e.g., controlling the observation interval between adjacent satellites to ≤2 minutes and the total time span to ≤10 minutes), and provides core parameters for generating the second satellite attitude control command. The ground processing and reconstruction module, as the core execution unit of the ground system, can first receive satellite downlink data and perform integrity verification, error detection, and time sequence reassembly. Then, it completes sub-pixel-level geometric registration through coordinate transformation, line-of-sight vector calculation, and DEM correction. It performs photometric normalization processing in conjunction with lighting and atmospheric parameters, and then uses a weighted iterative maximum a posteriori reconstruction algorithm for super-resolution fusion reconstruction. Finally, it calculates indicators such as SSIM and PSNR to evaluate the quality, selects qualified results, and feeds them back to the observation planning stage to optimize subsequent parameters, forming an observation-processing-optimization closed loop.

[0017] Correspondingly, such as Figure 1 As shown, the multi-frame super-resolution imaging method for optical remote sensing satellites includes: Step 110: Based on the requirements of the Earth observation mission, satellite orbit parameters and constellation configuration information, generate satellite attitude control commands that include preset sub-pixel level displacement strategies.

[0018] Among them, Earth observation mission requirements refer to the satellite imaging targets proposed by users or mission departments, including core parameters such as the geographical range of the ground target area, the desired spatial resolution, the allowable observation time window (e.g., avoiding cloud cover periods), and illumination requirements (e.g., solar altitude angle threshold); satellite orbital parameters are key geometric and dynamic parameters used to describe the satellite's on-orbit trajectory, including orbital inclination (the angle between the satellite's orbital plane and the Earth's equatorial plane), right ascension of the ascending node (the longitude of the orbital node intersecting the equator), orbital period (the time it takes for the satellite to orbit the Earth once), and orbital altitude (the average distance between the satellite and the Earth's surface), used to calculate the satellite's visibility and transit time of the ground target area; constellation configuration information is a parameter reflecting the overall deployment status of the satellite constellation, including the number of satellites in orbit within the constellation (usually ≥3 to ensure coordinated observation), the distribution of satellites in the orbital plane (e.g., the phase difference between adjacent orbits), the operational status of each satellite payload (e.g., whether the imaging sensor is available), and single... Satellite on-orbit priority and task allocation rules are used to optimize the timing and resource allocation of multi-satellite collaborative observations. Pre-set sub-pixel displacement strategies refer to controllable and quantifiable satellite attitude micro-adjustment schemes designed in advance to meet the needs of super-resolution reconstruction. By precisely controlling the satellite's pitch angle, yaw angle, and roll angle, ground targets are made to produce sub-pixel-level (usually 1 / 3 to 1 / 2 of the ground projection size of the imaging pixel) projection displacement in adjacent frames of imagery. Common modes include orthogonal stepping trajectories adapted to pushbroom / full-frame imaging and spiral scanning trajectories adapted only to full-frame imaging. Satellite attitude control commands are a set of commands generated by the ground system to drive satellite attitude adjustment and imaging tasks. They include specific parameters of the pre-set sub-pixel displacement strategy (such as displacement trajectory type and attitude adjustment amount), imaging timing (such as exposure interval), and the transit order and interval during multi-satellite collaborative observations. They can be divided into commands guiding independent acquisition by a single satellite and commands guiding multi-satellite collaborative acquisition (including collaborative timing).

[0019] In this embodiment, the ground system first receives the Earth observation mission requirements, defines the target area, desired resolution, and observation time constraints, and then uses satellite orbit parameters to calculate the visibility window of each satellite over the target area. Specifically, it can simulate satellite transit trajectories using parameters such as orbital inclination and orbital altitude to select imaging periods with suitable solar altitude angle and nadir angle. Simultaneously, it optimizes single-satellite observation priorities or multi-satellite collaborative transit intervals by combining constellation configuration information (such as the number of satellites in orbit and orbital plane distribution) to avoid image differences caused by ground object movement and illumination changes. Based on this, according to the satellite transit trajectory and satellite imaging mode (pushbroom or full-frame) within the imaging period, it designs a preset sub-pixel-level displacement strategy, determines the displacement trajectory type (orthogonal stepping or spiral scanning) and the corresponding three-axis attitude adjustment sequence, and finally integrates and encapsulates the preset sub-pixel-level displacement strategy to generate satellite attitude control commands that can drive the satellite attitude control system and imaging payload, which are then transmitted to the target satellite via the ground telemetry and control link.

[0020] This technical step deeply couples Earth observation mission requirements, satellite orbit parameters, constellation configuration information, and a pre-set sub-pixel-level displacement strategy, overcoming the limitations of existing multi-frame super-resolution technologies that rely on random displacement. The actively designed sub-pixel-level displacement can be precisely quantified, avoiding the accumulation of registration errors caused by natural jitter and orbital deviations. This provides stable and controllable geometric constraints for subsequent multi-frame image registration and super-resolution reconstruction. Simultaneously, by combining visibility analysis of orbit parameters with temporal optimization of constellation configuration, it ensures that the imaging period and collaborative observation logic are adapted to mission requirements, reducing the impact of illumination variations and atmospheric disturbances on image consistency. The generated satellite attitude control commands are directly linked to the super-resolution algorithm's requirements for sub-pixel displacement, enabling collaborative optimization of satellite attitude control and reconstruction algorithms. This supports high-resolution imaging without increasing the complexity of the optical system (contributing to lightweight satellite payloads and cost reduction), and ensures geometric and photometric consistency across multiple frames through controllable displacement. This lays a core foundation for generating super-resolution remote sensing images with low artifacts and high texture fidelity, effectively meeting the practical application needs of high-precision Earth observation.

[0021] Step 120: Receive multiple frames of ground target images collected by the satellite according to the satellite attitude control commands, as well as auxiliary data synchronously transmitted back by the satellite.

[0022] Among them, the multi-frame ground target imagery consists of multiple remote sensing images of the same target with controllable sub-pixel-level geometric differences, acquired by the satellite in accordance with attitude control commands when passing over the ground target area. Each satellite completes at least three repeated imaging operations during a single pass, and the ground target projection displacement between adjacent frames is 1 / 3 to 1 / 2 of the ground projection size of the imaging pixel. This data is the core data source for subsequent super-resolution fusion and reconstruction. The auxiliary data consists of multi-dimensional supplementary information recorded synchronously by the satellite when acquiring images, supporting the data processing at the ground end. This includes the satellite's three-axis attitude angles (pitch angle, yaw angle, roll angle), orbital parameters (ECEF position, velocity vector, and orbital altitude in the geocentric-ground-fixed coordinate system), nanosecond-level high-precision timestamps (to ensure data synchronization), environmental parameters (illuminance angle, solar azimuth angle, and atmospheric transparency), and digital elevation models, etc.

[0023] In specific application scenarios, after receiving satellite attitude control commands from the ground system, the satellite can first use its attitude control system (reaction flywheel, magnetic torque converter) to analyze the sub-pixel-level displacement strategy and attitude adjustment parameters in the commands, and fine-tune the three-axis attitude in real time to ensure that the lens is continuously aligned with the ground target area. Then, it drives the optical imaging payload to perform exposure acquisition according to the imaging sequence specified in the commands, and completes at least three repeated imaging sessions during a single pass to acquire multiple frames of ground target images with controllable sub-pixel-level differences. At the same time, the satellite synchronously activates the star sensor, gyroscope, and GNSS receiver during each imaging session. The satellite uses environmental sensors to collect auxiliary data such as attitude angles, orbital parameters, timestamps, illumination, and atmospheric parameters. Combined with a pre-set digital elevation model (DEM), multiple frames of images and auxiliary data are compressed and packaged according to mission number and stored in the on-board cache. After the satellite enters the coverage area of ​​the ground station, the packaged image data and auxiliary data are transmitted to the ground system through the satellite-to-ground communication link. After receiving the data, the ground system first performs integrity verification and error detection to eliminate data loss or corruption. Then, it completes the temporal association and reassembly of the images and auxiliary data according to the timestamps to prepare the data for subsequent processing.

[0024] This technical step, by receiving multiple frames of images acquired by the satellite according to preset instructions, ensures that the data source for subsequent super-resolution reconstruction has controllable and quantifiable sub-pixel-level geometric differences, avoiding the risk of inter-frame mismatch caused by traditional random displacement. The synchronously received auxiliary data can provide the ground end with accurate geometric references (attitude angles, orbital parameters, DEM) and radiometric calibration basis (illuminance, atmospheric parameters), effectively supporting the improvement of sub-pixel-level geometric registration accuracy and the consistency optimization of photometric normalization, and reducing processing errors caused by missing or asynchronous data.

[0025] Step 130: Perform sub-pixel-level geometric registration on multiple frames of ground target images based on auxiliary data, and perform photometric normalization processing on the image differences under different observation conditions.

[0026] Among them, sub-pixel-level geometric registration is a multi-frame image alignment technique with higher accuracy than pixel-level registration. Based on the geometric information (attitude angle, orbital parameters, DEM) in the auxiliary data, it calculates the difference in line-of-sight direction between frames and corrects terrain undulation deviations, so that the projection position deviation of the same ground object in multiple frames is controlled at the sub-pixel scale (usually less than 0.5 pixels), providing a precise geometric basis for subsequent fusion and reconstruction. Photometric normalization is used to address the problem of inconsistent brightness and contrast in multiple frames due to differences in observation conditions (such as changes in illumination angle and different atmospheric transparency). Based on the environmental parameters in the auxiliary data, it unifies the photometric characteristics of multiple frames through radiometric calibration, atmospheric correction and noise suppression, eliminating the interference of photometric differences on the fusion effect.

[0027] In this embodiment, the ground system can first preprocess the received auxiliary data to complete the time synchronization and alignment of multi-source data, and uniformly transform the satellite orbit coordinates to the geocentric-ground-fixed coordinate system (ECEF). Then, based on the preprocessed attitude angles and orbital parameters, the satellite line-of-sight vector of each frame image is calculated in combination with the WGS-84 Earth ellipsoid model. The vector is then transformed to the satellite local East-North-Sky (ENU) coordinate system and the inter-frame vector difference is solved. The geometric offset of the multi-frame image is quantified by the difference. Then, the registration deviation caused by terrain undulation is corrected by combining the DEM in the auxiliary data, and the sub-pixel level geometric registration of the multi-frame ground target image is completed. During the registration process, the illumination angle, solar azimuth angle and atmospheric transparency parameters in the auxiliary data are extracted simultaneously. Radiometric calibration is performed on the registered multi-frame image to eliminate the brightness deviation caused by atmospheric scattering. The illumination difference of different observation periods is balanced by block-by-block atmospheric correction. Finally, the random noise in the image is suppressed by weighted averaging or filtering strategy, and the photometric normalization processing of the multi-frame image is realized to ensure that the output registered image maintains a high degree of consistency in geometric position and photometric characteristics.

[0028] This technical step, through sub-pixel-level geometric registration, corrects terrain and inter-frame offset based on precise geometric information in auxiliary data, significantly improving the alignment accuracy of multi-frame images. It avoids reconstruction artifacts caused by the accumulation of geometric errors in traditional registration, providing a reliable spatial reference for super-resolution fusion. Simultaneously, photometric normalization processing, relying on environmental parameters in the auxiliary data, effectively eliminates inconsistencies in image photometric properties caused by differences in observation conditions such as illumination and atmosphere, unifying the radiometric characteristics of multi-frame images and reducing the interference of photometric differences on the restoration of fused details. The synergistic effect of these two steps enables multi-frame images to reach high-quality standards in both geometry and photometrics. This not only provides high-quality input data for subsequent super-resolution fusion reconstruction but also significantly improves the spatial resolution, texture realism, and radiometric consistency of the final reconstructed image, effectively meeting the requirements for image detail and reliability in high-precision Earth observation scenarios.

[0029] Step 140: Perform super-resolution fusion reconstruction on the registered and normalized multi-frame ground target images to obtain optical remote sensing images.

[0030] Super-resolution fusion reconstruction is a core processing technology that improves image resolution based on complementary observation information from multiple frames. It is used to mine different sub-pixel-level details in multiple frames of images (such as the edges and texture information of ground objects captured in different frames), combine spatial smoothing priors and satellite attitude control parameter constraints, and use pixel-level fusion algorithms to integrate effective information from multiple frames to restore high-resolution ground object structures from low-resolution input. Optical remote sensing images refer to the final output after super-resolution fusion reconstruction. Compared with the original single-frame images, their spatial resolution is significantly improved (usually at least twice the original resolution), and the true texture and spatial structure of ground objects are completely preserved. The surface projection positioning error is controllable and can be directly used for high-precision Earth observation applications such as target detection, ground object classification, and change recognition.

[0031] In the embodiments of this disclosure, the ground system can first calculate the inter-frame sub-pixel level displacement field based on the sub-pixel level registration results of the multi-frame images after registration and normalization. By using the spatial distribution characteristics of the displacement field, it can accurately identify the complementary detail distribution regions (such as the building corners and vegetation canopy textures that are clearly presented in some frames) and redundant noise regions of each frame image. Then, with the complementary detail distribution regions as the core, the weighted iterative maximum a posteriori reconstruction algorithm is used, combined with the satellite attitude control parameters and the spatial smoothing prior conditions, to perform pixel-level fusion of the multi-frame images. Specifically, complementary detail distribution areas can be assigned high fusion weights to highlight effective information, while redundant noise areas can be weighted to reduce interference, generating a high-resolution initial image. For detail deficiencies in complex terrain (such as mountains and canyons) or dynamically changing land features (such as flowing water bodies and temporary structures) in the initial image, adaptive weighting or block reconstruction strategies can be used for local fine-tuning to optimize the continuity of edge textures and structural integrity. Finally, based on preset quantitative indicators such as structural similarity (SSIM) and peak signal-to-noise ratio (PSNR), the reconstruction accuracy of the detail-corrected image is evaluated, and images that meet the preset resolution requirements are selected as the final optical remote sensing image output.

[0032] This technique, by deeply mining the complementary details of multiple frames of registered and normalized images, avoids the dependence on training data in traditional single-frame super-resolution reconstruction, fundamentally reducing the generation of illusory textures and ensuring the physical authenticity of ground textures. The weighted iterative maximum a posteriori reconstruction algorithm, combined with attitude control parameters and spatial smoothing priors, can effectively integrate sub-pixel-level details from multiple frames, significantly improving image spatial resolution, and suppress redundant noise and interference information through weight adjustment, ensuring image clarity. Local optimization strategies for complex terrain and dynamic areas can further compensate for the detail shortcomings of conventional fusion algorithms in special scenarios, ensuring the overall detail integrity and structural consistency of the reconstructed image. The final output optical remote sensing image not only has high resolution but also controllable surface projection positioning error, which can fully meet the needs of downstream applications such as target recognition and ground feature classification in high-precision Earth observation scenarios, effectively overcoming the technical limitations of limited accuracy and insufficient robustness in traditional multi-frame super-resolution reconstruction.

[0033] The multi-frame super-resolution imaging method for optical remote sensing satellites provided by this invention generates attitude control commands containing preset sub-pixel-level displacement strategies, deeply coupling satellite attitude control with the requirements of the super-resolution algorithm. This eliminates the dependence on random displacements such as natural satellite jitter and orbital deviations, providing precisely quantifiable sub-pixel displacement data for subsequent reconstruction. Simultaneously, it achieves high-resolution imaging without increasing the complexity of the optical system, contributing to lightweight and low-cost satellite payloads and reducing manufacturing, launch, and maintenance costs. Sub-pixel-level geometric registration and photometric normalization processing based on satellite-returned auxiliary data can effectively correct... This approach addresses registration bias caused by surface undulations, eliminates the issue of light and geometric mismatch in multi-frame images due to factors such as illumination angle and atmospheric transparency, suppresses random noise, and significantly reduces the accumulation of registration errors. Super-resolution fusion reconstruction based on multi-frame real observation data after registration normalization avoids the illusionary drawbacks of single-frame deep learning super-resolution methods, ensuring the physical realism of ground textures. Furthermore, it improves reconstruction accuracy by incorporating attitude control parameter constraints. This scheme is also adaptable to multi-satellite collaborative observation scenarios, further ensuring observation consistency through temporal collaboration under constellation configuration, enhancing robustness and cross-scenario generalization capabilities in complex environments. Overall, this scheme achieves collaborative optimization of satellite on-orbit imaging and ground-based super-resolution algorithms, applicable to single-satellite platforms and scalable to multi-satellite constellations. The final output optical remote sensing image resolution is significantly improved compared to the original single-frame data, with controllable surface positioning errors, meeting the practical application requirements of high-precision Earth observation.

[0034] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, and to fully illustrate the implementation of this embodiment, this embodiment also provides another multi-frame super-resolution imaging method for optical remote sensing satellites, such as... Figure 2 As shown, the method includes: Step 210: Based on the requirements of the Earth observation mission, satellite orbit parameters and constellation configuration information, generate satellite attitude control commands that include preset sub-pixel level displacement strategies.

[0035] The satellite attitude control commands include a first satellite attitude control command and a second satellite attitude control command. The first satellite attitude control command is used to instruct one or more satellites to independently acquire multiple frames of ground target images, and the second satellite attitude control command is used to instruct multiple satellites to collaboratively acquire multiple frames of ground target images according to the multi-satellite collaborative observation sequence.

[0036] Accordingly, for embodiments of this disclosure, the steps may include: Step 210-1: Based on the requirements of Earth observation missions, satellite orbital parameters and constellation configuration information, calculate the visibility window of each satellite over the ground target area, and select imaging periods that meet the illumination conditions and nadir point viewing angle conditions.

[0037] The visibility window refers to the specific time period during which a ground target area can enter the satellite's imaging field of view while one or more satellites are in orbit. In other words, it is the time interval during which the satellite has the capability to perform imaging operations on the target area. Its calculation depends on the satellite's orbital parameters and the geographic coordinates of the target area. The illumination condition is the solar illumination constraint that ensures the quality of image imaging. It is usually based on the solar altitude angle (e.g., solar altitude angle ≥ 15°) to ensure that the ground target is bright enough and the shadow coverage is small during imaging, so as to avoid the loss of image details due to insufficient illumination or strong light shadows. The nadir point viewing angle condition is the viewing angle deviation constraint between the satellite's nadir point (the intersection of the line connecting the satellite and the Earth's center on the Earth's surface) and the ground target area (e.g., viewing angle deviation ≤ 5°). It is used to reduce the geometric distortion of the image caused by excessive satellite observation angle and ensure the geometric consistency of multiple frames of images.

[0038] In this embodiment, the ground system first receives and parses the Earth observation mission requirements, clarifying the latitude and longitude range, resolution requirements, and time constraints of the target area. Then, it calls upon the orbital parameters of each satellite, simulates the on-orbit trajectory of each satellite using an orbital dynamics model, and calculates the visibility window of each satellite over the target area (i.e., the specific time period during which the satellite's field of view covers the target area) by combining the geographical coordinates of the target area. Simultaneously, it refers to constellation configuration information to exclude satellites with payload failures or low mission priority, selecting a list of satellites with observation capabilities. Based on this list, it further filters the visibility windows of each satellite, such as calculating the solar altitude angle within the window time period using a solar radiation model, retaining time segments that meet the illumination conditions. Then, it calculates the angular deviation between the satellite's nadir and the target area within each time segment using geometric modeling, eliminating time periods with excessive angular deviation. Finally, it integrates the filtering results to determine the imaging time period that is both within the satellite's visibility window and meets the illumination and nadir angular requirements, providing a time reference for subsequent design of sub-pixel-level displacement strategies and generation of satellite attitude control commands.

[0039] This technical step calculates the visibility window by integrating Earth observation mission requirements, satellite orbit parameters, and constellation configuration information. This ensures that participating satellites can accurately cover the target area, avoiding observation failures due to mismatch between satellite trajectories and targets. Simultaneously, by filtering illumination conditions and nadir point viewing angles, it effectively avoids factors affecting image quality such as insufficient illumination and geometric distortion, ensuring that subsequent multi-frame images have a good brightness foundation and geometric consistency. Furthermore, this process clarifies the optimal imaging time period, providing precise timing data for designing controllable sub-pixel-level displacement strategies and planning multi-satellite collaborative observation sequences. This reduces multi-frame image registration errors and photometric mismatches caused by unreasonable imaging time periods, laying a crucial foundation for the final high-definition, low-distortion optical remote sensing image output through super-resolution fusion reconstruction.

[0040] Step 210-2: Design a satellite three-axis attitude adjustment sequence based on the imaging period and satellite imaging mode, select the sub-pixel displacement trajectory of orthogonal stepping or spiral scanning, and form a preset sub-pixel level displacement strategy.

[0041] The orthogonal stepping mode involves alternating fine-tuning the attitude along the X and Y directions of the image plane, with each displacement being approximately 1 / 3 to 1 / 2 of the pixel size (suitable for pushbroom imaging and full-frame imaging). The spiral scanning mode rotates the optical axis at a fixed angular velocity during continuous exposure to achieve annular sub-pixel sampling (only applicable to full-frame imaging). All of these micro-displacements are achieved collaboratively by a satellite reaction flywheel and a magnetic torque converter, with the attitude adjustment typically controlled within tens of arcseconds to ensure the stability of the optical system.

[0042] For embodiments of this disclosure, the steps may include: determining the ground projection size of the satellite imaging pixels based on satellite orbit parameters; setting the sub-pixel displacement amplitude to 1 / 3 to 1 / 2 of the ground projection size; determining the initial geodetic coordinates of the satellite and the reference geodetic coordinates of the ground target area based on the satellite's transit trajectory during the imaging period; calculating the radius of curvature of the prime mover corresponding to the initial geodetic coordinates based on the semi-major axis and flattening parameters of the WGS-84 Earth ellipsoid model; transforming the geodetic coordinates of the satellite and the ground target area to the geocentric-Earth-fixed coordinate system based on the radius of curvature of the prime mover and calculating the coordinate difference; calculating the original line-of-sight vector through the coordinate difference and completing the normalization process to obtain the standard line-of-sight direction vector; establishing a transformation matrix from the geocentric-Earth-fixed coordinate system to the ENU coordinate system; and transforming the standard line-of-sight direction vector based on the transformation matrix. The line direction vector is transformed to the ENU coordinate system, and the initial azimuth and initial elevation angles required for satellite imaging are calculated. The sub-pixel displacement trajectory and roll angle control parameters are determined according to the satellite imaging mode, and the coordinate correction amount of the ground target area is generated for the sub-pixel displacement trajectory. The corrected geodetic coordinates of the ground target area are determined based on the coordinate correction amount, and the corrected azimuth and corrected elevation angles are determined based on the corrected geodetic coordinates. The azimuth adjustment amount is obtained by the difference between the initial azimuth and the corrected azimuth, and the elevation adjustment amount is obtained by the difference between the initial elevation and the corrected elevation. Based on the azimuth adjustment amount, elevation adjustment amount, and roll angle control parameters, a satellite three-axis attitude adjustment sequence is generated. The sub-pixel displacement trajectory and the satellite three-axis attitude adjustment sequence are integrated to form a preset sub-pixel level displacement strategy.

[0043] Among them, the ground projection size of a satellite imaging pixel refers to the actual geographical area (usually in meters) corresponding to a single pixel of the satellite imaging sensor on the ground. It is determined by orbital parameters such as satellite orbital altitude and optical system focal length, and is the core benchmark for determining sub-pixel displacement amplitude. Sub-pixel displacement amplitude refers to the small offset of the ground target projection position in adjacent frames of imagery designed to achieve super-resolution reconstruction, and is set to 1 / 3 to 1 / 2 of the pixel ground projection size. Initial geodetic coordinates are the geodetic coordinates when the satellite just enters the observation range of the target area during the imaging period. Reference geodetic coordinates are the standard geodetic coordinates of the ground target area, used to establish the initial geometric mapping relationship between the satellite and the target. WGS-84 Earth ellipsoid model. This is a globally applicable geodetic benchmark model used to define parameters such as the semi-major axis (a=6378137.0m) and flattening (f=1 / 298.257) of the Earth ellipsoid, providing a unified ellipsoidal benchmark for coordinate transformation and geometric modeling. The radius of curvature (N) of the prime meridian is the radius of curvature of a point on the WGS-84 ellipsoid along the prime meridian (the intersection of the section passing through that point and perpendicular to the meridian with the ellipsoidal surface), used for calculations to transform geodetic coordinates to the Earth-Centered Earth-Fixed (ECEF) coordinate system. The ECEF is a three-dimensional Cartesian coordinate system with the Earth's center of mass as the origin, the X-axis pointing to the intersection of the prime meridian and the equator, and the Z-axis pointing to the Earth's North Pole, facilitating spatial vector operations and global coordinate unification. The original line-of-sight vector... The vector difference between the satellite's ECEF coordinates and the ground target's ECEF coordinates reflects the original line-of-sight direction of the satellite pointing towards the target. The standard line-of-sight direction vector is a vector (magnitude of 1) normalized from the original line-of-sight vector, used to unify the quantization benchmark of the line-of-sight direction and adapt to subsequent attitude angle calculations. The ENU coordinate system (East-North-Sky coordinate system) is a local coordinate system of the satellite with its location as the origin, the X-axis pointing to local east, the Y-axis pointing to local north, and the Z-axis pointing to local zenith. The transformation matrix is ​​the matrix that realizes the vector transformation between the ECEF coordinate system and the ENU coordinate system. The initial azimuth / initial elevation angles are the initial imaging attitude angles of the satellite calculated based on the components of the standard line-of-sight direction vector in the ENU coordinate system. Azimuth refers to the clockwise angle between the line of sight and the north direction, and elevation refers to the angle between the line of sight and the horizon; satellite imaging mode refers to the way the satellite acquires images, including pushbroom imaging (continuously pushing along the orbital direction to acquire strip images) and full-frame imaging (acquiring the entire target area at once), which determines the sub-pixel displacement trajectory type; sub-pixel displacement trajectory is the path mode for satellite attitude adjustment, including orthogonal step trajectories adapted to pushbroom / full-frame and spiral scan trajectories adapted only to full-frame, used to generate controllable inter-frame sub-pixel differences; roll angle control parameter is the rotation angle control value of the satellite around the line of sight axis, which is usually set to a fixed value for pushbroom imaging (to maintain strip consistency), and can be finely adjusted for full-frame imaging (to adapt to the trajectory); coordinate correction amount ( , The first value represents the small correction value of the ground target geodetic coordinates designed based on the sub-pixel displacement trajectory, used to calculate the satellite attitude adjustment amount in reverse; the second value represents the satellite attitude angle recalculated based on the ground target geodetic coordinates after correction, and the difference between the second and third values ​​is the attitude adjustment amount; the third value represents the difference between the initial attitude angle and the corrected attitude angle, reflecting the specific values ​​of azimuth and pitch angles that the satellite needs to adjust to achieve sub-pixel displacement; the fourth value represents the three-axis attitude adjustment sequence, which is a sequence of instructions formed by sorting the azimuth adjustment amount, pitch adjustment amount and roll angle control parameters according to the execution time sequence of the satellite attitude control system, used to guide the satellite to adjust its attitude in real time.

[0044] For the embodiments of this disclosure, the ground projection size of the satellite imaging pixels can first be determined based on satellite orbital parameters (such as orbital altitude and optical focal length), the sub-pixel displacement amplitude can be set to 1 / 3 to 1 / 2 of the projection size, and then the initial geodetic coordinates of the satellite can be determined by combining the satellite's transit trajectory during the imaging period. The reference geodetic coordinates of the ground target area (λ, h) are then calculated; next, based on the semi-major axis (a=6378137.0m) and flattening (f=1 / 298.257) of the WGS-84 Earth ellipsoid model, the square of the eccentricity is first calculated ( Where e is the eccentricity of the WGS-84 ellipsoid, and then the formula for the radius of curvature of the prime meridian is used. (Parameter meanings: N is the radius of curvature of the prime meridian, a is the semi-major axis of the WGS-84 ellipsoid, e is the eccentricity.) Calculate N corresponding to the initial geodetic coordinates of the satellite (where N is the latitude of the initial geodetic coordinates of the satellite). Then, based on N, transform the geodetic coordinates of the satellite and the ground target to the ECEF coordinate system. The transformation formula is as follows: , , (Parameter meanings: x, y, z are ECEF coordinates, N is the radius of curvature of the prime mover, and h is the geodetic elevation of the satellite / ground target.) Latitude (For longitude), and calculate the difference between the two ECEF coordinates to obtain the original line-of-sight vector. (Parameter meaning:) The original line-of-sight vector. Let ECEF be the coordinate vector of the ground target. (For satellite ECEF coordinate vectors), The standard line-of-sight vector is obtained by normalization. (Parameter meaning:) The standard line-of-sight vector. Original line-of-sight vector The modulus length); then based on the initial geodetic coordinates of the satellite ( , Establish the transformation matrix from the ECEF coordinate system to the ENU coordinate system. (Parameter meaning:) The ENU transformation matrix is... For satellite longitude, (for satellite latitude), through Standard line-of-sight vector Transform to the ENU coordinate system to obtain the line of sight in the east ( ),north( ),sky( The directional components are then determined using the azimuth formula. (Parameter meaning: Yaw is the initial azimuth angle,) The eastern component of the line of sight. (Northward component of line of sight), pitch angle formula (Parameter meaning: Pitch is the initial pitch angle,) The initial azimuth and initial elevation angles required for satellite imaging are calculated (for the upward component of the line of sight); then, based on the satellite imaging mode (pushbroom / full-frame), the sub-pixel displacement trajectory (orthogonal stepping / helical scanning) and roll angle control parameters are determined (set to a fixed value for pushbroom and fine-tunable for full-frame), and the coordinate correction amount for the ground target area is generated for this trajectory. , ); Determine the corrected geodetic coordinates of the ground target based on the coordinate correction amount. , Repeat the aforementioned ECEF coordinate transformation, line-of-sight vector calculation, ENU transformation, and attitude angle solution process to obtain the corrected azimuth angle (Yaw') and corrected pitch angle (Pitch'). Then, use the difference between the initial azimuth angle and the corrected azimuth angle (h) to calculate the corrected azimuth angle. The azimuth adjustment is obtained by using the difference between the initial pitch angle and the corrected pitch angle. The pitch angle adjustment amount is obtained; finally, based on the azimuth angle adjustment amount, pitch angle adjustment amount and roll angle control parameters, the satellite three-axis attitude adjustment sequence is generated according to the execution time sequence of the satellite attitude control system, and the sub-pixel displacement trajectory is integrated with the sequence to form a preset sub-pixel level displacement strategy.

[0045] This technical step, through precise geometric modeling based on the WGS-84 ellipsoidal model and the ECEF coordinate system, combined with attitude angle calculation in the ENU coordinate system, enables the quantitative solution of the satellite's initial and corrected attitudes. This ensures the controllability of sub-pixel displacement amplitude (1 / 3 to 1 / 2 of the pixel projection size) and trajectory (orthogonal stepping / spiral scanning), avoiding registration errors caused by traditional random displacement. Simultaneously, by deriving attitude adjustment amounts inversely through coordinate correction and integrating the three-axis attitude adjustment sequence, the satellite attitude adjustment can accurately match the sub-pixel displacement requirements of super-resolution reconstruction, ensuring stable and quantifiable geometric differences between multiple image frames. Furthermore, designing differentiated roll angle control parameters and displacement trajectories for different imaging modes further enhances the adaptability and flexibility of attitude adjustment. Overall, the preset sub-pixel level displacement strategy formed in this step provides a core attitude reference for subsequent acquisition of multiple high-quality images by satellites. It can effectively reduce image geometric mismatch caused by attitude errors and lay a key technical foundation for sub-pixel level registration and super-resolution fusion reconstruction to output high-fidelity, high-resolution optical remote sensing images.

[0046] Step 210-3a: If a single-satellite independent acquisition task is performed, the preset sub-pixel level displacement strategy is encapsulated into the first satellite attitude control command by combining the on-orbit priority and task allocation status of the single satellite in the constellation configuration information.

[0047] Among them, the single-satellite independent acquisition mission refers to the mission mode in which a single satellite completes the acquisition of multiple frames of images of a ground target area independently, without the cooperation of other satellites. The core is to achieve preset sub-pixel level displacement through the attitude control of the single satellite itself, and acquire multiple frames of images that meet the requirements of super-resolution reconstruction. The on-orbit priority of a single satellite in the constellation configuration information is the mission execution order rule set for each satellite in the satellite constellation system. It is used to determine whether a single satellite has the authority to execute the current Earth observation mission first when multiple missions are concurrent, so as to avoid conflicts of satellite resources within the constellation. The mission allocation status of a single satellite is the current mission load and resource availability of a single satellite, including whether it has undertaken other observation missions, whether the imaging payload is working properly, and whether the onboard energy meets the mission requirements. It is used to determine whether the satellite has the conditions to execute the current single-satellite independent acquisition mission. The first satellite attitude control command is a satellite control command specifically used to guide the single-satellite independent acquisition mission. It encapsulates the preset sub-pixel level displacement strategy and the exclusive parameters required for the single satellite to execute the mission (such as exposure interval, data downlink trigger conditions), which can directly drive the single satellite to complete attitude adjustment and multi-frame image acquisition.

[0048] In this embodiment of the disclosure, when the ground system determines that a single-satellite independent acquisition task needs to be performed, the constellation configuration information can be retrieved first to extract the on-orbit priority and current task allocation status of the target satellite. The on-orbit priority is used to determine whether the satellite has higher priority than other non-urgent tasks and is authorized to perform the current observation task first. The task allocation status confirms that the satellite has no other task conflicts, its imaging payload is normal, and its onboard power is sufficient. If both conditions are met, the pre-designed sub-pixel-level displacement strategy (including sub-pixel displacement trajectory, three-axis attitude adjustment, number of imaging attempts, and timing) designed for this task is encapsulated according to the satellite command communication protocol format. Simultaneously, single-satellite independent acquisition-specific parameters (such as single exposure duration, image data buffering, and downlink trigger thresholds) are supplemented. After encapsulation, the instructions are checked for completeness and logic to ensure that the attitude adjustment parameters, imaging timing, and single-satellite hardware capabilities match. After successful verification, the first satellite attitude control command is generated and sent to the target satellite via the ground telemetry and control link to guide it in performing subsequent multi-frame image acquisition operations.

[0049] By combining the on-orbit priority of individual satellites, the execution of high-value Earth observation missions can be prioritized, avoiding observation delays caused by low-priority missions vying for resources and improving the rationality of constellation mission scheduling. By combining the mission allocation status of individual satellites, satellites with mission conflicts or payload malfunctions can be eliminated in advance, avoiding mission failures caused by the inability to execute commands after they are issued, and reducing resource waste. By encapsulating the preset sub-pixel-level displacement strategy into a dedicated first satellite attitude control command, it is possible to ensure that individual satellites accurately acquire the core parameters of attitude adjustment and imaging, avoiding attitude control deviations caused by missing parameters or incorrect formats. This ensures that the multi-frame images acquired by a single satellite have controllable and quantifiable sub-pixel-level displacements, reducing image geometric mismatch caused by attitude errors, providing a high-quality data source for subsequent sub-pixel-level registration and super-resolution fusion reconstruction, and ultimately helping to output high-resolution, low-artifact optical remote sensing images to meet the needs of high-precision Earth observation.

[0050] Step 210-3b: If a multi-satellite collaborative acquisition task is to be performed, the observation interval of adjacent satellites for the same target and the observation time span of all satellites are controlled according to the orbital phase difference and transit time difference of each satellite in the constellation configuration information, so as to form a multi-satellite collaborative observation sequence. The preset sub-pixel level displacement strategy and the multi-satellite collaborative observation sequence are encapsulated into the second satellite attitude control command.

[0051] Among them, the multi-satellite collaborative acquisition mission refers to a mission mode in which multiple satellites (preferably 3 or more, with 5 or more for optimal results) within a constellation jointly perform multi-frame image acquisition of the same ground target area according to a preset time sequence. The core is to obtain more multi-dimensional and complementary sub-pixel-level observation data through multi-satellite time and attitude coordination, providing richer data sources for super-resolution reconstruction. The orbital phase difference of each satellite in the constellation configuration information represents the angular position difference (expressed in radians or angles) of the satellites along their orbits within the same orbital plane or between adjacent orbital planes, reflecting the spatial correlation of the satellites' on-orbit distribution. The transit time difference of each satellite in the constellation configuration information is the time interval between different satellites entering the ground target area observation range (i.e., transiting the target area), determined by the satellite's orbital altitude and orbital phase difference. The observation interval of adjacent satellites on the same target is the interval between adjacent satellites participating in the collaborative acquisition performing observations on the same ground target. The time difference for imaging operations needs to be controlled within a short range (e.g., ≤2 minutes) to reduce the impact of ground movement and lighting changes on image consistency. The observation time span of all satellites refers to the total time from the first satellite starting to acquire images of the target area to the last satellite completing the acquisition, and it needs to be controlled within a limited range (e.g., ≤10 minutes) to avoid problems such as atmospheric disturbance and changes in ground condition caused by long-term observation. The multi-satellite collaborative observation sequence is a schedule for each satellite to perform acquisition tasks based on orbital phase difference and transit time difference, which clarifies the imaging start time, duration, and inter-frame interval of each satellite to ensure that the multi-satellite acquisition logic is orderly and the data is complementary. The second satellite attitude control command is a control command specifically used to guide the multi-satellite collaborative acquisition task. It encapsulates the preset sub-pixel level displacement strategy and the multi-satellite collaborative observation sequence, and can provide precise basis for attitude adjustment and imaging operations for each satellite participating in the collaboration.

[0052] In this embodiment of the present disclosure, when the ground system determines to perform a multi-satellite collaborative acquisition task, it first retrieves the constellation configuration information, extracts the orbital phase difference and transit time difference of each participating satellite, and calculates the precise time when each satellite arrives over the ground target area through the orbital dynamics model. Based on this time data, it controls the observation interval of adjacent satellites for the same target (ensuring the interval is short enough to reduce the impact of changes in the external environment), while constraining the observation time span of all satellites (ensuring the total duration is within a controllable range to ensure the stability of ground objects and illumination). Based on this, a multi-satellite collaborative observation sequence with a clear imaging order, start time, and inter-frame interval for each satellite is formulated. Subsequently, the preset sub-pixel level displacement strategy designed for each satellite in the early stage is integrated with the multi-satellite collaborative observation sequence and encapsulated into a unified second satellite attitude control command according to the satellite-to-ground communication protocol format. During the encapsulation process, it is necessary to verify whether the attitude parameters and imaging sequence of each satellite in the command match its real-time orbital state and payload capacity to avoid parameter conflicts. After the verification is passed, the second satellite attitude control command is synchronously sent to all participating satellites through the ground telemetry and control link to guide each satellite to perform multi-frame image acquisition according to the preset timing and attitude parameters of the command.

[0053] By controlling the observation interval and time span based on orbital phase difference and transit time difference, the impact of ground object movement, illumination angle changes, and atmospheric disturbances on the consistency of multi-frame images can be effectively reduced, ensuring high compatibility of images acquired by different satellites in terms of geometric and radiometric characteristics. Encapsulating the preset sub-pixel-level displacement strategy and multi-satellite collaborative observation sequence into a second satellite attitude control command ensures that all participating satellites acquire unified and accurate operating parameters, avoiding collaborative disorder caused by parameter inconsistencies. At the same time, it provides each satellite with a sub-pixel displacement scheme adapted to its own imaging mode, ensuring that images acquired by multiple satellites have controllable and quantifiable sub-pixel-level differences. Overall, this technical step, through multi-satellite temporal collaboration and unified parameter management, can provide richer and more consistent data sources for subsequent sub-pixel-level geometric registration and super-resolution fusion reconstruction, significantly improving the spatial resolution, texture realism, and radiometric consistency of the final optical remote sensing images, effectively meeting the demand for large-scale, high-detail images in high-precision Earth observation scenarios.

[0054] It should be noted that the ground system can control multiple satellites to perform the above-mentioned single-satellite independent acquisition tasks (i.e., only steps 210-3a need to be performed) according to actual working needs; or it can control multiple satellites to perform multi-satellite collaborative acquisition tasks (i.e., only steps 210-3b need to be performed); or it can control multiple satellites to perform the above-mentioned single-satellite independent acquisition tasks while also controlling multiple satellites to perform multi-satellite collaborative acquisition tasks (i.e., both steps 210-3a and 210-3b need to be performed simultaneously). No specific limitations are made here.

[0055] Step 220: Receive multiple frames of ground target images collected by the satellite according to the satellite attitude control commands, as well as auxiliary data synchronously transmitted back by the satellite.

[0056] In embodiments of this disclosure, the steps may include: receiving multiple frames of ground target images acquired by a satellite according to a first satellite attitude control command, as well as auxiliary data synchronously transmitted back by the satellite; and / or receiving multiple frames of ground target images acquired by multiple satellites according to a second satellite attitude control command, as well as auxiliary data synchronously transmitted back.

[0057] In specific application scenarios, when a satellite, in response to the first satellite attitude control command, acquires multiple frames of ground target imagery, it uses this imagery to establish a geometric mapping relationship with the ground target area based on a digital elevation model. Based on this geometric mapping relationship and a preset sub-pixel-level displacement strategy, the satellite passes over the ground target area. During this pass, the satellite's attitude is adjusted in real time via a reaction flywheel or other attitude control system, causing the ground target area to undergo sub-pixel-level projection displacement in adjacent image frames in orthogonal directions (e.g., alternating X-axis and Y-axis). Figure 3 , 4 (Different color frames correspond to the tiny step offsets of ground targets), and at the same time, multiple frames of ground target images are continuously acquired according to the command sequence to ensure that each frame of image has controllable and quantifiable sub-pixel level geometric differences.

[0058] In specific application scenarios, when multiple satellites acquire multiple frames of ground target images in response to the attitude control command of the second satellite, they are used to establish a geometric mapping relationship with the ground target area by combining the digital elevation model. Based on the geometric mapping relationship and the preset sub-pixel level displacement strategy, the satellites pass over the ground target area in sequence according to the multi-satellite collaborative observation time within a preset time period. During the passage, the satellite attitude is adjusted in real time through the reaction flywheel or other attitude control system to ensure that the images acquired by the multiple satellites are consistent in both geometric and radiometric aspects, and that sub-pixel level projection displacement is generated between adjacent frames acquired by each satellite. Among multiple satellites, the satellite that passes overhead first initiates imaging first, while the satellites that pass overhead later follow in sequence to ensure short observation intervals between adjacent satellites and controllable total observation span. During the transit of each satellite, the attitude control system, with the reaction flywheel as its core, analyzes the preset sub-pixel displacement strategy in the command in real time and dynamically fine-tunes the three-axis attitude: on the one hand, through precise attitude control, it causes sub-pixel-level projection displacement of the ground targets in adjacent frames acquired by itself; on the other hand, through unified attitude parameter calibration, it ensures that the images acquired by multiple satellites have small deviations in spatial position and uniform characteristics in brightness and contrast, and finally completes the collaborative acquisition of multiple frames of ground target images, forming a data source that is both complementary and consistent.

[0059] Step 230: Preprocess the satellite attitude angle, satellite orbit parameters and timestamps in the auxiliary data to complete the time synchronization and alignment of multi-source data, and uniformly transform the satellite orbit coordinates to the geocentric coordinate system.

[0060] Among them, satellite attitude angles refer to the three-dimensional angular parameters describing the satellite's attitude in orbit, including pitch angle, yaw angle, and roll angle, which are usually collected by high-precision star sensors and gyroscopes; satellite orbital parameters are geometric and dynamic parameters used to characterize the satellite's orbital trajectory, including orbital inclination, right ascension of the ascending node, orbital period, and orbital altitude, used to calculate the satellite's spatial position and transit trajectory; timestamps refer to the high-precision time information marked by the satellite for each set of attitude angles, orbital parameters, and image data, generated based on the onboard GPS clock or atomic clock; time synchronization and alignment of multi-source data refers to using timestamps as a reference, through interpolation compensation or delay calibration, to eliminate time deviations in the acquisition of data from different satellite sensors, ensuring that image data at a certain moment can accurately match the attitude angles and orbital parameters at the corresponding moment; the geocentric-fixed coordinate system (ECEF) refers to a three-dimensional Cartesian coordinate system with the Earth's center of mass as the origin, the X-axis pointing to the intersection of the Prime Meridian and the equator, the Y-axis pointing to the intersection of 90° East longitude and the equator, and the Z-axis pointing to the Earth's North Pole.

[0061] In this embodiment of the disclosure, after the ground system receives the auxiliary data transmitted from the satellite, it first performs preprocessing on the satellite attitude angle, satellite orbital parameters, and timestamp: outlier detection and smoothing filtering are performed on the attitude angle data, integrity verification and format standardization are performed on the orbital parameters, and format conversion is performed on the timestamp; then, using the high-precision timestamp as the core reference, the acquisition delay of different sensors is compensated by time interpolation algorithm, and the satellite attitude angle and satellite orbital parameters at the same observation time are time-synchronized and aligned with the corresponding multiple frames of ground target images, ensuring that each frame of image can be accurately associated with the attitude and orbital state at the moment of imaging; finally, based on the WGS-84 Earth ellipsoid model, the original satellite orbital coordinates are uniformly converted to the ECEF coordinate system using the conversion formula from geodetic coordinates to the geocentric Earth-fixed coordinate system, and the three-dimensional coordinates (x, y, z) of the satellite in this coordinate system are output, providing a unified spatial reference for subsequent calculation of the satellite line-of-sight vector and solving the image geometric offset.

[0062] Preprocessing of auxiliary data can effectively filter noise and outliers in attitude angles and orbital parameters, ensuring the reliability of basic data and avoiding interference from invalid data in subsequent calculations. Time synchronization and alignment of multi-source data can eliminate the misalignment of image, attitude, and orbital data caused by sensor acquisition delays, ensuring that each frame of image can be matched with accurate instantaneous attitude and orbital parameters, providing a temporal dimension guarantee for improving the accuracy of subsequent geometric registration. The unified transformation of satellite orbital coordinates to the geocentric-fixed coordinate system can solve the problem of benchmark differences in spatial position calculation under different coordinate systems, providing a unified spatial reference for subsequent operations such as satellite line-of-sight vector calculation and geometric offset solution, greatly reducing geometric errors caused by inconsistencies in coordinate systems. Overall, this lays a solid data foundation for achieving high-precision sub-pixel-level geometric registration, thereby ensuring the effectiveness of subsequent photometric normalization and super-resolution fusion reconstruction, and ultimately improving the spatial resolution and positioning accuracy of optical remote sensing images.

[0063] Step 240: Based on the preprocessed satellite attitude angles and satellite orbit parameters, and combined with the WGS-84 Earth ellipsoid model, calculate the satellite line-of-sight direction vector corresponding to each frame of ground target image, convert the satellite line-of-sight direction vector to the ENU coordinate system and calculate the vector difference, calculate the geometric offset of multiple frames of ground target image through the vector difference, and use the digital elevation model to correct the registration deviation caused by terrain undulation based on the geometric offset to complete the sub-pixel level geometric registration.

[0064] Among them, vector difference is the difference between the satellite line-of-sight vectors corresponding to different frames of imagery, which quantitatively reflects the line-of-sight deviation caused by satellite attitude adjustment or orbital motion in adjacent frames; geometric offset is the deviation of the projection position of the same ground feature on the image plane in multiple frames of imagery, calculated based on the line-of-sight vector difference, combined with satellite orbital altitude, satellite attitude angle, and optical system parameters (focal length, pixel size), which directly clarifies the degree of spatial offset that needs to be aligned between frames; digital elevation model (DEM) is a raster data model describing the terrain undulation of the ground target area, which can provide elevation information at different locations to correct the influence of terrain height differences on satellite line-of-sight projection and eliminate registration deviation caused by terrain factors; registration deviation caused by terrain undulation refers to the additional offset of the projection position of the same ground feature by the satellite line of sight in different frames due to uneven ground (such as the projection difference between mountainous and plain areas), which will lead to a decrease in registration accuracy if not corrected.

[0065] In this embodiment, the ground system first calls the preprocessed satellite attitude angles and orbital parameters, and combines them with the WGS-84 Earth ellipsoid model to transform the geodetic coordinates of the satellite and the ground target to the geocentric Earth-fixed (ECEF) coordinate system. The satellite line-of-sight vector corresponding to each frame of image is calculated by the difference between the two coordinates. Then, based on the geodetic coordinates at the time of satellite imaging, a transformation matrix from the ECEF coordinate system to the ENU coordinate system is established, and the line-of-sight vectors of all frames are uniformly transformed to the ENU coordinate system. The difference between the line-of-sight vectors of adjacent frames is calculated to obtain the vector difference. Based on this vector difference, the geometric offset of multiple frames of image is calculated in reverse by combining the focal length, pixel size and other parameters of the satellite optical system, and the projection position deviation of the same ground feature in different frames is clarified. Next, the digital elevation model (DEM) of the ground target area is called to extract the elevation data of the ground position corresponding to each frame of image. The satellite line-of-sight projection path is corrected according to the elevation difference to eliminate the additional registration deviation caused by terrain undulation. Finally, the sub-pixel level interpolation algorithm is used to accurately align each frame of image according to the corrected geometric offset, and the sub-pixel level geometric registration of multiple frames of ground target image is completed.

[0066] By combining the WGS-84 Earth ellipsoid model to calculate the line-of-sight direction vector, the consistency of the reference for calculating the spatial position of satellites and ground targets can be ensured, avoiding directional errors caused by coordinate system confusion. Transforming the line-of-sight vector to the ENU coordinate system makes the vector operation compatible with the satellite attitude logic, which can improve the accuracy of calculating inter-frame vector differences. The geometric offset obtained based on the vector difference can accurately quantify the inter-frame spatial deviation, providing a clear target for registration. The correction of terrain deviation by the digital elevation model can further eliminate the interference of terrain undulation on registration accuracy. The sub-pixel-level geometric registration achieved can significantly improve the spatial alignment accuracy of multi-frame images, effectively avoid super-resolution reconstruction artifacts caused by registration errors, and provide a high-quality spatial reference for subsequent photometric normalization and fusion reconstruction. This can significantly improve the detail reproduction, spatial resolution, and ground feature positioning accuracy of the final optical remote sensing image.

[0067] Step 250: Based on the illumination angle, solar azimuth angle and atmospheric transparency parameters in the auxiliary data, and in conjunction with the ground data and physical model, perform radiometric calibration and atmospheric correction on the multi-frame ground target images after sub-pixel geometric registration, and use a weighted average strategy or filtering strategy to suppress random noise in the images to complete photometric normalization processing.

[0068] Among them, the illumination angle in the auxiliary data is the angle between the sunlight and the normal to the ground target surface, recorded synchronously during satellite image acquisition. This directly affects the brightness of ground features in the image (the smaller the angle, the more sunlight the ground features receive, and the brighter the image). The solar azimuth angle in the auxiliary data is the angle between the projection of the sun onto the horizontal plane of the ground target area and the local due north direction during satellite image acquisition, reflecting the horizontal characteristics of sunlight. The atmospheric transparency parameter in the auxiliary data is used to characterize the degree of atmospheric attenuation of solar radiation (such as quantitative indicators of water vapor and aerosol content in the atmosphere). The lower the transparency, the greater the atmospheric scattering and absorption. The stronger the radiation, the more pronounced the image brightness attenuation; ground data refers to ground-based reference radiation data, topographic reference data, etc., used to assist in calibrating the radiometric accuracy of satellite imagery; the physical model refers to the atmospheric radiative transfer model, which simulates the transmission process of radiation in the atmosphere based on illumination and atmospheric parameters, providing a theoretical basis for radiometric calibration and atmospheric correction; radiometric calibration refers to the process of converting the digital quantization value (DN value) of satellite imagery into the absolute radiance received by the sensor using calibration coefficients provided by the sensor, aiming to establish a quantitative correspondence between image values ​​and the actual radiation of ground features; atmospheric correction refers to the process of eliminating atmospheric scattering and absorption effects on radiation based on the physical model. The system addresses noise interference by estimating and removing atmospheric path radiation and aerosol effects, and converting TOA reflectance to surface reflectance. A weighted averaging strategy, a noise suppression method, assigns different weights to image pixels based on their signal-to-noise ratio (SNR) (e.g., the noise difference between edge and flat areas). Pixels with higher SNR (detail areas) receive higher weights, while pixels with lower SNR (noisy areas) receive lower weights. This weighted calculation reduces the impact of noise on the image. A filtering strategy, another noise suppression method, uses specific filtering algorithms (e.g., Gaussian filtering, median filtering) to smooth random noise in the image (such as sensor electronic noise and atmospheric particle scattering). The image processing aims to prevent the formation of speckles and retain ground details as much as possible, avoiding the loss of details caused by excessive smoothing. Random noise is the irregular distribution of grayscale fluctuations in the image (such as random jumps in pixel grayscale values), which mainly comes from electronic noise and atmospheric disturbances of satellite imaging sensors. If not suppressed, it will lead to noise superposition during multi-frame fusion, reducing the clarity of the reconstructed image. Remote sensing data preprocessing (i.e., photometric normalization processing) refers to the complete processing flow that integrates radiometric calibration, atmospheric correction and noise suppression, ultimately unifying the surface reflectance of multiple frames of images, eliminating photometric mismatch caused by differences in observation conditions, and providing a photometrically consistent data source for subsequent super-resolution fusion reconstruction.

[0069] In this embodiment of the disclosure, the ground system first extracts the illumination angle, solar azimuth angle, and atmospheric transparency parameters corresponding to each frame of the image from the auxiliary data. Based on these parameters, the ground system performs radiometric calibration on the multi-frame ground target images that have completed sub-pixel-level geometric registration, using ground data and a physical model, and converts the digital quantization values ​​of the images into absolute radiance. Then, the physical model simulates the transmission process of radiation in the atmosphere to complete atmospheric correction, thereby eliminating the interference of atmospheric scattering and absorption on radiation and restoring the true radiation characteristics of ground objects. Subsequently, based on the noise distribution characteristics of the images, a weighted averaging strategy (giving high weight to detailed areas) or a filtering strategy (smoothing noisy areas) is selected to suppress random noise. Finally, the remote sensing data preprocessing of the multi-frame images is completed, so that each frame of the image maintains a high degree of uniformity in brightness and contrast.

[0070] This technical step combines auxiliary data and physical models to perform radiometric calibration and atmospheric correction, establishing a quantitative relationship between image values ​​and the actual radiation of ground features. This eliminates the interference of atmospheric and illumination differences on photometric values, restoring the true radiometric characteristics of ground features. Noise suppression through weighted averaging or filtering strategies reduces the impact of random noise while preserving ground feature details. Overall remote sensing data preprocessing enables multiple frames of images to have consistent photometric characteristics, providing high-quality input with geometric alignment and photometric uniformity for subsequent super-resolution fusion and reconstruction. This effectively reduces fusion artifacts and improves the clarity, texture realism, and radiometric consistency of the final optical remote sensing image, meeting the photometric accuracy requirements of high-precision Earth observation.

[0071] Step 260: Perform super-resolution fusion reconstruction on the registered and normalized multi-frame ground target images to obtain optical remote sensing images.

[0072] For embodiments of this disclosure, the steps may include: Step 260-1: Calculate the inter-frame sub-pixel level displacement field based on the sub-pixel level registration results of multi-frame ground target images, and identify the complementary detail distribution area and redundant noise area of ​​each frame ground target image through the spatial distribution characteristics of the inter-frame sub-pixel level displacement field.

[0073] The inter-frame sub-pixel displacement field, based on sub-pixel registration results, describes the spatial distribution of the positional offset of each pixel (or pixel block) in two adjacent frames at the sub-pixel scale. It is typically presented as a two-dimensional vector matrix, where the vector direction represents the displacement direction and the vector amplitude represents the displacement magnitude, quantifying the spatial distribution of inter-frame geometric differences. The spatial distribution characteristics of the inter-frame sub-pixel displacement field represent the variation and spatial correlation of the offsets within the displacement field. For example, some regions exhibit stable displacement amplitudes and consistent directions (potentially corresponding to redundant information), while others show differential fluctuations in displacement amplitude or direction (potentially corresponding to complementary details). This is the core criterion for distinguishing details from noise; complementary details... The segment distribution region is a spatial region in a multi-frame image that is unique to one or several frames and can supplement the missing ground feature details in other frames (such as a frame clearly showing the corner of a building, and another frame clearly showing the texture of vegetation). In the displacement field, this region usually shows a differential correlation between the displacement direction / amplitude and the surrounding area, corresponding to local changes in features such as the edges and textures of ground features. The redundant noise region is a repetitive information or random noise region that is common in multi-frame images and lacks effective support from ground feature details (such as uniform grayscale fluctuations on a flat road surface, or speckles caused by sensor electronic noise). In the displacement field, this region usually shows a small and irregular displacement amplitude, or highly repetitive displacement patterns across multiple frames, with no additional value for supplementing details.

[0074] In this embodiment of the disclosure, the ground system can first call the sub-pixel level registration results of multiple frames of ground target images, take the image with the best quality as the reference frame, and calculate the sub-pixel scale position offset of each pixel (or pixel block) between the remaining frames and the reference frame through optical flow method or cross-correlation matching algorithm to construct an inter-frame sub-pixel level displacement field covering the entire image; then analyze the spatial distribution characteristics of the displacement field, such as by statistically analyzing the displacement amplitude changes, direction consistency and inter-frame correlation of different regions, to identify regions with differential fluctuations in the displacement field (these regions correspond to unique ground feature details captured in different frames, i.e., complementary detail distribution regions), and at the same time filter out regions with small and irregular displacement amplitudes or highly repetitive displacement patterns in multiple frames (these regions have no additional detail supplementation, and are mostly repetitive information or random noise, i.e., redundant noise regions); throughout the process, it is necessary to combine prior knowledge of ground features (such as building edges usually correspond to abrupt displacement changes, and flat areas usually correspond to stable displacement) to optimize the recognition accuracy and ensure that the distinction between complementary details and redundant noise is accurate and reliable.

[0075] By calculating the inter-frame displacement field based on sub-pixel level registration results, the spatial differences between multiple frames of images can be accurately quantified, providing an objective quantitative basis for distinguishing details from noise and avoiding recognition bias caused by traditional subjective judgment. By using the spatial distribution characteristics of the displacement field to identify complementary detail distribution areas, the most valuable detail information in multiple frames of images can be efficiently located, providing a clear target for weight allocation during subsequent fusion and reconstruction (assigning high weights to complementary details). At the same time, the accurate identification of redundant noise areas can provide a target range for noise suppression (assigning low weights to redundant noise), avoiding the impact of noise superposition on reconstruction quality.

[0076] Step 260-2: Based on complementary detail distribution regions, a weighted iterative maximum a posteriori reconstruction algorithm is adopted. Combined with satellite attitude control parameters and spatial smoothing prior conditions, pixel-level fusion is performed on multiple frames of ground target images. Complementary detail distribution regions are given high fusion weights, and redundant noise regions are suppressed by weights to generate the initial image.

[0077] Among them, the weighted iterative maximum a posteriori reconstruction algorithm is a probabilistic image fusion algorithm that aims to maximize the posterior probability. It adjusts the pixel weights of each frame through multiple iterations, ensuring the fusion result conforms to the statistical characteristics of the multi-frame observation data while satisfying preset prior constraints (such as spatial smoothing), achieving a balance between detail preservation and noise suppression. The satellite attitude control parameters are the three-axis attitude adjustment data of the satellite during imaging missions, which can quantify the inter-frame geometric correlation of multiple images, providing geometric consistency constraints for the fusion process and avoiding fusion artifacts caused by inter-frame positional deviations. The spatial smoothing prior bar... The first image is a pre-defined constraint rule in super-resolution reconstruction, which requires the fused image to maintain grayscale continuity in space. This is used to suppress the generation of false textures that do not exist in real ground features and to ensure the rationality of the image structure. Pixel-level fusion is a fusion method that integrates the grayscale values ​​and detail information of corresponding pixel positions in multiple frames of images, with a single pixel as the processing unit. The first image is the first high-resolution image generated after fusion by the weighted iterative maximum a posteriori reconstruction algorithm. It has initially integrated the complementary details of multiple frames of images and suppressed redundant noise, but there may be shortcomings in local details (such as complex terrain areas) that need further optimization.

[0078] In this embodiment, the ground system can first call the identified complementary detail distribution regions and redundant noise regions, and use the weighted iterative maximum a posteriori reconstruction algorithm as the core framework. The satellite attitude control parameters are incorporated into the algorithm model as geometric constraints to ensure that the pixel position association of multiple frames of images during the fusion process is consistent with the geometry of the real imaging. At the same time, a spatial smoothing prior condition is introduced, and a reasonable threshold for the gray-level change of adjacent pixels is set to avoid the generation of false textures during the fusion process. During the algorithm iteration process, each pixel in the complementary detail distribution region is given a high fusion weight, and the unique ground object edges, textures and other details of the region are retained first. The pixels in the redundant noise region are weighted to suppress the influence of repetitive information and random noise. Through multiple rounds of iterative calculation (verifying the fit between the fusion result and the observation data after each round of iteration and adjusting the weight allocation strategy), the pixel contribution of each frame of image is gradually optimized, and finally, an initial high-resolution image integrating multiple frames of complementary details and suppressing redundant noise is generated.

[0079] This technical step, through a weighted iterative maximum a posteriori reconstruction algorithm combined with the geometric constraints of satellite attitude control parameters, ensures the consistency between the fusion result and the actual imaging geometry, avoiding fusion artifacts caused by inter-frame positional deviations. The introduction of spatial smoothing prior conditions can effectively suppress the generation of false textures and ensure the physical authenticity of ground textures. The strategy of assigning high weights to complementary detail distribution areas and suppressing redundant noise areas can maximize the preservation of unique effective details from multiple frames of images, reduce the interference of noise superposition, and significantly improve the spatial resolution and detail reproduction of the initial image.

[0080] Step 260-3: For areas with complex terrain or dynamic changes in land features in the initial image, use adaptive weighting or block reconstruction strategies to perform local fine-grained corrections.

[0081] Among them, the adaptive weighting strategy is a local optimization method that dynamically adjusts the fusion weights based on regional characteristics. By analyzing the detail clarity (such as edge gradient and texture contrast) of each frame of image in complex terrain or dynamic regions, it assigns higher weights to frames with better details and lower weights to frames that are blurry or have variations, thereby achieving dynamic adaptation and detail supplementation. The block reconstruction strategy is a local optimization method that divides the image into small blocks according to regional characteristics (such as by terrain slope or land cover type) and then performs reconstruction independently on each block. For complex terrain blocks, it adopts geometric correction parameters that adapt to elevation differences, and for dynamically changing blocks, it adopts inter-frame variation compensation parameters to avoid local detail distortion caused by globally uniform parameters.

[0082] In this embodiment of the disclosure, the ground system can first perform regional feature analysis on the initial image, identify complex terrain areas with dramatic topographic relief by combining the digital elevation model (DEM) in the auxiliary data, and locate dynamically changing areas of ground features by comparing the inter-frame differences of multiple images, thus clarifying the distribution of detail gaps in the initial image. Subsequently, an adaptation strategy is selected according to the type of gap area. For example, for complex terrain areas, a block reconstruction strategy is used to divide it into several small areas. Each small area is combined with the corresponding DEM elevation data to recalculate the satellite line-of-sight projection relationship and adjust the reconstruction parameters to correct the detail blur caused by terrain undulations. For dynamically changing areas of ground features, an adaptive weighting strategy is used to analyze the detail clarity and consistency of the area in each frame of the image. Frames that capture the stable state of dynamic ground features and have clearer details are given high weights, while frames with changes and deviations are given low weights. The complete outline and texture of dynamic ground features are supplemented by weight adjustment. Finally, the locally corrected area is stitched pixel-level with the intact area in the initial image, and the stitching traces are eliminated by edge smoothing processing to ensure that the corrected image maintains continuity and consistency in both overall structure and local details, thus completing the local fine-tuning optimization of the initial image.

[0083] This technical step, by specifically targeting areas with insufficient detail in the initial image, avoids the inefficiency and loss of detail in intact areas caused by repeated processing of the entire image, thus improving the accuracy of optimization. The adaptive weighting strategy can dynamically match the inter-frame detail differences in dynamic regions, and the block reconstruction strategy can adapt to the elevation features of complex terrain. The flexible application of these two strategies can effectively compensate for the insufficient adaptability of the global fusion algorithm in special regions, significantly improving the clarity of the outline of features in complex terrain and the integrity of details in dynamic regions. At the same time, the seamless stitching of the local correction with the initial image can ensure the overall structural coherence and radiometric consistency of the final image, so that the image not only has global high resolution, but also accurately restores the microscopic details of special regions, further meeting the needs of fine identification of complex terrain and dynamic features in high-precision Earth observation scenarios.

[0084] Step 260-4: Based on preset quantitative indicators, conduct reconstruction accuracy assessment on the initial version of the image after detail correction, and select the initial version of the image that meets the preset resolution requirements as the final optical remote sensing image.

[0085] Among them, the preset quantitative indicators are quantifiable evaluation standards used to objectively measure the quality of image reconstruction. They typically include structural similarity (SSIM), which reflects the consistency of image structure; peak signal-to-noise ratio (PSNR), which reflects sharpness and noise suppression effect; and surface projection positioning error, which reflects the accuracy of ground feature positioning. The threshold values ​​of the indicators are preset according to the accuracy requirements of the Earth observation task (e.g., SSIM ≥ 0.9, PSNR ≥ 35dB, positioning error ≤ 1 meter). The preset resolution requirements are the final image resolution standards preset according to the goals of the Earth observation task (e.g., target detection, ground feature classification). They typically include spatial resolution (e.g., at least twice that of the original single frame) and positioning resolution (e.g., surface projection positioning error ≤ 1 meter), which are the core criteria for selecting the final image. The final optical remote sensing image is the image result that has passed the reconstruction accuracy evaluation and meets the preset resolution requirements. It integrates complementary details from multiple frames of images, suppresses redundant noise, and corrects shortcomings in special areas. It has the characteristics of high spatial resolution, high texture realism, and high positioning accuracy, and can be directly used for downstream high-precision Earth observation applications.

[0086] In this embodiment, the ground system first retrieves pre-set quantitative evaluation indicators (such as SSIM, PSNR, and surface projection positioning error) and corresponding threshold standards. Then, it inputs the detailed-corrected initial image into the evaluation model to calculate specific values ​​for structural similarity, peak signal-to-noise ratio, and surface projection positioning error. Next, these calculated evaluation values ​​are compared one by one with the thresholds corresponding to the preset resolution requirements to determine whether the image meets the expected standards in all dimensions. If all evaluation values ​​meet or exceed the thresholds, it indicates that the image's resolution, detail accuracy, and positioning accuracy meet the mission requirements, and the detailed-corrected initial image is directly determined as the final optical remote sensing image. If one or more evaluation values ​​fail to meet the standards, the evaluation results are fed back to the earlier local fine-tuning correction stage. The correction parameters are adjusted (such as the adaptive weighted weight allocation ratio and the granularity of the block reconstruction area division) and the detailed-corrected image is regenerated. The accuracy evaluation is then performed again until an image that meets the preset resolution requirements is produced.

[0087] Reconstruction accuracy assessment based on preset quantitative indicators provides an objective and quantifiable standard for judging image quality, effectively avoiding biases caused by subjective evaluation, and ensuring that the final output optical remote sensing image consistently meets mission requirements in terms of resolution, detail integrity, and ground feature positioning accuracy. The feedback optimization mechanism for when evaluation standards are not met forms a closed loop of correction-evaluation-re-optimization, which can further compensate for potential detail gaps in preprocessing and improve the reliability of image quality.

[0088] For ease of understanding of this application, the following is combined with Figure 5The flowchart shown below provides a complete explanation of the technical solution in this application: First, observation planning is carried out to clarify the observation location, satellite transit time, observation start and end coordinates, and attitude control scheme during the observation period; then, single-satellite multi-frame acquisition is performed, with a single satellite initiating imaging from the observation starting point. After imaging, the attitude is adjusted to repeatedly align with the observation starting point, and then the ground target is sub-pixel displacement is generated through attitude fine-tuning and imaging is repeated; next, multi-satellite collaborative imaging is performed, with adjacent satellites in the constellation passing over the target area in sequence to complete imaging, and each satellite synchronously performing single-satellite multi-frame acquisition operations; during this process, attitude and geometric data are recorded synchronously, and each satellite records the corresponding orbital parameters and attitude-related parameters each time it images; finally, data downlink and fusion reconstruction are implemented. The satellite first labels and preprocesses the acquired data before downlinking it to the ground, and the ground processing module completes the registration of multi-frame images based on orbit, attitude, and other information, and then performs super-resolution reconstruction to finally obtain high-resolution optical remote sensing images.

[0089] In summary, the technical solution in this application, by generating attitude control commands containing preset sub-pixel-level displacement strategies, deeply couples satellite attitude control with the requirements of super-resolution algorithms. This eliminates the reliance on random displacements such as natural satellite jitter and orbital deviations, providing precisely quantifiable sub-pixel displacement data for subsequent reconstruction. Simultaneously, it achieves high-resolution imaging without increasing the complexity of the optical system, contributing to lightweight and low-cost satellite payloads and reducing manufacturing, launch, and maintenance costs. Furthermore, by performing sub-pixel-level geometric registration and photometric normalization processing based on auxiliary data transmitted from the satellite, it can effectively correct for the effects of terrain undulations. This approach addresses the misalignment of multiple images across different frames, eliminating issues caused by factors such as illumination angle and atmospheric transparency, thus suppressing random noise and significantly reducing registration error accumulation. Super-resolution fusion reconstruction is performed using multi-frame real-world observation data after registration normalization. This avoids the illusionary drawbacks of single-frame deep learning super-resolution methods, ensuring the physical realism of ground textures. Furthermore, it enhances reconstruction accuracy by incorporating attitude control parameter constraints. The scheme is also adaptable to multi-satellite collaborative observation scenarios, further ensuring observation consistency through temporal coordination within constellation configurations, enhancing robustness and cross-scenario generalization capabilities in complex environments. Overall, this scheme achieves collaborative optimization of satellite on-orbit imaging and ground-based super-resolution algorithms. It is applicable to single-satellite platforms and can be extended to multi-satellite constellations. The final output optical remote sensing image resolution is significantly improved compared to the original single-frame data, with controllable surface positioning errors, meeting the practical application requirements of high-precision Earth observation.

[0090] Furthermore, as Figure 1 and Figure 2 The specific implementation of the method shown in this embodiment provides a multi-frame super-resolution imaging device for optical remote sensing satellites, such as... Figure 6 As shown, the device may include: a generation module 61, a receiving module 62, a registration module 63, and a reconstruction module 64.

[0091] The generation module 61 can be used to generate satellite attitude control commands containing preset sub-pixel level displacement strategies based on the requirements of Earth observation missions, satellite orbit parameters and constellation configuration information. The receiving module 62 can be used to receive multiple frames of ground target images acquired by the satellite according to the satellite attitude control command, as well as auxiliary data synchronously transmitted back by the satellite; The registration module 63 can be used to perform sub-pixel-level geometric registration of multiple frames of ground target images based on auxiliary data, and perform photometric normalization processing for image differences under different observation conditions. The reconstruction module 64 can be used to perform super-resolution fusion reconstruction of registered and normalized multi-frame ground target images to obtain optical remote sensing images.

[0092] In some embodiments of this application, the satellite attitude control command includes a first satellite attitude control command, which instructs one or more satellites to independently acquire multiple frames of ground target images. The receiving module 62 is specifically used to receive multiple frames of ground target images acquired by the satellite according to the first satellite attitude control command, as well as auxiliary data synchronously transmitted back by the satellite. When the satellite acquires multiple frames of ground target images in response to the first satellite attitude control command, it is used to establish a geometric mapping relationship with the ground target area by combining a digital elevation model. Based on the geometric mapping relationship and a preset sub-pixel level displacement strategy, the satellite passes over the ground target area. During the passage, the satellite attitude is adjusted in real time by a reaction flywheel or other attitude control system, so that the ground target area generates a sub-pixel level projection displacement between adjacent frames.

[0093] In some embodiments of this application, the satellite attitude control command includes a second satellite attitude control command carrying a multi-satellite collaborative observation time sequence. The second satellite attitude control command is used to instruct multiple satellites to collaboratively acquire multiple frames of ground target images according to the multi-satellite collaborative observation time sequence. The receiving module 62 is specifically used to receive multiple frames of ground target images acquired by multiple satellites according to the second satellite attitude control command, as well as synchronously transmitted auxiliary data. When multiple satellites acquire multiple frames of ground target images in response to the second satellite attitude control command, they are used to establish a geometric mapping relationship with the ground target area by combining a digital elevation model. Based on the geometric mapping relationship and a preset sub-pixel level displacement strategy, they pass over the ground target area sequentially according to the multi-satellite collaborative observation time sequence within a preset time period. During the passage, the satellite attitude is adjusted in real time by a reaction flywheel or other attitude control system so that the images acquired by multiple satellites maintain consistency in both geometric and radiometric dimensions, and sub-pixel level projection displacement is generated between adjacent frames acquired by each satellite.

[0094] In some embodiments of this application, the generation module 61 can be specifically used to calculate the visibility window of each satellite to the ground target area based on the requirements of the Earth observation mission, satellite orbit parameters and constellation configuration information, and to filter the imaging period that meets the illumination conditions and the nadir point viewing angle conditions; to design the satellite three-axis attitude adjustment sequence according to the imaging period and satellite imaging mode, to select the sub-pixel displacement trajectory of orthogonal stepping or spiral scanning, and to form a preset sub-pixel level displacement strategy; if a single-satellite independent acquisition mission is performed, the preset sub-pixel level displacement strategy is encapsulated into a first satellite attitude control command in combination with the on-orbit priority and task allocation status of the single satellite in the constellation configuration information; and / or, if a multi-satellite collaborative acquisition mission is performed, the observation interval of adjacent satellites to the same target and the observation time span of all satellites are controlled according to the orbital phase difference and transit time difference of each satellite in the constellation configuration information to form a multi-satellite collaborative observation sequence, and the preset sub-pixel level displacement strategy and the multi-satellite collaborative observation sequence are encapsulated into a second satellite attitude control command.

[0095] In some embodiments of this application, the generation module 61 can be specifically used to determine the ground projection size of the satellite imaging pixels based on the satellite orbit parameters, set the sub-pixel displacement amplitude to 1 / 3 to 1 / 2 of the ground projection size, and determine the initial geodetic coordinates of the satellite and the reference geodetic coordinates of the ground target area by combining the satellite's transit trajectory during the imaging period; calculate the radius of curvature of the primordial and maxima corresponding to the initial geodetic coordinates based on the semi-major axis and flattening parameters of the WGS-84 Earth ellipsoid model, transform the geodetic coordinates of the satellite and the ground target area to the geocentric-Earth-fixed coordinate system based on the radius of curvature of the primordial and maxima and calculate the coordinate difference, calculate the original line-of-sight vector through the coordinate difference and complete the normalization process to obtain the standard line-of-sight direction vector; establish a transformation matrix from the geocentric-Earth-fixed coordinate system to the ENU coordinate system, and based on the transformation matrix... The standard line-of-sight vector is transformed to the ENU coordinate system, and the initial azimuth and initial elevation angles required for satellite imaging are calculated. Sub-pixel displacement trajectories and roll angle control parameters are determined based on the satellite imaging mode, and coordinate corrections for the ground target area are generated for the sub-pixel displacement trajectories. Corrected geodetic coordinates of the ground target area are determined based on the coordinate corrections, and corrected azimuth and elevation angles are determined based on the corrected geodetic coordinates. The azimuth adjustment is obtained by the difference between the initial azimuth and the corrected azimuth, and the elevation adjustment is obtained by the difference between the initial elevation and the corrected elevation. Based on the azimuth adjustment, elevation adjustment, and roll angle control parameters, a satellite three-axis attitude adjustment sequence is generated. The sub-pixel displacement trajectories and the satellite three-axis attitude adjustment sequence are integrated to form a preset sub-pixel-level displacement strategy.

[0096] In some embodiments of this application, the registration module 63 can be specifically used to preprocess the satellite attitude angle, satellite orbit parameters, and timestamps in the auxiliary data, complete the time synchronization and alignment of multi-source data, and uniformly transform the satellite orbit coordinates to the geocentric-ground-fixed coordinate system; based on the preprocessed satellite attitude angle and satellite orbit parameters, combined with the WGS-84 Earth ellipsoid model, calculate the satellite line-of-sight direction vector corresponding to each frame of ground target image, transform the satellite line-of-sight direction vector to the ENU coordinate system and calculate the vector difference, and calculate the geometric offset of multiple frames of ground target images through the vector difference; use the digital elevation model to correct the registration deviation caused by terrain undulation based on the geometric offset, and complete the sub-pixel level geometric registration; based on the illumination angle, solar azimuth angle, and atmospheric transparency parameters in the auxiliary data, perform radiometric calibration and atmospheric correction on the multi-frame ground target images after sub-pixel level geometric registration, and use a weighted average strategy or filtering strategy to suppress random noise in the images, and complete the photometric normalization processing.

[0097] In some embodiments of this application, the reconstruction module 64 can be specifically used to calculate the inter-frame sub-pixel level displacement field based on the sub-pixel level registration results of multiple frames of ground target images, identify the complementary detail distribution region and redundant noise region of each frame of ground target images through the spatial distribution characteristics of the inter-frame sub-pixel level displacement field; based on the complementary detail distribution region, a weighted iterative maximum a posteriori reconstruction algorithm is used, combined with satellite attitude control parameters and spatial smoothing prior conditions, to perform pixel-level fusion of multiple frames of ground target images, assigning high fusion weight to the complementary detail distribution region and suppressing the redundant noise region to generate an initial image; for the detail deficiencies in complex terrain or dynamically changing areas of ground features in the initial image, an adaptive weighted or block reconstruction strategy is used for local fine-tuning correction; based on preset quantization indicators, the reconstruction accuracy of the detail-corrected initial image is evaluated, and the initial image that meets the preset resolution requirements is selected as the final optical remote sensing image.

[0098] It should be noted that other corresponding descriptions of the functional units involved in the multi-frame super-resolution imaging device for optical remote sensing satellites provided in this embodiment can be found in [reference needed]. Figure 1 and Figure 2 The corresponding descriptions in [the document] will not be repeated here.

[0099] Based on the above, Figure 1 and Figure 2 Accordingly, this embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the above-described method. Figure 1 and Figure 2 The multi-frame super-resolution imaging method for optical remote sensing satellites is shown.

[0100] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause an electronic device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0101] Based on the above, Figure 1 and Figure 2 The method shown, and Figure 6 To achieve the above objectives, the present application also provides an electronic device, specifically a personal computer, tablet computer, server, or other network device, as shown in the virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to achieve the above-described objectives. Figure 1 and Figure 2 The multi-frame super-resolution imaging method for optical remote sensing satellites is shown.

[0102] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0103] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0104] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0105] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platform, or it can be implemented by hardware.

[0106] This invention generates attitude control commands containing preset sub-pixel-level displacement strategies, deeply coupling satellite attitude control with the requirements of super-resolution algorithms. This eliminates reliance on random displacements such as natural satellite jitter and orbital deviations, providing precisely quantifiable sub-pixel displacement data for subsequent reconstruction. Simultaneously, it achieves high-resolution imaging without increasing the complexity of the optical system, contributing to lightweight and low-cost satellite payloads and reducing manufacturing, launch, and maintenance costs. Furthermore, sub-pixel-level geometric registration and photometric normalization based on satellite-returned auxiliary data effectively corrects registration errors caused by terrain undulations. This method eliminates the optical and geometric mismatch issues caused by factors such as illumination angle and atmospheric transparency in multi-frame images, suppresses random noise, and significantly reduces the accumulation of registration errors. Super-resolution fusion reconstruction is performed based on multi-frame real observation data after registration normalization. This avoids the illusionary drawbacks of single-frame deep learning super-resolution methods, ensuring the physical realism of ground textures, and improves reconstruction accuracy by incorporating attitude control parameter constraints. Furthermore, this scheme is adaptable to multi-satellite collaborative observation scenarios, further ensuring observation consistency through temporal collaboration under constellation configuration, enhancing robustness in complex environments and cross-scenario generalization capabilities. Overall, this scheme achieves collaborative optimization of satellite on-orbit imaging and ground-based super-resolution algorithms, applicable to single-satellite platforms and scalable to multi-satellite constellations. The final output optical remote sensing image resolution is significantly improved compared to the original single-frame data, and the surface positioning error is controllable, meeting the practical application requirements of high-precision Earth observation.

[0107] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0108] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A multi-frame super-resolution imaging method for optical remote sensing satellites, characterized in that, The method is applied to ground systems, including: Based on the requirements of Earth observation missions, satellite orbit parameters, and constellation configuration information, satellite attitude control commands containing preset sub-pixel level displacement strategies are generated. Receive multiple frames of ground target images acquired by the satellite according to the satellite attitude control commands, as well as auxiliary data synchronously transmitted back by the satellite; Based on the auxiliary data, sub-pixel-level geometric registration is performed on the multiple frames of ground target images, and photometric normalization is performed on the image differences under different observation conditions. Super-resolution fusion reconstruction is performed on the registered and normalized multi-frame ground target images to obtain optical remote sensing images.

2. The method according to claim 1, characterized in that, The satellite attitude control command includes a first satellite attitude control command, which is used to instruct one or more satellites to independently acquire multiple frames of ground target images. The receiving satellite acquires multiple frames of ground target images according to the satellite attitude control commands, as well as the auxiliary data synchronously transmitted back by the satellite, including: The system receives multiple frames of ground target images acquired by the satellite according to the first satellite attitude control command, as well as auxiliary data synchronously transmitted back by the satellite. When the satellite acquires the multiple frames of ground target images in response to the first satellite attitude control command, it uses a digital elevation model to establish a geometric mapping relationship with the ground target area. Based on the geometric mapping relationship and the preset sub-pixel level displacement strategy, the satellite passes over the ground target area. During the pass, the satellite attitude is adjusted in real time through a reaction flywheel or other attitude control system, so that the ground target area generates a sub-pixel level projection displacement between adjacent frames.

3. The method according to claim 2, characterized in that, The satellite attitude control command includes a second satellite attitude control command carrying a multi-satellite collaborative observation time sequence. The second satellite attitude control command is used to instruct multiple satellites to collaboratively acquire multiple frames of ground target images according to the multi-satellite collaborative observation time sequence. The receiving satellite acquires multiple frames of ground target images according to the satellite attitude control commands, as well as the auxiliary data synchronously transmitted back by the satellite, including: The system receives multiple frames of ground target images acquired by the multiple satellites according to the attitude control command of the second satellite, as well as synchronously transmitted auxiliary data. When the multiple satellites acquire the multiple frames of ground target images in response to the attitude control command of the second satellite, they use a digital elevation model to establish a geometric mapping relationship with the ground target area. Based on the geometric mapping relationship and the preset sub-pixel displacement strategy, the satellites pass over the ground target area sequentially according to the multi-satellite collaborative observation time sequence within a preset time period. During the passage, the satellite attitude is adjusted in real time through a reaction flywheel or other attitude control system to ensure that the images acquired by the multiple satellites maintain consistency in both geometric and radiometric dimensions, and that sub-pixel projection displacement is generated between adjacent frames acquired by each satellite.

4. The method according to claim 3, characterized in that, The process of generating satellite attitude control commands, including preset sub-pixel level displacement strategies, based on Earth observation mission requirements, satellite orbit parameters, and constellation configuration information, includes: Based on the Earth observation mission requirements, the satellite orbit parameters, and the constellation configuration information, the visibility window of each satellite over the ground target area is calculated, and imaging time periods that meet the lighting conditions and the nadir point viewing angle conditions are selected. Based on the imaging period and satellite imaging mode, a satellite three-axis attitude adjustment sequence is designed, and sub-pixel displacement trajectories of orthogonal stepping or spiral scanning are selected to form a preset sub-pixel level displacement strategy. If a single-satellite independent acquisition task is performed, then, based on the on-orbit priority and task allocation status of the single satellite in the constellation configuration information, the preset sub-pixel level displacement strategy is encapsulated into the first satellite attitude control command; and / or, If a multi-satellite collaborative acquisition task is performed, the observation interval of adjacent satellites for the same target and the observation time span of all satellites are controlled according to the orbital phase difference and transit time difference of each satellite in the constellation configuration information, so as to form a multi-satellite collaborative observation sequence. The preset sub-pixel level displacement strategy and the multi-satellite collaborative observation sequence are encapsulated into the second satellite attitude control command.

5. The method according to claim 4, characterized in that, The step of designing a satellite three-axis attitude adjustment sequence based on the imaging period and satellite imaging mode, selecting sub-pixel displacement trajectories for orthogonal stepping or helical scanning, and forming a preset sub-pixel level displacement strategy includes: Based on the satellite orbit parameters, the ground projection size of the satellite imaging pixels is determined, and the sub-pixel displacement amplitude is set to 1 / 3 to 1 / 2 of the ground projection size. Combined with the satellite's transit trajectory during the imaging period, the initial geodetic coordinates of the satellite and the reference geodetic coordinates of the ground target area are determined. Based on the semi-major axis and flattening parameters of the WGS-84 Earth ellipsoid model, the radius of curvature of the prime meridian corresponding to the initial geodetic coordinates is calculated. Based on the radius of curvature of the prime meridian, the geodetic coordinates of the satellite and the ground target area are transformed to the geocentric geodetic coordinate system and the coordinate difference is calculated. The original line-of-sight vector is calculated through the coordinate difference and normalized to obtain the standard line-of-sight direction vector. Establish a transformation matrix from the geocentric-fixed coordinate system to the ENU coordinate system. Based on the transformation matrix, transform the standard line-of-sight direction vector to the ENU coordinate system and calculate the initial azimuth and initial elevation angles required for satellite imaging. The sub-pixel displacement trajectory and roll angle control parameters are determined based on the satellite imaging mode, and the coordinate correction amount of the ground target area is generated for the sub-pixel displacement trajectory. Based on the coordinate correction amount, the corrected geodetic coordinates of the ground target area are determined, and the corrected azimuth and corrected pitch angles are determined based on the corrected geodetic coordinates. The azimuth adjustment amount is obtained by the difference between the initial azimuth and the corrected azimuth, and the pitch angle adjustment amount is obtained by the difference between the initial pitch angle and the corrected pitch angle. Based on the azimuth adjustment amount, the pitch adjustment amount, and the roll angle control parameters, a satellite three-axis attitude adjustment sequence is generated. The sub-pixel displacement trajectory is integrated with the satellite three-axis attitude adjustment sequence to form a preset sub-pixel level displacement strategy.

6. The method according to claim 1, characterized in that, Based on the auxiliary data, sub-pixel-level geometric registration is performed on the multi-frame ground target images. Simultaneously, photometric normalization is performed to address image differences under different observation conditions, including: The satellite attitude angles, satellite orbit parameters and timestamps in the auxiliary data are preprocessed to complete the time synchronization and alignment of multi-source data, and the satellite orbit coordinates are uniformly transformed to the geocentric and geofixed coordinate system. Based on the preprocessed satellite attitude angles and satellite orbit parameters, the satellite line-of-sight vector corresponding to each frame of ground target imagery is calculated using the WGS-84 Earth ellipsoid model. The satellite line-of-sight vector is then transformed to the ENU coordinate system and the vector difference is calculated. The geometric offset of the multi-frame ground target imagery is calculated using the vector difference. The digital elevation model is then used to correct the registration deviation caused by terrain undulations based on the geometric offset, thus completing sub-pixel-level geometric registration. Based on the illumination angle, solar azimuth angle, and atmospheric transparency parameters in the auxiliary data, combined with ground data and physical models, radiometric calibration and atmospheric correction are performed on the multi-frame ground target images after sub-pixel-level geometric registration. A weighted averaging strategy or filtering strategy is used to suppress random noise in the images, and photometric normalization is completed.

7. The method according to claim 1, characterized in that, The process of super-resolution fusion and reconstruction of registered and normalized multi-frame ground target images to obtain optical remote sensing images includes: The inter-frame sub-pixel level displacement field is calculated based on the sub-pixel level registration results of multi-frame ground target images. The complementary detail distribution region and redundant noise region of each frame ground target image are identified by the spatial distribution characteristics of the inter-frame sub-pixel level displacement field. Based on the complementary detail distribution region, a weighted iterative maximum a posteriori reconstruction algorithm is used to perform pixel-level fusion of the multi-frame ground target images by combining satellite attitude control parameters and spatial smoothing prior conditions. The complementary detail distribution region is given a high fusion weight, and the redundant noise region is weighted to suppress the noise, thus generating the initial image. To address the shortcomings in detail in areas with complex terrain or dynamic changes in land features in the initial imagery, an adaptive weighted or block-based reconstruction strategy is employed for localized fine-grained correction. Based on preset quantitative indicators, the reconstruction accuracy of the initial image after detail correction is evaluated, and the initial image that meets the preset resolution requirements is selected as the final optical remote sensing image.

8. A multi-frame super-resolution imaging device for optical remote sensing satellites, characterized in that, The device is applied to a ground system and includes: The generation module is used to generate satellite attitude control commands containing preset sub-pixel level displacement strategies based on the requirements of Earth observation missions, satellite orbit parameters and constellation configuration information. The receiving module is used to receive multiple frames of ground target images acquired by the satellite according to the satellite attitude control command, as well as auxiliary data synchronously transmitted back by the satellite; The registration module is used to perform sub-pixel-level geometric registration of the multi-frame ground target images based on the auxiliary data, and to perform photometric normalization processing for image differences under different observation conditions. The reconstruction module is used to perform super-resolution fusion reconstruction on registered and normalized multi-frame ground target images to obtain optical remote sensing images.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

10. An electronic device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

Images