Remote sensing image production method and system

Abstract

The invention provides a remote sensing image production method and system, and relates to the technical field of remote sensing image automatic production, and the method comprises the steps: obtaining original remote sensing image data; processing the original remote sensing image data according to a preset image processing rule to obtain a processed image of the original remote sensing image data; performing quality detection on the processed image to obtain a quality detection result of the processed image; according to a quality inspection result, judging whether the processed image reaches a qualified standard or not; when the processed image does not reach the qualified standard, locally repairing the processed image according to the quality inspection result until the processed image reaches the qualified standard; and when the processed image reaches the qualified standard, taking the processed image as an output image. Through accurate quality detection, a local repair strategy and an intelligent repair mechanism, the production efficiency is improved, and the quality of the output image is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing image automation production, in particular to a remote sensing image production method and system. BACKGROUND

[0002] At present, the existing remote sensing image automation production system generally follows a linear process of collection, preprocessing, quality inspection and output. First, the original data downloaded from the satellite is preprocessed, such as radiation, geometry, atmospheric correction and image fusion. Then, the quality of the entire image is evaluated in the quality inspection link. Finally, the qualified image is stored or distributed.

[0003] In the related art, when the quality inspection detects quality defects such as cloud pollution, blur and color cast, the complete process from preprocessing to quality inspection will be restarted, resulting in increased resource consumption. Not only the computing power is wasted, but also the production cost of remote sensing images is increased due to multiple rework, and the production task of remote sensing images is delayed. SUMMARY

[0004] The problem solved by the present application is how to improve the efficiency and quality of remote sensing image production.

[0005] To solve the above problems, the present application provides a remote sensing image production method and system.

[0006] In a first aspect, the present application provides a remote sensing image production method, comprising: obtaining original remote sensing image data; processing the original remote sensing image data according to a preset image processing rule to obtain a processed image of the original remote sensing image data; performing quality inspection on the processed image to obtain a quality inspection result of the processed image; determining whether the processed image meets a qualified standard according to the quality inspection result; when the processed image does not meet the qualified standard, performing local repair on the processed image according to the quality inspection result until the processed image meets the qualified standard; when the processed image meets the qualified standard, taking the processed image as a produced image.

[0007] Optionally, the processing of the original remote sensing image data according to the preset image processing rule to obtain a processed image comprises: dividing the original remote sensing image data into a plurality of tiles with the same size according to division parameters of the preset image processing rule; performing radiation correction, geometry correction and atmospheric correction on each tile in sequence according to correction parameters of the preset image processing rule to obtain a corrected tile; The corrected tiles are combined to obtain the processed image.

[0008] Optionally, the step of performing quality inspection on the processed image to obtain the quality inspection result of the processed image includes: Pixel-level defect detection is performed on the processed image to obtain defect information for each tile in the processed image. The defect information includes defect type, defect coordinates, and defect area. The defect type, defect coordinates, and defect area of ​​each tile are summarized to obtain the defect type set, total defect area, and defect density of the processed image. The set of defect types, the total defect area, and the defect density of the processed image are used as the quality inspection result of the processed image.

[0009] Optionally, determining whether the processed image meets the qualification standard based on the quality inspection result includes: The quality inspection results are compared with multiple preset standards in the qualification standards to obtain the comparison results corresponding to the defect type set, the total defect area, and the defect density, respectively. Based on the comparison results, it is determined whether the processed image meets the qualification standard; Specifically, if any defect type in the defect type set does not belong to the qualified type in the preset standard, or the total defect area is greater than or equal to the maximum qualified area in the preset standard, or the defect density is greater than or equal to the maximum qualified density in the preset standard, the processed image is determined to have failed to meet the qualified standard. When all the defect types in the defect type set belong to the qualified type, the total defect area is less than the maximum qualified area, and the defect density is less than the maximum qualified density, the processed image is determined to meet the qualified standard.

[0010] Optionally, the step of performing local repair on the processed image based on the quality inspection results until the processed image meets the qualification standard includes: Based on the quality inspection results, determine the coordinates of the defects in the processed image; Based on the defect coordinates, the defective tiles in the processed image are determined; Based on the defect information in the defective tile, a local repair task corresponding to the defective tile is generated; The local repair task is performed on the defective tile until the processed image meets the acceptance criteria.

[0011] Optionally, generating a local repair task corresponding to the defective tile based on the defect information in the defective tile includes: Based on the defect type in the defective tile, determine the corresponding repair algorithm and repair parameters for the defective tile; The defect coordinates and defect area in the defective tile are extracted to obtain a multidimensional feature vector of the defective tile. The multidimensional feature vector is input into a dynamic parameter model for prediction to obtain the target parameter value of the repair parameter. Based on the repair algorithm corresponding to the defective tile and the target parameter value of the repair parameter, the local repair task of the defective tile is constructed.

[0012] Optionally, determining the repair algorithm and repair parameters corresponding to the defective tile based on the defect type in the defective tile includes: When the defect type is cloud pollution, the repair algorithm is determined to be a space-time interpolation algorithm, and the threshold of the number of cloudless tiles in the defective tiles, the number of time series frames, and the fusion weight coefficient are used as the repair parameters. When the defect type is fuzzy, the repair algorithm is determined to be an attention-guided reconstruction algorithm, and the edge enhancement coefficient, texture fidelity weight and number of iterations in the defect tile are used as the repair parameters; When the defect type is color cast, the repair algorithm is determined to be a color migration correction algorithm, and the reference color gamut range, brightness scaling factor, and saturation correction coefficient in the defective tile are used as the repair parameters.

[0013] Optionally, performing the local repair task on the defective tile includes: When a defective tile exists in the processed image, the local repair task corresponding to the defective tile is executed directly. When there are multiple defective tiles in the processed image, a priority score for the local repair task is obtained by using a multi-dimensional priority model based on the defect information of the defective tiles corresponding to the local repair task. The local repair tasks are sorted according to the priority scores to obtain the task execution order, and the local repair tasks of all the defective tiles are executed according to the task execution order.

[0014] Optionally, it also includes: When the processed image fails to meet the qualification standard, the number of defective tiles in the processed image is obtained; When the number of defective tiles is greater than or equal to a preset number threshold, spatiotemporal clustering is performed on the defect information of all defective tiles to obtain the defect pattern of the processed image. Based on the defect pattern, the correction parameters in the preset image processing rules are updated, and the updated preset image processing rules are applied in real time to the subsequent processing of the original remote sensing image data.

[0015] Secondly, the present invention provides a remote sensing image production system, comprising: The raw data acquisition unit is used to acquire raw remote sensing image data; The preprocessing unit is used to process the original remote sensing image data according to preset image processing rules to obtain a processed image; A quality inspection unit is used to perform quality inspection on the processed image and obtain the quality inspection result of the processed image; The judgment unit is used to determine whether the processed image meets the qualification standard based on the quality inspection results. The repair unit is used to perform local repair on the processed image according to the quality inspection results when the processed image fails to meet the qualification standard, until the processed image meets the qualification standard. The output unit is used to output the processed image as the output image when the processed image meets the qualification standard.

[0016] The remote sensing image production method and system of this invention significantly improve the efficiency and quality of remote sensing image production through a systematic processing flow and an intelligent repair mechanism. Specifically, in the remote sensing image production process, raw remote sensing image data is first acquired and processed according to preset image processing rules to obtain preliminary processed images. These preset image processing rules are based on long-term production experience, effectively improving the basic quality of the images. Subsequently, the processed images undergo quality inspection to obtain detailed quality inspection results. These results not only accurately determine whether the processed images meet the acceptance standards but also provide basic data for subsequent repair work. Quality inspection quickly identifies problems in the processed images, such as cloud contamination, blurring, and color cast, providing a basis for targeted repair and avoiding the resource waste and time consumption caused by indiscriminate repair of the entire image in traditional methods. When the processed image does not meet the acceptance standards, local repair is performed based on the quality inspection results until the image meets the acceptance standards. Local repair is based on the accurate identification and location of defects, repairing only the problematic parts rather than repeatedly processing the entire image. This method not only reduces the computational resources and time required for restoration but also effectively avoids new problems that may be introduced by global restoration. The implementation of local restoration makes resource utilization more rational in the image production process, significantly improving production efficiency. This invention introduces an intelligent restoration mechanism into the local restoration process. By analyzing quality inspection results, it automatically restores the processed images, thereby achieving optimal restoration results. This not only improves the success rate of restoration but also further enhances the quality of the produced images. Compared with traditional manual restoration or fixed-pattern restoration, the intelligent restoration of this invention can better adapt to different types of defects, ensuring that the restored images meet requirements in terms of quality and detail. In summary, the remote sensing image production method of this invention, through precise quality inspection, local restoration strategies, and an intelligent restoration mechanism, not only improves production efficiency but also ensures the quality of the produced images. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the remote sensing image production method according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the tile repair process according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a remote sensing image production system according to an embodiment of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0019] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0020] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0021] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0022] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties. The collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0023] Combination Figure 1 As shown, an embodiment of the present invention provides a method for producing remote sensing images, comprising: Acquire raw remote sensing image data.

[0024] Specifically, raw remote sensing image data typically originates from satellite sensors, containing information on the reflection and radiation of a region's surface. Due to various factors affecting satellite imaging, such as atmospheric conditions and sensor performance, raw images often contain noise and defects. Therefore, acquiring high-quality raw data is fundamental to ensuring effective subsequent processing. In practice, it is necessary to ensure the integrity and accuracy of the data, while recording relevant metadata, such as imaging time and sensor parameters, for correction and analysis in subsequent processing. For example, consider acquiring raw remote sensing image data from a high-resolution satellite sensor (such as GF-2), covering a geographical area. This image is in TIFF format, has a resolution of 1 meter, and covers an area of ​​100 square kilometers. During data acquisition, relevant metadata, including imaging time, sensor model, and geographical location, is recorded.

[0025] The original remote sensing image data is processed according to preset image processing rules to obtain the processed image of the original remote sensing image data.

[0026] Specifically, preprocessing raw remote sensing image data is a crucial step in improving image quality. Preprocessing includes radiometric correction, geometric correction, and atmospheric correction. Radiometric correction aims to adjust the brightness values ​​of the image and eliminate the influence of atmospheric and sensor noise; geometric correction is used to correct geometric distortions in the image and align it with the geographic coordinate system; atmospheric correction removes the effects of atmospheric scattering and absorption, restoring the true reflectance of ground features. Preprocessing significantly improves image quality, making it more suitable for subsequent quality inspection and applications. The preset image processing rules are based on a deep understanding of the characteristics of remote sensing images, ensuring that the processed image achieves high standards in quality and accuracy. In a preferred embodiment of the invention, the preset image processing rules set corresponding correction parameters in the radiometric, geometric, and atmospheric correction steps, and perform radiometric, geometric, and atmospheric corrections according to these parameters. In a preferred embodiment of the invention, after acquiring the raw remote sensing image data, it is processed according to the preset image processing rules. First, the raw remote sensing image data is divided into multiple 256×256 pixel standard tiles, each with an independent coordinate identifier. Then, radiometric correction, geometric correction, and atmospheric correction are performed on each tile sequentially. Radiometric correction eliminates the effects of atmospheric and sensor noise by adjusting brightness values; geometric correction aligns the tiles with the geographic coordinate system by correcting geometric distortions; atmospheric correction restores the true reflectivity of ground features by removing atmospheric scattering and absorption. Finally, all corrected tiles are reassembled to obtain the processed image.

[0027] The processed image is subjected to quality inspection to obtain the quality inspection result of the processed image.

[0028] Specifically, quality inspection, through pixel-level defect detection, can accurately identify common defects in images such as cloud contamination, blurring, and color cast. For example, using an improved U-Net++ network to perform pixel-level defect detection on the processed image, the results show the following defects: cloud contamination (Tile_3_7 and Tile_4_8, affecting areas of 10% and 15% respectively), blurring (Tile_5_9, affecting area of ​​5%), and color cast (Tile_6_10, affecting area of ​​8%). Summarizing this defect information yields the quality inspection results for the processed image. These results are output in a structured format, including information such as defect type, defect coordinates, and defect area. This presentation of the quality inspection results not only allows for rapid problem localization but also provides detailed data for subsequent restoration, directly impacting the efficiency and effectiveness of image restoration.

[0029] Based on the quality inspection results, determine whether the processed image meets the qualification standard.

[0030] Specifically, acceptance criteria typically include indicators corresponding to defect type, total defect area, and defect density. By comparing the quality inspection results with preset acceptance criteria, it is possible to accurately determine whether the image meets application requirements. If the image does not meet the acceptance criteria, it needs to be repaired; if it meets the criteria, it can be directly output. This judgment process not only improves the automation level of production but also ensures the consistency of the quality of the produced images.

[0031] When the processed image fails to meet the qualification standard, the processed image is partially repaired according to the quality inspection results until the processed image meets the qualification standard.

[0032] Specifically, local restoration, when an image fails to meet quality standards, involves repairing only the defective areas based on quality inspection results, rather than repeatedly processing the entire image. This local restoration strategy not only saves computational resources and time but also avoids new problems that global restoration might introduce. The selection of the restoration algorithm and the optimization of its parameters are crucial to ensuring the restoration effect. Through an intelligent restoration mechanism, corresponding restoration schemes can be generated based on different defect types and characteristics. The implementation of local restoration makes resource utilization more rational in the image production process, significantly improving production efficiency.

[0033] When the processed image meets the qualification standard, the processed image is used as the output image.

[0034] Specifically, once the processed image meets the acceptable standards, it is used as the output image. The output image undergoes rigorous quality testing and precise local restoration, ensuring its quality. This high-quality output image better meets the needs of various application scenarios, providing accurate and reliable data support for agricultural monitoring, urban planning, and environmental assessment. Remote sensing images produced using this method not only show a significant improvement in quality but also achieve a qualitative leap in production efficiency. The high-quality and high-efficiency production of the output image provides strong technical support for the widespread application of remote sensing imagery in various fields.

[0035] The remote sensing image production method of this invention significantly improves the efficiency and quality of remote sensing image production through a systematic processing flow and an intelligent repair mechanism. Specifically, in the remote sensing image production process, raw remote sensing image data is first acquired and processed according to preset image processing rules to obtain preliminary processed images. These preset image processing rules are based on long-term production experience, effectively improving the basic quality of the images. Subsequently, the processed images undergo quality inspection to obtain detailed inspection results. These results not only accurately determine whether the processed images meet the acceptance standards but also provide basic data for subsequent repair work. Quality inspection quickly identifies problems in the processed images, such as cloud contamination, blurring, and color cast, providing a basis for targeted repair and avoiding the resource waste and time consumption caused by indiscriminate repair of the entire image in traditional methods. When the processed image does not meet the acceptance standards, local repair is performed based on the quality inspection results until the image meets the acceptance standards. Local repair is based on the accurate identification and location of defects, repairing only the problematic parts rather than repeatedly processing the entire image. This method not only reduces the computational resources and time required for restoration but also effectively avoids new problems that may be introduced by global restoration. The implementation of local restoration makes resource utilization more rational in the image production process, significantly improving production efficiency. This invention introduces an intelligent restoration mechanism into the local restoration process. By analyzing quality inspection results, it automatically restores the processed images, thereby achieving optimal restoration results. This not only improves the success rate of restoration but also further enhances the quality of the produced images. Compared with traditional manual restoration or fixed-pattern restoration, the intelligent restoration of this invention can better adapt to different types of defects, ensuring that the restored images meet requirements in terms of quality and detail. In summary, the remote sensing image production method of this invention, through precise quality inspection, local restoration strategies, and an intelligent restoration mechanism, not only improves production efficiency but also ensures the quality of the produced images.

[0036] Optionally, processing the original remote sensing image data according to preset image processing rules to obtain a processed image includes: According to the division parameters of the preset image processing rules, the original remote sensing image data is divided into multiple tiles of the same size and specifications; According to the correction parameters of the preset image processing rules, each tile is sequentially subjected to radiometric correction, geometric correction and atmospheric correction to obtain the corrected tile; The corrected tiles are combined to obtain the processed image.

[0037] Specifically, in the remote sensing image production method of this invention, the original remote sensing image data is divided into multiple tiles, each with the same size and specifications, and the tiles are divided according to the geocoding of the image data and preset division rules. Tile division can effectively break down the large amount of data in the original image, thereby improving the efficiency and flexibility of image processing. Through tile division, the image data is easier to process in parallel and distributed storage, greatly improving the efficiency and quality of image processing. Then, according to preset image processing rules, radiometric correction, geometric correction, and atmospheric correction are performed on each tile sequentially based on correction parameters. Among them, radiometric correction parameters include sensor response characteristics, solar illumination conditions, atmospheric conditions, etc. These parameters are used to adjust the radiance value of the image, eliminate the influence of atmospheric and sensor noise, and obtain the true radiance information of ground objects. Geometric correction parameters cover satellite orbital parameters, Earth curvature, terrain undulation, etc. This information is used to correct geometric distortions in the image, align the image with the geographic coordinate system, and ensure the geometric accuracy of the image. Atmospheric correction parameters include atmospheric molecular scattering, aerosol scattering, water vapor content, etc. These parameters are used to remove the influence of the atmosphere on imaging, restore the actual reflectance of ground objects, and improve the spectral accuracy of the image. Radiometric correction aims to eliminate or reduce the influence of atmospheric conditions, sensor performance, and other factors on image brightness values, thereby improving the radiometric accuracy of the image. Geometric correction corrects geometric distortions in the image, aligning it with the geographic coordinate system and enhancing the geometric accuracy of the image. Atmospheric correction further removes the effects of atmospheric scattering and absorption, restoring the true reflectance of ground objects and improving the spectral accuracy of the image. After correcting each tile, these corrected tiles are recombined to obtain the final processed image. This processing method effectively improves the overall quality and accuracy of the image, providing high-quality input for subsequent quality inspection and local repair steps. In a preferred embodiment of the invention, parallel processing is used for each tile, particularly through distributed image preprocessing, which significantly improves the efficiency of remote sensing image production. Specifically, parallel processing allows different tiles of the image to be distributed across multiple processors or computing nodes for simultaneous processing, which greatly shortens the overall image processing time. For distributed image preprocessing, the tiled image data is distributed across different computing units. Each computing unit independently executes preset image processing rules on the tiles assigned to it, including radiometric correction, geometric correction, and atmospheric correction. This distributed processing approach fully leverages the advantages of multi-core processors or cluster computing, achieving efficient task allocation and resource utilization.

[0038] In this embodiment of the invention, tile division enables parallel image processing, significantly reducing processing time. By performing radiometric, geometric, and atmospheric corrections on each tile, the radiometric, geometric, and spectral accuracy of the image can be effectively improved, thereby enhancing the overall image quality. Recombining the corrected tiles results in a processed image that better meets the needs of subsequent quality inspection and local restoration, improving the efficiency and output quality of the entire image production process.

[0039] Optionally, the step of performing quality inspection on the processed image to obtain the quality inspection result of the processed image includes: Pixel-level defect detection is performed on the processed image to obtain defect information for each tile in the processed image. The defect information includes defect type, defect coordinates, and defect area. The defect type, defect coordinates, and defect area of ​​each tile are summarized to obtain the defect type set, total defect area, and defect density of the processed image. The set of defect types, the total defect area, and the defect density of the processed image are used as the quality inspection result of the processed image.

[0040] Specifically, the quality inspection process includes pixel-level defect detection of the processed image to accurately identify defect information in each tile. During this process, various defect types can be detected, such as cloud contamination, blurring, and color cast, and the specific coordinates and area occupied by these defects within the tile can be determined. This can be achieved using image analysis techniques, such as deep learning algorithms. In a preferred embodiment of the invention, an improved U-Net++ network can be used for pixel-level defect segmentation, outputting structured defect information (JSON format). After acquiring the defect information, the defect type, defect coordinates, and defect area of ​​each tile are summarized. This involves not only simple data collection but also data integration and analysis to generate a comprehensive quality report for the entire processed image. Specifically, the summary results are displayed as a defect type set, total defect area, and defect density. The defect type set lists all defect types present in the image; the total defect area is the sum of the areas of all defective regions; and the defect density is the frequency of defect occurrence per unit area. These quality indicators collectively constitute the quality inspection results of the processed image, providing precise guidance for subsequent image restoration and optimization.

[0041] In a preferred embodiment of the invention, a pre-processed remote sensing image is subjected to quality inspection. This image, sourced from the GF-2 satellite, covers geographic information of a specific region, has a resolution of 1 meter, and is in TIFF format. First, the image is divided into multiple 256×256 pixel tiles, each with independent coordinates. Then, a modified U-Net++ network based on deep learning is used to perform pixel-level defect detection on each tile. The detection results show that some tiles have defects such as cloud contamination, blurring, and color cast. For example, cloud contamination was detected at Tile_3_7 and Tile_4_8, affecting areas of 10% and 15%, respectively; blurring was detected at Tile_5_9, affecting an area of ​​5%; and color cast was detected at Tile_6_10, affecting an area of ​​8%. This defect information is summarized to form a defect type set (cloud contamination, blurring, color cast), a total defect area (38%), and a defect density (0.38 defective tiles per square kilometer). These detailed quality inspection results provide precise guidance for subsequent local repairs, ensuring the efficiency and targeted nature of the repair work. This pixel-level defect detection and comprehensive quality assessment allows for accurate evaluation of image quality and provides reliable data support for subsequent repair and optimization.

[0042] In this embodiment of the invention, pixel-level defect detection accurately identifies various defects in tiles, including their type, location, and size, improving the meticulousness of quality inspection and ensuring that even minute defects are not missed. Secondly, by summarizing the defect information of each tile, a comprehensive quality inspection result is generated, providing operators with an intuitive and comprehensive overview of image quality. This not only helps to quickly locate problem areas but also guides the prioritization and specific methods of repair work, thereby achieving efficient and accurate image restoration.

[0043] Optionally, determining whether the processed image meets the qualification standard based on the quality inspection result includes: The quality inspection results are compared with multiple preset standards in the qualification standards to obtain the comparison results corresponding to the defect type set, the total defect area, and the defect density, respectively. Based on the comparison results, it is determined whether the processed image meets the qualification standard; Specifically, if any defect type in the defect type set does not belong to the qualified type in the preset standard, or the total defect area is greater than or equal to the maximum qualified area in the preset standard, or the defect density is greater than or equal to the maximum qualified density in the preset standard, the processed image is determined to have failed to meet the qualified standard. When all the defect types in the defect type set belong to the qualified type, the total defect area is less than the maximum qualified area, and the defect density is less than the maximum qualified density, the processed image is determined to meet the qualified standard.

[0044] Specifically, determining whether a processed image meets the acceptance criteria is a crucial step, involving comparing the quality inspection results with multiple preset standards within the acceptance criteria. The quality inspection results encompass key indicators such as the defect type set, total defect area, and defect density, while the acceptance criteria are set with corresponding thresholds based on actual application needs, such as the allowable range of defect types, the maximum acceptable area, and the maximum acceptable density. By comparing these indicators one by one, the quality status of the processed image can be accurately assessed. For example, if defect types outside the preset standards are found in the defect type set, such as color cast, or if the total defect area exceeds the preset maximum acceptable area, or if the defect density exceeds the preset maximum acceptable density, then the processed image can be definitively determined not to meet the acceptance criteria. Conversely, if all defect types are within the allowable range, and the total defect area and density are both below the preset thresholds, then the image can be determined to be acceptable. In a preferred embodiment of the invention, the quality inspection results of a certain processed image show that the defect type set includes cloud contamination and blurring, the total defect area is 18%, and the defect density is 0.15 defect tiles per square kilometer. The preset acceptance criteria are: allowed defect types are cloud contamination and blurring, the maximum acceptable area is 20%, and the maximum acceptable density is 0.2 defective tiles per square kilometer. Based on the above judgment logic, the quality inspection results are compared with the acceptance criteria one by one. Since all defect types in the image are within the allowable range, and the total defect area (18%) is less than the maximum acceptable area (20%), and the defect density (0.15) is also less than the maximum acceptable density (0.2), the processed image is ultimately determined to meet the acceptance criteria and can be directly used as output image for subsequent applications.

[0045] In this embodiment of the invention, by comparing the quality inspection results with the acceptance standards in detail, a comprehensive and meticulous evaluation of the processed image quality can be achieved. This not only effectively identifies various problems in the images but also accurately determines whether these problems are within acceptable limits, thereby avoiding the risk of substandard images flowing into subsequent application stages. Secondly, by comparing the quality inspection results with the acceptance standards in detail, clear quality control standards are established for image production, making the production process more standardized and regulated. Production personnel can conduct unified quality inspection and evaluation of images according to the preset acceptance standards, ensuring that each produced image meets the predetermined quality requirements.

[0046] Optionally, the step of performing local repair on the processed image based on the quality inspection results until the processed image meets the qualification standard includes: Based on the quality inspection results, determine the coordinates of the defects in the processed image; Based on the defect coordinates, the defective tiles in the processed image are determined; Based on the defect information in the defective tile, a local repair task corresponding to the defective tile is generated; The local repair task is performed on the defective tile until the processed image meets the acceptance criteria.

[0047] Specifically, in the remote sensing image production method of this invention, local repair of substandard processed images first involves accurately locating the defect coordinates in the image based on quality inspection results. These defect coordinates directly guide the repair operation to the specific location of the defect, avoiding blind processing of the entire image and greatly improving repair efficiency. Next, based on the defect coordinates, defective tiles can be identified. Subsequently, for each defective tile, a corresponding local repair task is automatically generated based on detailed defect information. The local repair tasks for different tiles are different, established based on the unique defect characteristics of each tile. These tasks encompass all aspects, from the selection of repair algorithms to the setting of specific repair parameters, ensuring the accuracy and effectiveness of the repair work. For example, for cloud contamination issues, a spatiotemporal interpolation algorithm might be used, combined with information from surrounding cloudless areas for repair; while for blurring issues, a deep learning-based super-resolution reconstruction algorithm might be employed. During the actual repair process, these local repair tasks are executed one by one, and the image quality is re-inspected after each repair to ensure the repair effect meets expectations. This closed-loop feedback mechanism continues to operate until the entire image reaches the preset qualification standard.

[0048] In a preferred embodiment of the present invention, firstly, the defect coordinates are accurately located based on the quality inspection results, and defective tiles are identified. For example, Tile_3_7 has cloud contamination affecting 10% of the area, and Tile_5_9 has blurriness affecting 5% of the area. Subsequently, local repair tasks are generated for these tiles according to the defect type. For the cloud-contaminated tile Tile_3_7, the STI-AG spatiotemporal interpolation algorithm is invoked, utilizing the texture information of surrounding cloudless tiles and time-series reference frames to dynamically generate repair parameters and perform the repair operation. For the blurry tile Tile_5_9, the attention-guided reconstruction algorithm is enabled, focusing on the edge and texture areas for repair. After the repair is completed, the repaired tiles are automatically re-inspected. If the total defect area of ​​the repaired image is reduced to 18% and the defect density is reduced to 0.15 defective tiles per square kilometer, meeting the qualification standard, the repair is considered successful, and the repaired image is used as the final output image.

[0049] In this embodiment of the invention, by accurately locating the coordinates of defects and identifying defective tiles, the invention can achieve precise repair of problem areas, avoiding the resource waste and time consumption caused by repeatedly processing the entire image in traditional methods. This not only significantly reduces the computational resources required for repair but also effectively avoids new problems that may be introduced by global repair, thereby significantly improving production efficiency.

[0050] Optionally, generating a local repair task corresponding to the defective tile based on the defect information in the defective tile includes: Based on the defect type in the defective tile, determine the corresponding repair algorithm and repair parameters for the defective tile; The defect coordinates and defect area in the defective tile are extracted to obtain a multidimensional feature vector of the defective tile. The multidimensional feature vector is input into a dynamic parameter model for prediction to obtain the target parameter value of the repair parameter. Based on the repair algorithm corresponding to the defective tile and the target parameter value of the repair parameter, the local repair task of the defective tile is constructed.

[0051] Specifically, the generation process of the local repair task demonstrates a high degree of intelligence and targeting. Firstly, the process involves identifying the defect type within the defective tile to determine the appropriate repair algorithm and parameters. For example, for cloud contamination defects, a spatiotemporal interpolation algorithm might be chosen, while for fuzzy defects, an attention-guided reconstruction algorithm might be employed. This algorithm selection mechanism ensures that the most suitable repair method is applied to different types of defects, thereby improving the repair effect. Next, feature extraction is performed on the defect coordinates and area within the defective tile to generate a multi-dimensional feature vector. By quantifying the location and extent of the defect, data support is provided for subsequent optimization of repair parameters. Subsequently, the extracted multi-dimensional feature vector is input into a dynamic parameter model for prediction, yielding the target parameter values ​​for the repair. This dynamic parameter model is typically based on machine learning techniques, such as the LightGBM gradient boosting decision tree algorithm, which can predict the optimal repair parameters based on historical data and defect feature vectors. Finally, based on the determined repair algorithm and the predicted target parameter values, a local repair task for the defective tile is constructed. This method of generating local restoration tasks not only allows for customized restoration based on the specific defect characteristics of each tile, but also enables dynamic adjustment of restoration parameters to adapt to restoration needs in different scenarios, thereby significantly improving the efficiency and quality of image restoration. Specifically, the TPR-Engine, a dual-engine collaborative control center, can be used to generate tile-level local restoration tasks.

[0052] In a preferred embodiment of the present invention, the defect type in the defective tile is first determined based on the quality inspection results. For cloud contamination defects, the STI-AG spatiotemporal interpolation algorithm is selected as the repair algorithm, and initial repair parameters are determined based on the defect area and coordinates. For fuzzy defects, the attention-guided reconstruction algorithm is selected, and initial repair parameters are extracted based on the defect features. Next, feature extraction is performed on the defect coordinates and area in the defective tile to generate a multi-dimensional feature vector. For example, for Tile_3_7, the feature vector includes defect coordinates (3,7), defect area (10%), defect type (cloud contamination), and other relevant features. Similarly, for Tile_5_9, the feature vector includes defect coordinates (5,9), defect area (5%), defect type (fuzzy), etc. These multi-dimensional feature vectors are input into a dynamic parameter model for prediction. The dynamic parameter model is based on the LightGBM gradient boosting decision tree algorithm, which predicts the optimal repair parameters by analyzing historical repair data and defect features. For example, for Tile_3_7, the model predicts a threshold of 5 for the number of neighboring cloudless tiles, 3 for the time series frames, and a fusion weight coefficient of 0.7. For Tile_5_9, the model predicts an edge enhancement coefficient of 1.2, a texture fidelity weight of 0.8, and 5 iterations. Based on the predicted target values ​​of the repair parameters and the selected repair algorithm, a local repair task is constructed for each defective tile. For example, for Tile_3_7, the repair task includes calling the STI-AG spatiotemporal interpolation algorithm to repair using the predicted parameters. For Tile_5_9, the repair task includes calling the attention-guided reconstruction algorithm to repair using the predicted parameters. The repair tasks are described in structured data format, including target tile identifier, repair algorithm type, dynamic repair parameters, and priority information.

[0053] In this embodiment of the invention, by determining the repair algorithm and parameters according to the defect type, precise repair of different defects can be achieved, avoiding the problem of poor repair results caused by fixed repair algorithms and parameters in traditional methods. The repair strategy is dynamically adjusted according to the specific characteristics of the defect, thereby significantly improving the success rate of repair and image quality. Furthermore, by extracting multi-dimensional feature vectors of defect coordinates and area and inputting them into a dynamic parameter model for prediction, more accurate target values ​​for repair parameters can be obtained, further improving the reliability and stability of the repair effect.

[0054] Optionally, determining the repair algorithm and repair parameters corresponding to the defective tile based on the defect type in the defective tile includes: When the defect type is cloud pollution, the repair algorithm is determined to be a space-time interpolation algorithm, and the threshold of the number of cloudless tiles in the defective tiles, the number of time series frames, and the fusion weight coefficient are used as the repair parameters. When the defect type is fuzzy, the repair algorithm is determined to be an attention-guided reconstruction algorithm, and the edge enhancement coefficient, texture fidelity weight and number of iterations in the defect tile are used as the repair parameters; When the defect type is color cast, the repair algorithm is determined to be a color migration correction algorithm, and the reference color gamut range, brightness scaling factor, and saturation correction coefficient in the defective tile are used as the repair parameters.

[0055] Specifically, when dealing with cloud contamination, a spatiotemporal interpolation algorithm is chosen for restoration. This algorithm utilizes information from adjacent cloudless tiles in the time series to fill in cloud-contaminated areas through interpolation. Restoration parameters include a threshold for the number of cloudless tiles, determining the number of cloudless areas referenced during interpolation; the number of time series frames, which involves the number of image frames referenced in the time dimension to ensure temporal consistency of the restoration results; and a fusion weight coefficient to control the mixing ratio between the original and interpolated data, ensuring a natural transition between the restored area and its surrounding environment. For blurring defects, an attention-guided reconstruction algorithm is selected as the restoration method. This algorithm uses a deep learning model to pay special attention to edge and texture details in the image, thereby enhancing image sharpness. Edge enhancement coefficients in the restoration parameters are used to strengthen edge features, making boundaries more distinct; texture fidelity weights ensure that while improving sharpness, the authenticity and naturalness of the original texture are maintained; and the number of iterations determines the number of times the algorithm runs to achieve the best deblurring effect. When dealing with color cast issues, a color migration correction algorithm is selected. This algorithm adjusts the color distribution of the image to match the reference color gamut. The reference color gamut range in the repair parameters defines the standard for the target color; the brightness scaling factor is used to adjust the overall brightness of the image to match the brightness level of the target color gamut; and the saturation correction coefficient fine-tunes the color saturation to ensure the accuracy and consistency of the image colors.

[0056] In this embodiment of the invention, the spatiotemporal interpolation algorithm effectively utilizes time-series information to generate natural and realistic restoration results when dealing with cloud contamination. The attention-guided reconstruction algorithm significantly enhances image details and improves image clarity and readability when addressing blur issues. The color migration correction algorithm accurately adjusts color distribution and restores the true colors of the image when dealing with color casts. By matching the most suitable restoration algorithm and its parameters to each defect type, the restoration quality and efficiency are significantly improved.

[0057] Optionally, performing the local repair task on the defective tile includes: When a defective tile exists in the processed image, the local repair task corresponding to the defective tile is executed directly. When there are multiple defective tiles in the processed image, a priority score for the local repair task is obtained by using a multi-dimensional priority model based on the defect information of the defective tiles corresponding to the local repair task. The local repair tasks are sorted according to the priority scores to obtain the task execution order, and the local repair tasks of all the defective tiles are executed according to the task execution order.

[0058] Specifically, when only one defective tile exists in the image, the corresponding local repair task is executed directly without additional scheduling operations, ensuring the efficiency of the repair process. However, when multiple defective tiles exist, a multi-dimensional priority model is introduced to optimize the execution order of repair tasks. Specifically, the multi-dimensional priority model comprehensively considers various information about the defective tile, including defect type, defect area, and defect coordinates, to calculate the priority score for each local repair task. For example, cloud contamination defects may be given a higher priority due to their significant impact on the overall image quality, or blurry defects appearing in critical areas of the image (such as urban areas or farmland) will also have their repair task priority increased accordingly. Through the multi-dimensional priority model, the repair order can be rationally arranged based on factors such as the severity and location of defects, prioritizing the defects with the greatest impact on image quality, thereby achieving efficient resource utilization and maximizing the repair effect. After determining the task execution order, the local repair tasks for all defective tiles will be executed sequentially according to this order, ensuring that each defect is properly handled, ultimately bringing the entire image to a qualified standard.

[0059] In a preferred embodiment of the invention, when performing a local repair task, multiple defective tiles are first identified. To perform repair efficiently, a multi-dimensional priority model is used to determine the priority score of each local repair task. The multi-dimensional priority model comprehensively considers the defect type (cloud contamination, blurring, and color cast), defect area (cloud contamination 10%, blurring 5%, color cast 8%), and defect coordinates (Tile_3_7, Tile_5_9, Tile_6_10). Among these, cloud contamination repair has a medium complexity, blurring repair has a high complexity, and color cast repair has a low complexity. The historical delay ratio for cloud contamination repair is 10%, for blurring repair it is 20%, and for color cast repair it is 5%.

[0060] The weight formula for the multidimensional priority model is: ; The weights are respectively , , , Calculate the priority score for each defective tile: Regarding Tile_3_7 (cloud contamination), urgency level Defect area ratio Repairing complexity Historical delay ratio The priority score is: ; For Tile_5_9 (fuzzy), urgency level Defect area ratio Repairing complexity Historical delay ratio The priority score is: ; For Tile_6_10 (color cast), urgency level Defect area ratio Repairing complexity Historical delay ratio The priority score is: ; Based on the calculation results, the priority scores are sorted from highest to lowest as follows: Tile_5_9 (0.475)>Tile_3_7 (0.48)>Tile_6_10 (0.309); Therefore, the task execution order is as follows: Tile_5_9 → Tile_3_7 → Tile_6_10.

[0061] In this embodiment of the invention, by comprehensively considering various defect information, the defects that most need priority repair can be accurately identified, ensuring that critical issues are resolved in a timely manner. The priority-based repair strategy not only improves repair efficiency but also avoids wasting resources, allowing limited computing resources to be utilized to the maximum extent for repair tasks that have the greatest impact on image quality.

[0062] Optionally, it also includes: When the processed image fails to meet the qualification standard, the number of defective tiles in the processed image is obtained; When the number of defective tiles is greater than or equal to a preset number threshold, spatiotemporal clustering is performed on the defect information of all defective tiles to obtain the defect pattern of the processed image. Based on the defect pattern, the correction parameters in the preset image processing rules are updated, and the updated preset image processing rules are applied in real time to the subsequent processing of the original remote sensing image data.

[0063] Specifically, when processed images fail to meet the acceptable standards, the number of defective tiles is further analyzed. If the number of defective tiles exceeds a preset threshold, it indicates a widespread quality problem in the image. At this point, spatiotemporal clustering analysis is performed on the defect information of all defective tiles. This process aims to identify defect groups with similar characteristics and distribution patterns from a large amount of defect data. By analyzing the temporal and spatial distribution characteristics of these defects, potential defect patterns are uncovered. For example, certain areas may frequently experience a certain type of defect due to specific atmospheric conditions or sensor performance issues. This spatiotemporal clustering analysis not only considers the type and location of defects but also incorporates multi-dimensional information such as imaging time and season, thus enabling a more comprehensive understanding of the causes and distribution patterns of defects. Based on the analyzed defect patterns, the correction parameters in the preset image processing rules are updated. This update is targeted; for example, if cloud pollution is frequently found in a certain area, the atmospheric correction parameters are adjusted to increase compensation for the impact of clouds on that area. The updated correction parameters are applied in real time to subsequent image processing, thereby improving processing efficiency and quality and reducing the probability of similar defects occurring in future production. This mechanism embodies the self-learning and self-optimization capabilities of this invention, enabling the entire production system to automatically adjust based on problems encountered in actual production, continuously optimize the production process, and improve image quality. Specifically, the PPSO-Engine, a dual-engine collaborative control center, can be used to perform reverse optimization of preset image processing rules.

[0064] In a preferred embodiment of the present invention, when the processed image fails to meet the qualification standard, the number of defective tiles is first obtained, and 12 defective tiles are found, exceeding a preset threshold (e.g., the threshold is set to 10). This indicates that there are relatively common quality problems in the image, requiring further analysis. To gain a deeper understanding of the distribution patterns of these defects, spatiotemporal clustering analysis is performed on the defect information of all defective tiles. Specifically, an improved Geo-DBSCAN algorithm is used for spatiotemporal clustering analysis, which collects information such as defect type, defect coordinates, defect area, and imaging time for each defective tile. Through the spatiotemporal clustering algorithm, the following defect patterns are identified: Atmospheric disturbance pattern, with frequent cloud pollution in the northwest region of the image (Tile_3_7, Tile_4_8, Tile_5_9), and the imaging time of these tiles is concentrated in spring. Blurred pattern, with blurry defects appearing in the southeast region of the image (Tile_6_10, Tile_7_11, Tile_8_12), and the imaging time of these tiles is concentrated in summer. A color cast pattern was observed in the central region of the image (Tile_9_13, Tile_10_14, Tile_11_15), with these tiles primarily imaged in autumn. Based on these defect patterns, the correction parameters in the preset image processing rules were updated. These included: for the cloud pollution pattern, increasing the water vapor compensation factor in the atmospheric correction parameters from the default value of 1.0 to 1.5 to better handle cloud pollution under high water vapor conditions in spring; for the blur pattern, adjusting the focus parameter in the geometric correction parameters from the default value of 0.8 to 0.95 to improve the sharpness of summer images; and for the color cast pattern, optimizing the color balance parameter in the radiometric correction parameters from the default value of 0.5 to 0.7 to improve color accuracy in autumn images. The updated correction parameters will be applied in real-time during subsequent image processing. For example, when processing new GF-2 satellite imagery, these optimized parameters will be automatically applied to better correct potential defects during preprocessing, thereby improving the overall image quality.

[0065] In this embodiment of the invention, by acquiring the number of defective tiles and setting a preset threshold, batch quality issues requiring focused attention can be effectively identified, avoiding overreaction to sporadic defects and improving response efficiency to large-scale quality problems. Secondly, spatiotemporal clustering analysis provides a deeper understanding of defect distribution patterns, not only helping to identify specific regions or conditions where defects frequently occur, but also providing a scientific basis for optimizing correction parameters. In this way, image processing rules can be adjusted in a targeted manner, thereby reducing the occurrence of similar defects in subsequent processing and improving overall production efficiency and image quality.

[0066] In summary, in the preferred embodiment of the present invention, combined with Figure 2As shown, the defect data received from the QDLM (Quality Inspection and Defect Location Module) serves as the trigger point for initiating the repair process. Next, the defect data is parsed to extract the specific coordinates of the defective tiles (Tile_X_Y), thereby accurately locating the defective areas in the image. Based on this defect information, repair parameters are dynamically generated. These parameters are tailored to specific defects (such as cloud contamination or blurring) to guide subsequent repair operations. To quickly verify the repair effect, only the repaired tiles can be output instead of the entire image.

[0067] Combination Figure 3 As shown, another embodiment of the remote sensing image production system of the present invention includes: The raw data acquisition unit is used to acquire raw remote sensing image data; The preprocessing unit is used to process the original remote sensing image data according to preset image processing rules to obtain a processed image; A quality inspection unit is used to perform quality inspection on the processed image and obtain the quality inspection result of the processed image; The judgment unit is used to determine whether the processed image meets the qualification standard based on the quality inspection results. The repair unit is used to perform local repair on the processed image according to the quality inspection results when the processed image fails to meet the qualification standard, until the processed image meets the qualification standard. The output unit is used to output the processed image as the output image when the processed image meets the qualification standard.

[0068] The remote sensing image production system of the present invention has the same advantages over the prior art as the aforementioned remote sensing image production method, and will not be repeated here.

[0069] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A method for producing remote sensing images, characterized in that, include: Acquire raw remote sensing image data; The original remote sensing image data is processed according to preset image processing rules to obtain the processed image of the original remote sensing image data. The processed image is subjected to quality inspection to obtain the quality inspection result of the processed image; Based on the quality inspection results, determine whether the processed image meets the qualification standard; When the processed image fails to meet the qualification standard, the processed image is partially repaired according to the quality inspection results until the processed image meets the qualification standard. When the processed image meets the qualification standard, the processed image is used as the output image.

2. The remote sensing image production method according to claim 1, characterized in that, The step of processing the original remote sensing image data according to preset image processing rules to obtain a processed image includes: According to the division parameters of the preset image processing rules, the original remote sensing image data is divided into multiple tiles of the same size and specifications; According to the correction parameters of the preset image processing rules, each tile is sequentially subjected to radiometric correction, geometric correction and atmospheric correction to obtain the corrected tile; The corrected tiles are combined to obtain the processed image.

3. The remote sensing image production method according to claim 2, characterized in that, The process of performing quality inspection on the processed image to obtain the quality inspection result of the processed image includes: Pixel-level defect detection is performed on the processed image to obtain defect information for each tile in the processed image. The defect information includes defect type, defect coordinates, and defect area. The defect type, defect coordinates, and defect area of ​​each tile are summarized to obtain the defect type set, total defect area, and defect density of the processed image. The set of defect types, the total defect area, and the defect density of the processed image are used as the quality inspection result of the processed image.

4. The remote sensing image production method according to claim 3, characterized in that, The step of determining whether the processed image meets the qualification standard based on the quality inspection results includes: The quality inspection results are compared with multiple preset standards in the qualification standards to obtain the comparison results corresponding to the defect type set, the total defect area, and the defect density, respectively. Based on the comparison results, it is determined whether the processed image meets the qualification standard; Specifically, if any defect type in the defect type set does not belong to the qualified type in the preset standard, and / or the total defect area is greater than or equal to the maximum qualified area in the preset standard, and / or the defect density is greater than or equal to the maximum qualified density in the preset standard, the processed image is determined to have failed to meet the qualified standard. When all the defect types in the defect type set belong to the qualified type, the total defect area is less than the maximum qualified area, and the defect density is less than the maximum qualified density, the processed image is determined to meet the qualified standard.

5. The remote sensing image production method according to claim 3, characterized in that, The step of performing local repair on the processed image based on the quality inspection results until the processed image meets the qualification standard includes: Based on the quality inspection results, determine the coordinates of the defects in the processed image; Based on the defect coordinates, the defective tiles in the processed image are determined; Based on the defect information in the defective tile, a local repair task corresponding to the defective tile is generated; The local repair task is performed on the defective tile until the processed image meets the acceptance criteria.

6. The remote sensing image production method according to claim 5, characterized in that, The step of generating a local repair task corresponding to the defective tile based on the defect information in the defective tile includes: Based on the defect type in the defective tile, determine the corresponding repair algorithm and repair parameters for the defective tile; The defect coordinates and defect area in the defective tile are extracted to obtain a multidimensional feature vector of the defective tile. The multidimensional feature vector is input into a dynamic parameter model for prediction to obtain the target parameter value of the repair parameter. Based on the repair algorithm corresponding to the defective tile and the target parameter value of the repair parameter, the local repair task of the defective tile is constructed.

7. The remote sensing image production method according to claim 6, characterized in that, The step of determining the repair algorithm and repair parameters corresponding to the defective tile based on the defect type in the defective tile includes: When the defect type is cloud pollution, the repair algorithm is determined to be a space-time interpolation algorithm, and the threshold of the number of cloudless tiles in the defective tiles, the number of time series frames, and the fusion weight coefficient are used as the repair parameters. When the defect type is fuzzy, the repair algorithm is determined to be an attention-guided reconstruction algorithm, and the edge enhancement coefficient, texture fidelity weight and number of iterations in the defect tile are used as the repair parameters; When the defect type is color cast, the repair algorithm is determined to be a color migration correction algorithm, and the reference color gamut range, brightness scaling factor, and saturation correction coefficient in the defective tile are used as the repair parameters.

8. The remote sensing image production method according to claim 5, characterized in that, Performing the local repair task on the defective tile includes: When a defective tile exists in the processed image, the local repair task corresponding to the defective tile is executed directly. When there are multiple defective tiles in the processed image, a priority score for the local repair task is obtained by using a multi-dimensional priority model based on the defect information of the defective tiles corresponding to the local repair task. The local repair tasks are sorted according to the priority scores to obtain the task execution order, and the local repair tasks of all the defective tiles are executed according to the task execution order.

9. The remote sensing image production method according to claim 5, characterized in that, Also includes: When the processed image fails to meet the qualification standard, the number of defective tiles in the processed image is obtained; When the number of defective tiles is greater than or equal to a preset number threshold, spatiotemporal clustering is performed on the defect information of all defective tiles to obtain the defect pattern of the processed image. Based on the defect pattern, the correction parameters in the preset image processing rules are updated, and the updated preset image processing rules are applied in real time to the subsequent processing of the original remote sensing image data.

10. A remote sensing image production system, characterized in that, include: The raw data acquisition unit is used to acquire raw remote sensing image data; The preprocessing unit is used to process the original remote sensing image data according to preset image processing rules to obtain a processed image; A quality inspection unit is used to perform quality inspection on the processed image and obtain the quality inspection result of the processed image; The judgment unit is used to determine whether the processed image meets the qualification standard based on the quality inspection results. The repair unit is used to perform local repair on the processed image according to the quality inspection results when the processed image fails to meet the qualification standard, until the processed image meets the qualification standard. The output unit is used to output the processed image as the output image when the processed image meets the qualification standard.

Images