Radiation correction and terrain correction based remote sensing data preprocessing method and system

Abstract

This application discloses a remote sensing data preprocessing method and system based on radiometric correction and topographic correction, relating to the field of SAR remote sensing signal error correction technology. In the SAR data preprocessing process for forest resource monitoring, this scheme assesses the cloud cover status of SAR images and determines whether cloud and cloud shadow shielding processing is necessary based on the assessment's pass / fail status. After the SAR image assessment, SAR control point features are matched, and SAR data processing is determined based on the matching's pass / fail status. After SAR control point matching, the reliability of the SAR data is analyzed, and SAR image coordinate and data optimization is determined based on the analysis results. This solves the problem of limited applicability of remote sensing data in forest resource monitoring applications.

Description

Technical Field

[0001] This invention relates to the field of SAR remote sensing signal error correction technology, and in particular to a remote sensing data preprocessing method and system based on radiometric correction and terrain correction. Background Technology

[0002] With the rapid development and widespread application of remote sensing technology, its advantages of large-area synchronous observation, high timeliness, and relatively low data acquisition cost have been fully utilized. In particular, synthetic aperture radar, with its unique advantages of all-weather, all-day operation and strong penetration, has become a core technical means for remote sensing monitoring of forest resources and marine resources. As a core component of terrestrial ecosystems, forests are key carriers for global carbon cycling, biodiversity maintenance, and the construction of ecological security barriers. Therefore, conducting precise and efficient forest resource monitoring work, such as biomass estimation, vegetation cover inversion, and dynamic monitoring of forest stand structure, plays an important role in forestry management, ecological environmental protection, and climate change response.

[0003] Traditional forest monitoring relies on ground-based sample plot surveys. While highly accurate, these methods suffer from drawbacks such as being time-consuming, labor-intensive, costly, having limited spatial coverage, and difficulty in achieving large-scale dynamic updates. They are no longer sufficient to meet the demands of refined and routine forest resource monitoring in the new era. Against this backdrop, remote sensing technology, with its advantages of wide coverage, real-time dynamic observation, low cost, and high efficiency, has become the core technology for acquiring raw SAR (Synthetic Aperture Radar) data for forest resource monitoring, including parameters such as band, resolution, and imaging swath width. However, raw SAR data is susceptible to interference from various factors such as sensor performance, atmospheric conditions, and topographic relief, making it difficult to directly meet the requirements of monitoring and analysis. Therefore, improving the quality of SAR data preprocessing is a crucial prerequisite for ensuring the accuracy of forest monitoring.

[0004] Existing patents: For example, Chinese invention patent application CN116996112B discloses a real-time preprocessing method for remote sensing satellite data, which includes: by adopting a streaming processing architecture, transmitting data from the preprocessing stages of frame synchronization, descrambling, decoding, packetization, auxiliary data and image extraction, decompression, image stitching, relative radiometric correction, and system geometric correction in a slice streaming mode. Data in each processing stage is transmitted through an inter-server network or memory, replacing the traditional batch processing method of transmitting and exchanging data via disk. The data slice size and the parallelism of the SAR data sequential stream can be adjusted according to the actual performance test of the system.

[0005] For example, the Chinese invention patent application CN111404593B discloses a method and apparatus for processing satellite remote sensing data, which includes: acquiring satellite remote sensing broadcast data and external auxiliary data received by a front-end receiving device and obtained through preprocessing operations; acquiring pre-configured timing scheduling information based on the reception status information of the satellite remote sensing broadcast data; and standardizing the satellite remote sensing broadcast data by real-time scheduling the corresponding processing flow based on the timing scheduling information, and generating multi-level satellite remote sensing products based on the external auxiliary data; wherein the satellite remote sensing broadcast data includes: first-level broadcast data from various geostationary orbit radiation imagers, first-level broadcast data from interferometric atmospheric vertical sounders, and first-level broadcast data from lightning imagers.

[0006] Existing patents have constructed an efficient technical architecture around remote sensing satellite data preprocessing. They employ a streaming processing architecture to achieve sliced ​​streaming transmission and parallelism control, replacing the traditional disk batch processing mode. Furthermore, they utilize timed scheduling to achieve standardized processing of remote sensing broadcast data and multi-level product generation, providing core technical support for efficient SAR data processing. Building upon this foundation, for specific target application scenarios such as forest resource monitoring, further refined processing of the entire SAR data process is still required. The specific processing steps are as follows:

[0007] The system receives raw detection signals from SAR satellites via a satellite-to-ground radio frequency link from remote sensing satellites. After basic signal processing, including radio frequency signal demodulation, frame synchronization, and descrambling decoding, SAR images are acquired from archives or programming interfaces of SAR data publishing organizations such as EarthExplorer and ScienceBase. These SAR images directly contain raw digital quantized values ​​observed and recorded by satellite sensors, representing unprocessed relative records of ground object radiation energy, such as the raw digital values ​​of Sentinel-1 SAR images. SAR data with suitable spatiotemporal resolution and imaging range are selected and matched, while SAR data with excessive cloud cover or substandard quality is removed. Subsequently, auxiliary information such as digital elevation models, atmospheric parameters, and sensor parameters are collected. Atmospheric parameters refer to quantitative indicators characterizing the physical state and optical properties of the atmosphere along the remote sensing signal transmission path, such as atmospheric transmittance and atmospheric water vapor content. Sensor parameters refer to the inherent technical specifications and operating parameters of remote sensing sensors, such as optical sensors, radar sensors, and infrared sensors, such as spatial resolution and spectral resolution.

[0008] Next, the original digital quantization values ​​are converted into physically meaningful radiance values ​​using sensor parameters to complete system radiometric correction. Then, by combining atmospheric parameters and atmospheric radiative transfer models (such as the 6S model), the effects of atmospheric scattering and absorption are eliminated, and the surface reflectance is retrieved. Subsequently, geometric refinement is performed, and a rigorous rational function model or collinearity equation is constructed using a digital elevation model to perform orthorectification on the image. Based on this geometric fine correction, the SAR data acquired from the SAR imagery is then geometrically registered. Control point features are matched using matching algorithms (such as SIFT (Scale-Invariant Feature Transform) and SURF (Speeded Up Robust Features)). For example, in the registration of SAR images in forest areas, SIFT is used to extract control point features such as forest boundaries and topographic inflection points, ensuring that all SAR data pixels are strictly aligned spatially. To address the interference of topography on radiation, topographic radiation correction models (such as C-correction and SCS (Satellite Calibration Spectrometer) correction) are used in addition to orthorectification. Topographic factors such as slope and aspect derived from the digital elevation model are used to quantify and eliminate the differences in solar illumination caused by topographic undulations, thereby obtaining the topographically normalized surface reflectance. For SAR data, the inherent speckle noise can severely affect the analysis. Subsequently, adaptive filtering algorithms (such as Lee filtering and Gamma-MAP filtering) can be used to process the data as needed to effectively suppress speckle noise. Finally, according to the spatial resolution requirements of the specific monitoring task, scale unification is achieved through resampling, and the boundary vector file of the target area is used for precise cropping to obtain the final qualified and standardized SAR data.

[0009] Existing technologies rely on clear and stable control point features, such as forest boundaries and topographic inflection points, for geometric registration and on topographic factors derived from digital elevation models for radiometric correction. However, the shading and movement of clouds and cloud shadows above forest areas can interfere with the identification of control point features that can be used for registration. This leads to incomplete extraction of control point features and a significant increase in the mismatch rate of control point feature matching based on matching algorithms (such as SIFT and SURF). Consequently, SAR data cannot achieve pixel-level spatial alignment, further resulting in errors in pixel-level spatial alignment of SAR data. In addition, the dynamic shading effect of clouds and their shadows, combined with the static illumination differences caused by topographic undulations, makes it difficult for topographic radiometric correction models (such as C-correction and SCS correction) to accurately quantify and eliminate composite radiometric distortion. The failure to effectively eliminate radiometric distortion further exacerbates the errors in SAR data.

[0010] Based on this, pixel-level spatial alignment errors in SAR data may lead to disordered mapping relationships of pixel spatial coordinates, resulting in inconsistent spatial extent definition of target areas (specific forest areas, forest monitoring sample areas, ecological protection forest belts, etc.) during the scale unification process. At the same time, gray value anomalies caused by composite radiation distortion can disrupt the stability of local statistical distribution, leading to reduced accuracy in the selection of filtering parameters (such as window size and smoothing coefficient), making it impossible to effectively separate noise from effective signals, and even smoothing out real surface change details.

[0011] Ultimately, due to the accumulation of the aforementioned errors, the spatial coordinates of SAR data cannot be accurately matched with the geographic coordinates of the vector boundary. This leads to misjudgment of the target area during the cropping process, making it difficult to obtain standardized SAR data with accurate spatial range and qualified data quality. Consequently, the support for SAR data processing in remote sensing data is insufficient, ultimately limiting the applicability of remote sensing data in forest resource monitoring application scenarios. Summary of the Invention

[0012] This invention provides a remote sensing data preprocessing method and system based on radiometric correction and terrain correction, which can acquire standardized SAR data with accurate spatial range and qualified data quality, thereby improving the reliability of remote sensing data processing. The technical solution provided by this application is as follows:

[0013] Firstly, a remote sensing data preprocessing method based on radiometric correction and topographic correction is provided. The specific implementation of this method is as follows:

[0014] The cloud cover of SAR images is assessed, and the suitability of the assessment determines whether cloud and cloud shadow shielding processing is necessary. After the SAR image assessment, SAR control point features are matched, and the suitability of the matching determines whether processing to improve radiation distortion reduction is necessary. After the SAR control point matching is completed, the reliability of the SAR data is analyzed to reflect the suitability of the SAR data processing, and the analysis determines whether optimization to improve the cropping accuracy of the target area is necessary.

[0015] Secondly, a remote sensing data preprocessing system based on radiometric correction and topographic correction is provided. This system includes:

[0016] The system comprises three modules: a SAR image cloud cover assessment module, a SAR control point feature matching module, and a SAR data reliability analysis module. The SAR image cloud cover assessment module evaluates the cloud cover status of SAR images and determines whether cloud and cloud shadow shielding processing is necessary based on the assessment's pass / fail status. The SAR control point feature matching module matches SAR control point features after SAR image assessment and determines whether processing to improve radiation distortion reduction is necessary based on the matching's pass / fail status. The SAR data reliability analysis module analyzes the reliability of SAR data after SAR control point matching to reflect the pass / fail status of SAR data processing and determines whether optimization to improve target area cropping accuracy is necessary based on the analysis results.

[0017] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0018] 1. By assessing the cloud cover status of SAR images and determining whether cloud and cloud shadow shielding processing is necessary based on the assessment results, it helps to avoid the distortion interference of clouds and cloud shadows on the radiometric and topographic features of SAR images. This enhances the quality control capability of SAR data preprocessing data sources, enables precise screening and targeted processing of images with excessive cloud shadow interference, and thus achieves the goal of providing high-quality, cloud-free base images for subsequent radiometric and topographic correction. After meeting the assessment criteria, SAR control point features are matched, and the determination of whether SAR data processing is necessary based on the matching results can effectively improve the pixel-level spatial alignment accuracy of SAR data. This strengthens the reliability of spatial reference support for radiometric and topographic correction, enhances the quality control of control point extraction and matching, ensures the accuracy of the evaluation of radiometric distortion elimination effect, and analyzes the reliability of SAR data after meeting the matching qualification conditions. Based on the analysis, it is determined whether to optimize SAR image coordinates and data, which helps to accurately control the spatial positioning accuracy and target area coverage integrity of SAR data. It reflects the quality closed-loop control level of the entire SAR data preprocessing process, enhances the matching degree between preprocessed data and real geographic boundaries, and thus achieves the goal of outputting high-quality and high-reliability SAR data and ensuring the practical value of radiometric and topographic correction results.

[0019] 2. Cloud and cloud shadow shielding can effectively reduce the impact of clouds and cloud shadows on SAR images. Compared with existing technologies, the obscuring and movement of clouds and cloud shadows above forest areas can interfere with the identification of control point features that can be used for registration, resulting in incomplete extraction of control point features. This significantly increases the mismatch rate of control point feature matching based on matching algorithms (such as SIFT and SURF), making it impossible for SAR data to achieve pixel-level spatial alignment. This further leads to errors in the pixel-level spatial alignment of SAR data. This solution can reduce the interference of cloud shadows on the extraction of control point features in forest areas through precise cloud and cloud shadow shielding, ensuring the integrity and saliency of effective control point feature extraction, greatly reducing the mismatch rate of control point feature matching, ensuring that SAR data can stably achieve pixel-level spatial alignment and eliminate alignment errors.

[0020] 3. SAR data processing can improve the quality of SAR data. Compared with existing technologies, pixel-level spatial alignment errors in SAR data may lead to disordered mapping relationships of pixel spatial coordinates, resulting in inconsistent spatial range definition of target areas (specific forest areas, forest monitoring sample areas, ecological protection forest belts, etc.) during the scale unification process. At the same time, gray value anomalies caused by composite radiation distortion can disrupt the stability of local statistical distribution, leading to reduced accuracy in the selection of filtering parameters (such as window size and smoothing coefficient). This solution can achieve unified definition and accurate matching of the spatial range of the target area, while eliminating gray value anomalies caused by composite radiation distortion, restoring the stability of local statistical distribution, ensuring the accuracy and adaptability of filtering parameter selection, and improving the accuracy and reliability of SAR data preprocessing for forest resource monitoring target areas.

[0021] 4. SAR image coordinate and data optimization helps improve the accuracy of target area cropping. Compared with existing technologies, due to the accumulation of the above-mentioned errors, the spatial coordinates of SAR data and the geographic coordinates of the vector boundary cannot be accurately matched. This leads to misjudgment of the target area during the cropping process, making it difficult to obtain standardized SAR data with accurate spatial range and qualified data quality. This solution helps to achieve accurate alignment between the spatial coordinates of SAR data and the geographic coordinates of the vector boundary, thereby ensuring that the cropped area completely covers the target area and successfully obtaining standardized SAR data with accurate spatial range, stable radiation characteristics, and qualified data quality. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1This is the upper part of the flowchart of the remote sensing data preprocessing method based on radiometric correction and terrain correction provided in the embodiments of the present invention;

[0024] Figure 2 This is the second half of the flowchart of the remote sensing data preprocessing method based on radiometric correction and terrain correction provided in the embodiments of the present invention;

[0025] Figure 3 This is a block diagram of a remote sensing data preprocessing method based on radiometric correction and terrain correction provided in an embodiment of the present invention;

[0026] Figure 4 This is a flowchart of SAR data processing for a remote sensing data preprocessing method based on radiometric correction and terrain correction provided in an embodiment of the present invention.

[0027] Figure 5 This is a schematic diagram of the structure of a remote sensing data preprocessing system based on radiometric correction and terrain correction provided in an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0029] Before providing a detailed explanation of the embodiments of this application, the application scenarios of these embodiments will be described first.

[0030] Example 1: This embodiment of the invention provides a remote sensing data preprocessing method based on radiometric correction and topographic correction, such as... Figure 1 The flowchart shown is the upper part of the remote sensing data preprocessing method based on radiometric correction and terrain correction. Figure 2 This is a flowchart of the lower half of the remote sensing data preprocessing method based on radiometric correction and terrain correction provided in an embodiment of the present invention. The processing flow of this method may include the following steps:

[0031] First, a SAR image cloud cover assessment is performed to determine if the SAR image cloud cover is greater than or equal to the SAR image cloud cover reference value. If so, the corresponding SAR image is marked as a SAR image to be processed, and cloud and cloud shadow masking is performed. It is then determined whether the cloud and cloud shadow coverage of the re-acquired SAR image after cloud and cloud shadow masking is still greater than the corresponding threshold. If so, an abnormal cloud and cloud shadow masking processing alert is sent to designated personnel. Otherwise, SAR control point feature matching is performed. If the SAR image cloud cover is less than the SAR image cloud cover reference value, the corresponding SAR image is marked as a qualified SAR image, and SAR control point feature matching is performed. The SAR control point extraction result is compared with the preset SAR control point extraction result. If the SAR control point extraction result is greater than or equal to the preset SAR control point extraction result, the control point features are matched based on the preset matching algorithm. If the SAR control point extraction result is less than the preset SAR control point extraction result, the control point features are enhanced based on the preset multi-scale feature extraction algorithm. When matching control point features using a preset matching algorithm, the SAR control point matching rate is compared with the preset SAR control point matching rate. If the SAR control point matching rate is greater than the preset SAR control point matching rate, SAR data processing is performed. If the SAR control point matching rate is not greater than the preset SAR control point matching rate, terrain-constrained control point matching is performed. It is then determined whether the preset requirements are met after the terrain-constrained control point matching operation. If yes, SAR data processing continues; otherwise, a SAR control point matching anomaly alert is sent to the preset personnel. For SAR data processing, it is determined whether the root mean square deviation of the acquired SAR data is less than the preset SAR data root mean square deviation. If yes, SAR data reliability analysis is performed to determine whether the SAR overlap is greater than or equal to the SAR overlap calibration value and whether the boundary offset error is less than or equal to the preset boundary offset error. If yes, a SAR data processing qualified alert is sent to the preset personnel; otherwise, SAR data is processed based on the polarization decomposition method to output standardized SAR data based on coordinate unification and parameter optimization methods.

[0032] In this application, a database storing various types of preset data is established before the design of the remote sensing data preprocessing method based on radiometric correction and terrain correction. The database includes, but is not limited to, SAR image cloud shadow coverage reference values, preset number of control point features extracted, preset SAR control point matching rate, and SAR overlap calibration values.

[0033] The database contains data obtained through on-orbit observations of SAR satellites, ground range tests, and field surveys, as well as data thresholds and compliance requirements determined by industry standards in the field of SAR remote sensing. For extreme scenarios that are difficult to cover by actual data (such as high-altitude mountains, areas with strong cloud cover, and low-overlap imaging scenarios), simulation experiments are conducted using SAR imaging simulation software (such as SAR scape and Gamma SAR) to generate supplementary data.

[0034] This database adopts a hierarchical business logic data structure, which organizes data according to a four-layer architecture: basic configuration layer, correction parameter layer, quality assessment layer, and scenario adaptation layer. All data items are structured data with clearly defined field definitions, data types, accuracy requirements, and value ranges. It adapts to the full-process call requirements of SAR data preprocessing for radiometric correction and terrain correction, while also being compatible with data expansion and updates.

[0035] In addition, PostgreSQL / MySQL relational databases were selected as the core storage medium to store the above four layers of structured data (numerical and character parameters), which is suitable for the high-frequency, fast and accurate parameter retrieval and calling requirements in the SAR data preprocessing process.

[0036] In this embodiment, as Figure 3 The diagram shown illustrates a remote sensing data preprocessing method based on radiometric correction and topographic correction. SAR image cloud cover assessment: In the SAR data preprocessing process for forest resource monitoring, the cloud cover status of SAR images is assessed. Based on the assessment's pass / fail status, cloud and cloud shadow shielding processing is deemed necessary to reduce the impact of clouds and cloud shadows on SAR images. This enhances the quality control capability of SAR data preprocessing data sources and enables precise screening and targeted processing of images with excessive cloud shadow interference.

[0037] SAR control point feature matching: After the SAR image evaluation is completed, the SAR control point features are matched to reflect the completeness of control point extraction and the accuracy of matching. The matching results determine whether to perform SAR data processing to improve the radiation distortion elimination effect. This helps to improve the pixel-level spatial alignment accuracy of SAR data and strengthens the reliability of the spatial reference support for radiation correction and terrain correction.

[0038] SAR data reliability analysis: After SAR control point matching is completed, the reliability of SAR data is analyzed to reflect the qualification of SAR data processing. Based on the analysis, it is determined whether to optimize SAR image coordinates and data to improve the cropping accuracy of target areas. The target area refers to the area pre-set by the personnel to monitor forest resources. This helps to enhance the matching degree between pre-processed data and real geographical boundaries, thereby achieving the output of high-quality and high-reliability SAR data.

[0039] Furthermore, the cloud cover of SAR images is assessed, and the specific process is as follows: Based on the ratio of the area of ​​the cloud-covered region to the area of ​​the entire SAR image obtained by the all-sky imager, the corresponding ratio is expressed as the SAR image cloud cover degree used to evaluate the interference of the cloud-covered region; if the SAR image cloud cover degree is greater than or equal to the average SAR image cloud cover degree required by preset personnel for historical remote sensing data processing as the SAR image cloud cover reference value, then the corresponding SAR image is marked as a SAR image to be processed and cloud and cloud cover shielding processing is performed; otherwise, the corresponding SAR image is marked as a qualified SAR image and SAR control point feature matching is performed.

[0040] Furthermore, the specific process for cloud and cloud shadow shielding is as follows:

[0041] The preset monitoring time points are randomly selected by preset personnel within the time period for forest resource monitoring. The SAR images to be processed are then input into a preset deep learning (such as U-Net, MaskR-CNN) cloud detection model to help distinguish between clouds and ground objects and obtain the cloud coverage area mask.

[0042] A cloud cover area mask is a binary image used to accurately mark cloud-covered areas in an image. Based on the cloud cover area mask, the feature threshold range in the qualified SAR image is extracted and compared with the feature thresholds in the SAR image to be processed, including backscattering coefficient values ​​and ground object pixel gray values, to eliminate misjudged areas.

[0043] It should be added that the feature threshold range refers to a screening standard with a clear numerical range defined for specific stable land cover types (such as woodland and mountains), used to distinguish stable land cover from clouds. Specifically, it includes: the backscattering coefficient threshold range used to characterize the scattering ability of SAR images and the texture feature threshold range used to reflect the spatial distribution law of the gray value of land cover pixels. Both are the optimal ranges that can meet the needs of remote sensing data preprocessing, obtained by the pre-set personnel based on multiple historical experiments. Specific stable land cover types refer to land cover categories that have stable spatial distribution, fixed scattering characteristics, and significant differences from the scattering characteristics of clouds, including: woodland, mountains, etc.

[0044] The backscattering coefficient threshold range represents the intensity of the echo scattered in the direction of radar transmission after electromagnetic waves interact with ground targets. It is a numerical range with clear upper and lower limits defined based on the backscattering characteristics of SAR images.

[0045] The texture feature threshold range represents a numerical interval with clear upper and lower limits, defined based on the spatial distribution patterns of the grayscale values ​​of ground object pixels (such as texture attributes like uniformity, roughness, and correlation) for a specific stable ground object type. The backscattering coefficient value is compared with the backscattering coefficient threshold range of stable ground objects in the benchmark threshold library. The grayscale value of the ground object pixel is compared with the texture feature threshold range of stable ground objects in the benchmark threshold library. The backscattering coefficient value refers to the ratio of the electromagnetic wave power scattered in the direction of the radar sensor to the electromagnetic wave power incident on the ground surface area, monitored by the SAR sensor. The grayscale value of the ground object pixel refers to the digitized brightness value corresponding to a single pixel of a specific ground object in the SAR image, monitored by optical remote sensing.

[0046] Based on the comparison results, the region attributes are determined. Only when both the backscattering coefficient value and the grayscale value of the ground object pixel fall within the feature threshold range is the region determined to be non-cloud. If any feature threshold is greater than or less than the maximum and minimum values ​​within the feature threshold range, the region is determined to meet the characteristics of a cloud and is a real cloud area. Based on the above determination results, the initial cloud coverage area mask is corrected by changing the label in the mask corresponding to all regions determined to be non-cloud from "cloud" to "non-cloud". All pixel labels determined to be "real cloud areas" are retained, and the optimized cloud coverage area mask is finally obtained.

[0047] Determine whether the cloud shadow coverage of the re-acquired SAR image after cloud and cloud shadow masking processing is still greater than the corresponding threshold, i.e., the cloud shadow coverage of the SAR image is greater than or equal to the SAR image cloud shadow coverage reference value; if it is still greater, send an abnormal cloud and cloud shadow masking processing prompt to the preset personnel; otherwise, mark the corresponding SAR image as a qualified SAR image and perform SAR control point feature matching.

[0048] Example A: In the preprocessing scenario of SAR images in mountainous forests, clouds are easily confused with the texture of dense forest, leading to a high false positive rate in traditional cloud detection algorithms and a SAR image cloud shadow coverage detection error below the acceptable standard. Using synchronously acquired GF-2 images (1m spatial resolution) that clearly distinguish cloud layers from forest boundaries as a true reference standard for SAR image cloud shadow coverage, a cloud detection model based on U-Net deep learning is adopted to optimize detection. This is combined with entinel-1 SAR images from a preset detection time period to assist in distinguishing clouds from ground features. The specific process is as follows:

[0049] Radiometric and geometric corrections were performed on the GF-2 imagery to ensure that its geographic coordinates were consistent with the SAR imagery to be processed; registration and radiometric normalization were performed on the entinel-1 SAR imagery to eliminate differences between SAR images.

[0050] Using manually labeled cloud regions in GF-2 images as label data, and combining the backscattering coefficient features and texture features of SAR images to construct a training sample set, the U-Net model is trained to improve the ability to identify thin and fragmented clouds. The SAR image to be processed, GF-2 image, and entinel-1 SAR image are input into the trained U-Net model, and the cloud coverage area mask is output.

[0051] The cloud cover area mask is a binary image generated by processing SAR images based on cloud detection algorithms (such as U-Net, Mask R-CNN and other deep learning models) to accurately label cloud-covered areas in the images. The feature threshold range of stable land features such as forests and mountains in multi-temporal cloudless SAR images is extracted and compared with the features of suspected cloud areas in the SAR images to be processed to eliminate misjudged areas caused by high scattering of forests.

[0052] In this embodiment, cloud and cloud shadow shielding processing helps improve the accuracy of cloud detection and the effectiveness of cloud-ground object differentiation in SAR images, reduces the interference of false and missed cloud detections on preprocessing, and filters out unqualified images that still have cloud shadow interference and excessive positioning deviations, ensuring the implementation accuracy of radiometric correction and terrain correction, and improving the reliability and usability of SAR data preprocessing results.

[0053] In this embodiment, performing SAR image cloud cover assessment helps to accurately screen out SAR images with excessive cloud interference and implement cloud shielding processing, reducing the distortion interference of cloud shadows on SAR data radiation characteristics and terrain features, and ensuring the integrity of terrain feature control point extraction and the accuracy of control point matching for qualified SAR images.

[0054] Furthermore, the specific process of SAR control point feature matching is as follows:

[0055] The number of control point features extracted from qualified SAR images is statistically analyzed based on the SAR imaging platform. The ratio of the number of control point features obtained to the preset number of control point features extracted is expressed as the SAR control point extraction result used to evaluate the completeness of control point feature extraction. The preset number of control point features extracted is the number of control point features required for this remote sensing data preprocessing.

[0056] The SAR control point extraction results are compared with the preset SAR control point extraction results; if the SAR control point extraction results are greater than or equal to the preset SAR control point extraction results, the features of the control points are matched based on the preset matching algorithm.

[0057] Conversely, based on pre-defined multi-scale feature extraction algorithms, such as multi-scale SIFT, to enhance control point features, such as terrain inflection points and forest boundaries, the number of effective control points extracted can be increased. Specifically, this refers to:

[0058] The terrain gradient value is obtained by using a terrestrial 3D laser scanner. The terrain gradient value represents the slope in the target area that reflects the degree of terrain undulation and the degree of abrupt change in the boundary of land features. The larger the terrain gradient value, the more obvious the features of terrain inflection points and forest boundaries, and the easier it is to be identified as an effective control point.

[0059] If the terrain gradient value is greater than or equal to the preset first threshold of terrain gradient, the corresponding target area is marked as a high gradient area, and the SAR control point extraction results are monitored in real time. Based on the original terrain gradient value, a prompt is sent to the preset personnel to adjust the terrain gradient value by a preset adjustment ratio until the SAR control point extraction result is equal to the preset SAR control point extraction result. The adjustment operation is stopped. The features of the control points are matched based on the preset matching algorithm to ensure the stability of the control point extraction and avoid missing edge control points with relatively prominent control point features but not reaching the benchmark threshold.

[0060] If the terrain gradient value is less than the preset first threshold but greater than the preset second threshold, the corresponding target area is marked as a medium gradient region. The features of the control points are directly matched based on the preset matching algorithm to ensure the stability of the control point extraction. If the terrain gradient value is less than or equal to the preset second threshold, the corresponding target area is marked as a low gradient region. Based on the original terrain gradient value, a prompt is sent to the preset personnel to increase the terrain gradient value by a preset adjustment ratio. The SAR control point extraction results are monitored in real time until the SAR control point extraction results are equal to the preset SAR control point extraction results. The increase operation stops, and the features of the control points are matched based on the preset matching algorithm to ensure the stability of the control point extraction. This process also removes noise points caused by flat terrain and uniform ground features, reducing interference from invalid control points. The preset first threshold and the preset second threshold are both set by the preset personnel based on historical experience.

[0061] The number of control point features that were successfully matched by high-resolution satellite data is counted and expressed as the number of matched control point pairs. The ratio of the number of matched control point pairs to the total number of matched control point pairs is expressed as the SAR control point matching rate, which reflects the accuracy of control point matching in qualified SAR images.

[0062] The SAR control point matching rate is compared with the preset SAR control point matching rate, which is represented by the average of the SAR control point matching rates obtained in the historical time period. If the SAR control point matching rate is greater than the preset SAR control point matching rate, it indicates that the SAR data has achieved pixel-level spatial alignment, and SAR data processing to evaluate the radiation distortion reduction effect is performed. Otherwise, terrain-constrained control point matching operation is performed.

[0063] In this embodiment, performing SAR control point feature matching helps ensure the sufficiency and completeness of qualified SAR image control point feature extraction, enhances the identifiability of key terrain areas (terrain inflection points, forest boundaries) features, improves the accuracy and quantity of effective control point extraction, thereby improving the pixel-level spatial alignment effect of SAR data and ensuring the accuracy of radiation distortion reduction effect assessment.

[0064] Furthermore, the specific process of terrain-constrained control point matching is as follows: Based on the preset matching algorithm (such as SIFT, SURF) and terrain constraints, the matched control point pairs are screened to eliminate mismatched points and reduce the mismatch rate of control point matching.

[0065] Iterative adjustments are made to the terrain constraints, specifically: Pre-set thresholds for slope difference and aspect difference between matched point pairs are compared with the slope difference and aspect difference of the matched point pairs corresponding to the already matched control point pairs; only control point pairs whose slope difference is less than or equal to the slope difference threshold and whose aspect difference is less than or equal to the aspect difference threshold are retained; control point pairs exceeding the threshold range are identified as mismatches and removed; the SAR control point extraction results and SAR control point matching rate are verified after removal to enhance the matching response of control point features and improve overall matching accuracy; the terrain constraints include: slope difference between matched point pairs and aspect difference between the slope difference and aspect difference of the matched point pairs. Aspect difference; the slope of the matched point pair is represented by the slope value of the corresponding surface location of each matched point during remote sensing image registration or terrain matching, obtained by lidar; the aspect difference is represented by the absolute value of the difference of the aspect values ​​of the corresponding surface location of each matched point, obtained by lidar. The terrain constraints include: the difference threshold of the matched point pair and the aspect difference threshold, both of which are preset by the preset personnel; if the preset requirements are met after the terrain constraint control point matching operation is completed, that is, the newly acquired SAR control point extraction result is greater than the preset SAR control point extraction result, then SAR data processing is performed; otherwise, a SAR control point matching anomaly prompt is sent to the preset personnel.

[0066] In this embodiment, performing terrain-constrained control point matching can accurately eliminate mismatched points in control point matching, effectively reduce the mismatch rate of control point matching, ensure the reliability of matching point pairs, enhance the control point feature matching response, and significantly improve the overall accuracy of control point matching.

[0067] Furthermore, the specific process of SAR data processing is as follows: The root mean square deviation (RMSD) of the backscattering coefficient of the SAR image, used to characterize the scattering ability of ground objects to SAR signals, is obtained through the SAR sensor and expressed as the SAR data RMSD. The backscattering coefficient of the SAR image refers to the coefficient extracted from the SAR image that characterizes the scattering ability of ground objects to microwave signals emitted by the SAR sensor. It is determined whether the obtained SAR data RMSD is less than the preset SAR data RMSD, where the preset SAR data RMSD is represented by the average value of the SAR data RMSD obtained over a historical time period. If it is less, SAR data reliability analysis is performed; otherwise, the SAR data is processed based on the polarization decomposition method to improve the accuracy of radiation distortion quantification.

[0068] It should be added that, such as Figure 4 The diagram shows a SAR data processing flowchart of the remote sensing data preprocessing method based on radiometric correction and terrain correction provided in this embodiment of the invention. If the SAR control point matching rate is not greater than the preset SAR control point matching rate, a terrain-constrained control point matching operation is performed. It is then determined whether the preset requirements are met after the terrain-constrained control point matching operation. If yes, SAR data processing continues; otherwise, a SAR control point matching anomaly alert is sent to a preset personnel. Regarding SAR data processing, it is determined whether the root mean square deviation of the acquired SAR data is less than the preset root mean square deviation of SAR data. If yes, SAR data reliability analysis is performed, first obtaining the SAR overlap and boundary offset error. If the SAR overlap is greater than or equal to the SAR overlap calibration value and the boundary offset error is less than or equal to the preset boundary offset error, a SAR data processing qualified prompt is sent to the preset personnel. Otherwise, based on the coordinate unification and parameter optimization method, specifically including: converting the vector boundary of the target forest area into the pixel coordinates of the SAR data based on the corrected preset spatial mapping relationship, and re-determining whether the SAR overlap is greater than or equal to the SAR overlap calibration value and the boundary offset error is less than or equal to the preset boundary offset error, a SAR data processing qualified prompt is sent to the preset personnel. Otherwise, a SAR data processing abnormal prompt is sent to the preset personnel.

[0069] In this embodiment, SAR data processing can accurately quantify the radiometric accuracy of the backscattering coefficients of SAR images after radiometric and topographic correction, thereby improving the radiometric consistency of the preprocessed SAR data and the reliability of the ground object scattering feature characterization.

[0070] Furthermore, the specific process of processing SAR data based on polarization decomposition is as follows: Based on a preset polarization decomposition method, such as VV polarization (Vertical transmit Vertical receive) and VH polarization (Vertical transmit Horizontal receive), the backscattered signals under different polarization modes are decomposed into polarization characteristic components of surface scattering, volume scattering, and secondary scattering.

[0071] Specifically, backscattering signal refers to the electromagnetic wave emitted by synthetic aperture radar, which, after illuminating ground features (such as forests, mountains, etc.), is scattered by the ground features and returns to the SAR sensor along a path opposite to the incident direction; surface scattering refers to the signal corresponding to the scattering that occurs when electromagnetic waves are incident on the surface of a continuous medium; volume scattering refers to the signal that returns to the sensor after electromagnetic waves penetrate into the interior of a discrete medium with a certain volume and undergo multiple scatterings with multiple particles in the medium; secondary scattering refers to the signal corresponding to the scattering of electromagnetic waves after they interact with the surfaces of two different scattering bodies.

[0072] The decomposed polarization feature components are used as auxiliary parameters to input into a preset correction model, such as a traditional C-correction / SCS correction model, which outputs standardized SAR data. Specifically, polarization feature components such as entropy, angle, anti-entropy, or volume scattering power are extracted based on polarization decomposition algorithms such as Cloude-Pottier or Freeman-Durden in the traditional C-correction / SCS correction model. Then, these components that can characterize the differences in the scattering mechanism of ground objects are used as auxiliary parameters to reconstruct the correction factor in the traditional C-correction / SCS correction model that only depends on the local incident angle, upgrading it to a composite function of the incident angle and polarization features. The parameters of the C-correction / SCS correction model are optimized by fitting SAR data. Finally, after accuracy verification, the correction process of coupled polarization features is performed on the entire area to output standardized SAR data that eliminates the interference of terrain and ground object scattering.

[0073] SAR data reliability analysis is performed based on the output standardized SAR data; standardized SAR data refers to SAR data with accurate spatial range, stable radiation characteristics, and qualified data quality.

[0074] In this embodiment, the polarization decomposition method helps to improve the accuracy of radiation distortion transformation and correction, outputs standardized SAR data with accurate spatial range and stable radiation characteristics, ensures the authenticity and effectiveness of the radiation physical quantities on which terrain correction is based, and enhances the radiation consistency of the preprocessed SAR data and the reliability of the ground object scattering characteristic characterization.

[0075] Furthermore, the specific process of SAR data reliability analysis is as follows:

[0076] The target area image is acquired by SAR payload, and after geometric correction, the target area is cropped out. The geographical range of the cropped area is derived, and the SAR overlap is obtained to evaluate the degree of matching between the cropped area and the real vector boundary. It refers to the ratio of the intersection area and the union area of ​​the SAR cropped area and the high-precision vector boundary.

[0077] The boundary offset error used to quantify the degree of deviation of the clipping boundary refers to the maximum Euclidean distance between the corresponding points of the SAR clipping region and the high-precision vector boundary. It is obtained by extracting the corresponding points of the two types of boundaries, calculating the Euclidean distance of each point, and taking the maximum value.

[0078] Determine whether the obtained SAR overlap is greater than or equal to the SAR overlap calibration value set in advance by the preset personnel, and whether the boundary offset error is less than or equal to the preset boundary offset error.

[0079] If yes, a SAR data processing pass notification is sent to the designated personnel. Otherwise, the cropping accuracy is improved based on coordinate unification and parameter optimization methods. Specifically, this involves converting the vector boundary (geographic coordinates) of the target forest area into pixel coordinates of the SAR data based on the corrected preset spatial mapping relationship. This specifically involves obtaining the corresponding geographic reference values ​​based on the geographic coordinates of all nodes on the vector boundary of the target area. These coordinates are discrete sets of points, each with a specific geographic reference value. The geographic coordinates of each node are substituted into the corrected spatial mapping relationship for mapping and obtaining the corresponding floating-point pixel coordinates. The floating-point pixel coordinates are rounded based on the geographic reference values. All the converted pixel coordinates are connected sequentially to obtain the pixel boundary of the target forest area in the SAR data.

[0080] Adjusting the boundary buffer of the cropping window specifically means: for every 1 pixel increase in the boundary offset error compared to the preset boundary offset error, increase the buffer by 0.5 pixels to ensure that the cropped area covers the entire target area; re-acquire the target area after coordinate unification and parameter optimization, and re-acquire the SAR overlap and boundary offset error. If the corresponding conditions are still not met, i.e., the SAR overlap is less than the SAR overlap calibration value, or the boundary offset error is greater than the preset boundary offset error, then send a SAR data processing anomaly prompt to the preset personnel; otherwise, send a SAR data processing qualified prompt to the preset personnel.

[0081] The specific process for obtaining the preset spatial mapping relationship is as follows: Collect raw auxiliary data of SAR images (such as orbital parameters and imaging geometric information), ground control points (GCPs, which can be extracted through GNSS measurements or high-precision DOM / DSM), and geographic coordinate system benchmarks of the target area (such as CGCS2000); based on the geometric principles of SAR imaging (such as the range-Doppler equation), use ground control points to correct the geometric distortion of the raw SAR images; fit the geographic coordinates of the control points with the pixel coordinates of the SAR images using algorithms such as least squares, weighted least squares, and robust estimation algorithms to generate a preliminary spatial mapping function; calculate the residual of the corrected control points, that is, the deviation between the actual pixel coordinates and the coordinates predicted by the mapping function, that is, the Euclidean distance between the actual pixel coordinates and the coordinates predicted by the mapping function in the pixel space. If the residual exceeds the threshold (generally less than or equal to 0.5 pixels), the control points are re-selected until the accuracy requirements are met, and finally the corrected preset spatial mapping relationship is obtained.

[0082] In this embodiment, performing SAR data reliability analysis helps to accurately quantify the degree of matching between the SAR clipping region and the real vector boundary, as well as the boundary offset deviation, ensuring complete coverage of the target area, avoiding the application deviation of radiometric correction and terrain correction results caused by clipping deviation, and ensuring that the final output SAR data has accurate spatial positioning and complete target area.

[0083] It should be added that, such as Figure 5 The diagram shows the structure of a remote sensing data preprocessing system based on radiometric correction and terrain correction provided in this embodiment of the invention. Specifically, it includes: a SAR image cloud cover assessment module, a SAR control point feature matching module, and a SAR data reliability analysis module. The SAR image cloud cover assessment module is used to assess the cloud cover status of SAR images during SAR data preprocessing for forest resource monitoring, and determines whether cloud and cloud shadow shielding processing is performed to reduce the impact of clouds and cloud shadows on SAR images based on the assessment's pass / fail status. The SAR control point feature matching module, after meeting the assessment's pass / fail status, matches SAR control point features to reflect the completeness and matching accuracy of control point extraction, and determines whether SAR data processing is performed to evaluate the radiometric distortion reduction effect based on the matching's pass / fail status. The SAR data reliability analysis module, after meeting the matching's pass / fail status, analyzes the reliability of SAR data to reflect the pass / fail status of SAR data processing, and determines whether SAR image coordinate and data optimization is performed to improve the cropping accuracy of the target area based on the analysis.

[0084] Example 2: Example 1 has already achieved basic spatial alignment and radiometric correction of SAR data. Building on this, Example 2 comprehensively considers the problem of inconsistent target area spatial definition caused by pixel-level spatial alignment errors, as well as derivative problems such as inaccurate selection of filter parameters, difficulty in separating noise and effective signals, and loss of details of real surface changes caused by composite radiometric distortion. This Example 2 achieves accurate support for subsequent core applications such as scale unification, filtering and denoising, and surface change detection using SAR data, ultimately obtaining high-standard SAR data with accurate spatial range, pure signal quality, and complete surface features. The specific process is as follows:

[0085] The specific process of SAR control point feature matching is as follows: For the acquired control point features, a preset interpolation algorithm is used to complete the spatial alignment of SAR data at the pixel level; the vector boundary of the target area is superimposed, and the root mean square error of registration and the geographic coordinate deviation of the control points are obtained to evaluate the uniformity of the spatial range definition of the target area during the scale unification process. Specifically, based on the selected n valid control points, the sum of the squares of the abscissa and ordinate deviations of all control points is accumulated, then divided by the number of control points n, and finally the square root of the result is taken as the root mean square error of the SAR image; the true geographic coordinates of the control points are obtained, and for each pair of corresponding control points, the abscissa and ordinate deviations are calculated separately. The offset and ordinate offset are used to obtain the offset of each control point in geospatial space, which is expressed as the control point geographic coordinate offset. The vector boundary of the target area refers to the use of vector boundary data (characterized by point, line, and surface geometric elements and corresponding attribute information) describing the geographic range of the target area under a unified geospatial reference system (such as the WGS84 latitude and longitude coordinate system) to delineate a precise geospatial constraint range, so as to achieve accurate spatial matching between SAR image pixels and the geographic range of the target area, covering: the latitude and longitude range of ecological protection forest belts, etc. The resolution of SAR images is obtained through high-resolution calibration targets, and the geographic coordinate offset is obtained through inertial measurement units.

[0086] If both indicators meet the standards, the spatial scope of the target area is defined. Otherwise, the alignment optimization process is initiated. Specifically, a prompt is sent to the designated personnel to increase the density of verification control points based on the original density, and the root mean square deviation of SAR data is monitored in real time until the root mean square deviation of SAR data in the target area equals the preset root mean square deviation of SAR data. The increase in the density of verification control points is then stopped to ensure uniform coverage of the target area. High-precision three-dimensional coordinates of all verification control points in the target area are collected using equipment such as GNSS feature threshold receivers or total stations. The coordinate data is then imported into feature threshold SAR image processing software to obtain the number of control points per unit area, and finally the original density of verification control points and the density of verification control points are obtained.

[0087] Based on the Gauss-Kruger projection correction model, coordinate correction is performed on control points in qualified SAR images using GNSS measured coordinates. Specifically, the geodetic coordinates of the GNSS measured control points are converted into Cartesian coordinates under this projection. Then, the original pixel coordinates of the control points in the qualified SAR images are converted into initial Gauss-Kruger plane coordinates through image geocoding parameters. Based on polynomial fitting algorithms and affine transformation algorithms, the deviation relationship between the initial plane coordinates of the SAR image control points and the GNSS measured plane coordinates is fitted, and the model parameters are calculated. Finally, the calculated parameters are used to perform coordinate correction on all control points in the SAR image, achieving accurate alignment between the SAR image and the Gauss-Kruger projection coordinate system.

[0088] The resolution and geographic coordinate deviation of the SAR image after the alignment optimization process are reacquired until both indicators meet the standards. Then, SAR data reliability analysis is performed to ensure that the spatial range of the target area is uniformly defined. Among them, the GNSS measured coordinates refer to the three-dimensional spatial coordinates of the ground target point in a specific coordinate system directly obtained by on-site measurement through GNSS (Global Navigation Satellite System).

[0089] Meeting both criteria means that the SAR image resolution is greater than or equal to the preset SAR image resolution, and the geographic coordinate deviation is less than or equal to the preset geographic coordinate deviation. The preset SAR image resolution and the preset geographic coordinate deviation are both represented by the average values ​​obtained from the corresponding historical time periods.

[0090] On the one hand, it can be explained that bilinear interpolation is used for the SAR image to be registered with geometric offset and the reference SAR image as a benchmark. Specifically, firstly, based on the mapping relationship between the pixel coordinates of the SAR image to be registered and the pixel coordinates (or geographic coordinates) of the reference SAR image, the coordinates of each target pixel on the SAR image to be registered are mapped to the sub-pixel position of the reference image. This position is a floating-point coordinate with a decimal part. Then, the four nearest neighboring integer pixels around this sub-pixel position are found. The corresponding weights are calculated based on the distance between the sub-pixel coordinates and the four neighboring pixels. The closer the pixel is, the greater the weight. Then, the pixel values ​​of the four neighboring pixels are multiplied by the corresponding weights respectively, and the products are added to obtain the pixel value of this sub-pixel position. This completes the pixel alignment, thereby completing SAR image monitoring and rapid preprocessing.

[0091] On the other hand, if the images after interferometric SAR processing and terrain correction are aligned, cubic convolution interpolation is used. Similarly, a mapping relationship is first established at the control points between the pixel coordinates of the SAR image to be registered and the pixel coordinates (or geographic coordinates) of the reference SAR image. The target pixel coordinates on the reference grid such as the orthophoto grid are mapped to the sub-pixel floating-point coordinates of the SAR image to be corrected. Then, 16 neighboring pixels in a 4×4 range around this sub-pixel position are determined. Based on the cubic convolution kernel function, the corresponding weights are calculated according to the positional relationship between the sub-pixel coordinates and the 16 neighboring pixels. The pixel values ​​of the 16 neighboring pixels are multiplied by their corresponding weights and summed to obtain the pixel value of the target pixel, thus completing pixel alignment, preserving SAR image details and boundary sharpness, and reducing SAR image blurring.

[0092] In summary, this application's embodiments, by assessing the cloud cover status of SAR images and determining whether cloud and cloud shadow shielding processing is necessary based on the assessment's pass rate, help avoid distortion interference from clouds and cloud shadows on the radiometric and topographic features of SAR images. This enhances the quality control capability of SAR data preprocessing data sources, enables precise screening and targeted processing of images with excessive cloud shadow interference, and thus achieves the goal of providing high-quality, cloud-free base images for subsequent radiometric and topographic corrections. After meeting the assessment's pass rate, SAR control point features are matched, and SAR data processing is determined based on the matching's pass rate, effectively improving the pixel-level spatial density of SAR data. Alignment accuracy enhances the reliability of spatial reference support for radiometric and topographic corrections, strengthens quality control of control point extraction and matching, and ensures the accuracy of radiometric distortion elimination effect evaluation. After meeting the matching requirements, the reliability of SAR data is analyzed, and the SAR image coordinates and data optimization is evaluated based on the analysis. This helps to accurately control the spatial positioning accuracy and target area coverage integrity of SAR data, reflects the quality closed-loop control level of the entire SAR data preprocessing process, and enhances the matching degree between preprocessed data and real geographic boundaries. Thus, the goal of outputting high-quality, high-reliability SAR data and ensuring the practical value of radiometric and topographic correction results is achieved.

[0093] The above-disclosed embodiments are merely some examples of the present invention and should not be construed as limiting the scope of the present invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A remote sensing data preprocessing method based on radiometric correction and topographic correction, characterized in that, The method includes: The extent to which SAR images are covered by cloud shadows is assessed, and the appropriate level of cloud and cloud shadow shielding is determined based on the assessment results. The specific process for cloud and cloud shadow shielding is as follows: The SAR images to be processed, acquired at preset monitoring time points, are input into a preset deep learning cloud detection model to obtain the cloud coverage area mask. The feature threshold range in qualified SAR images is extracted based on the cloud coverage area mask, and compared with the feature threshold in the SAR image to be processed to eliminate misjudged areas. The backscattering coefficient values ​​are compared with the backscattering coefficient threshold ranges of stable ground objects in the benchmark threshold library; The grayscale values ​​of ground feature pixels are compared with the texture feature threshold ranges of stable ground features in the benchmark threshold library; The region attributes are determined based on the comparison results. Only when the backscattering coefficient value and the gray value of the ground object pixel of the region to be verified both fall within the feature threshold range is the region determined to be non-cloudy. If any of the feature thresholds exceeds the feature threshold range, the area is determined to meet the characteristics of a cloud and is a real cloud area. Determine whether the cloud shadow coverage of the re-acquired SAR image after cloud and cloud shadow masking processing is still greater than the corresponding threshold; If it is still greater than the specified value, a cloud and cloud shadow shielding processing anomaly prompt will be sent to the preset personnel; otherwise, the corresponding SAR image will be marked as a qualified SAR image and SAR control point feature matching will be performed. After the SAR image assessment is completed, the features of the SAR control points are matched, and the suitability of the matching is used to determine whether to perform processing to improve the radiation distortion reduction effect. After SAR control point matching is completed, the reliability of SAR data is analyzed to reflect the quality of SAR data processing. Based on the analysis, it is determined whether to perform optimization to improve the cropping accuracy of the target area.

2. The remote sensing data preprocessing method based on radiometric correction and topographic correction as described in claim 1, characterized in that, The specific process for assessing the cloud cover status of SAR images is as follows: Obtain the ratio of the area of ​​the cloud-covered region to the area of ​​the entire SAR image, and express the corresponding ratio as the cloud cover of the SAR image; If the cloud cover of a SAR image is greater than or equal to the reference value for cloud cover of a SAR image, the corresponding SAR image is marked as a SAR image to be processed and cloud and cloud cover masking is performed. Otherwise, the corresponding SAR image is marked as a qualified SAR image and SAR control point feature matching is performed.

3. The remote sensing data preprocessing method based on radiometric correction and topographic correction as described in claim 2, characterized in that, The specific process of SAR control point feature matching is as follows: The number of control point features extracted from qualified SAR images is counted. The ratio of the number of control point features acquired to the preset number of control point features extracted is expressed as the SAR control point extraction result. Compare the SAR control point extraction results with the preset SAR control point extraction results; If the SAR control point extraction result is greater than or equal to the preset SAR control point extraction result, then the features of the control points are matched based on the preset matching algorithm; Conversely, enhancing control point features based on a pre-defined multi-scale feature extraction algorithm specifically refers to: Obtain terrain gradient values; If the terrain gradient value is greater than or equal to the preset first threshold of terrain gradient, the corresponding target area is marked as a high gradient area, and the SAR control point extraction result is monitored in real time. Based on the original terrain gradient value, a prompt is sent to the preset personnel to adjust the terrain gradient value by a preset adjustment ratio until the SAR control point extraction result is equal to the preset SAR control point extraction result, and the adjustment operation stops. The features of the control point are matched based on the preset matching algorithm. If the terrain gradient value is less than the preset first threshold of terrain gradient, but greater than the preset second threshold of terrain gradient, the corresponding target area is marked as the medium gradient area, and the features of the control point are directly matched based on the preset matching algorithm. If the terrain gradient value is less than or equal to the preset second threshold of terrain gradient, the corresponding target area is marked as a low gradient area. Based on the original terrain gradient value, a prompt is sent to the preset personnel to increase the terrain gradient value by a preset adjustment ratio. The SAR control point extraction results are monitored in real time until the SAR control point extraction results are equal to the preset SAR control point extraction results. The adjustment operation is stopped, and the features of the control points are matched based on the preset matching algorithm. The number of successfully matched control point features is counted and expressed as the number of matched control point pairs; The ratio of the number of matched control point pairs to the total number of matched control point pairs is expressed as the SAR control point matching rate. Compare the SAR control point matching rate with the preset SAR control point matching rate; If the SAR control point matching rate is greater than the preset SAR control point matching rate, SAR data processing to evaluate the radiation distortion elimination effect is performed; otherwise, terrain-constrained control point matching operation is performed.

4. The remote sensing data preprocessing method based on radiometric correction and topographic correction as described in claim 3, characterized in that, The specific process of the terrain constraint control point matching operation is as follows: Based on the preset matching algorithm and terrain constraints, the matched control point pairs are filtered to remove mismatched points; Verify the SAR control point extraction results and SAR control point matching rate after removal; If the terrain constraint control point matching operation meets the preset requirements, SAR data processing will be performed; otherwise, an abnormal SAR control point matching prompt will be sent to the preset personnel.

5. The remote sensing data preprocessing method based on radiometric correction and topographic correction as described in claim 4, characterized in that, The specific process of SAR data processing is as follows: The root mean square deviation between the backscattering coefficient and the preset backscattering coefficient is obtained and expressed as the root mean square deviation of SAR data. Determine whether the root mean square deviation of the acquired SAR data is less than the preset root mean square deviation of SAR data; If the value is less than 1, then SAR data reliability analysis is performed; otherwise, SAR data is processed based on polarization decomposition to improve the accuracy of radiation distortion.

6. The remote sensing data preprocessing method based on radiometric correction and topographic correction as described in claim 5, characterized in that, The specific process of processing SAR data based on the polarization decomposition method is as follows: Based on the preset polarization decomposition method, backscattered signals under different polarization modes are decomposed into polarization characteristic components of surface scattering, volume scattering, and secondary scattering. Surface scattering refers to the signal that is scattered when an electromagnetic wave is incident on the surface of a continuous medium. Volume scattering refers to the signal that returns to the sensor after electromagnetic waves penetrate into a discrete medium with a certain volume, undergo multiple scatterings with multiple particles in the medium, and then return to the sensor. The secondary scattering refers to the signal that returns to the sensor after an electromagnetic wave interacts with the surfaces of two different scattering bodies in succession. The decomposed polarization feature components are used as auxiliary parameters to input into the preset correction model, and standardized SAR data are output. Perform SAR data reliability analysis based on the output standardized SAR data; The standardized SAR data refers to SAR data that has accurate spatial range, stable radiation characteristics, and qualified data quality.

7. The remote sensing data preprocessing method based on radiometric correction and topographic correction as described in claim 2, characterized in that, The specific process of SAR control point feature matching is as follows: Based on the acquired control point features, a preset interpolation algorithm is used to complete the spatial alignment of SAR data at the pixel level; By overlaying the vector boundary of the target area, the root mean square error of registration and the geographical coordinate deviation of the control points are obtained. Obtain SAR image resolution and geographic coordinate deviation; If both indicators meet the criteria, the spatial scope of the target area is defined; otherwise, the alignment optimization process is initiated, which specifically refers to: Send a prompt to the preset personnel to increase the density of verification control points based on the original density of verification control points, and monitor the root mean square deviation of SAR data in real time until the root mean square deviation of SAR data in the target area equals the preset root mean square deviation of SAR data, and then stop increasing the density of verification control points. Based on the Gauss-Kruger projection correction model, coordinate correction is performed on control points in qualified SAR images by combining GNSS measured coordinates. Reacquire the SAR image resolution and geographic coordinate deviation after the alignment optimization process until both indicators meet the standards, and perform SAR data reliability analysis to ensure that the spatial range of the target area is uniformly defined; The two indicators being met means that the SAR image resolution is greater than or equal to the preset SAR image resolution, and the geographic coordinate deviation is less than or equal to the preset geographic coordinate deviation.

8. The remote sensing data preprocessing method based on radiometric correction and topographic correction as described in claim 7, characterized in that, The specific process of SAR data reliability analysis is as follows: Obtain SAR overlap; Obtain the boundary offset error; Determine whether the obtained SAR overlap is greater than or equal to the SAR overlap calibration value, and whether the boundary offset error is less than or equal to the preset boundary offset error; If yes, a SAR data processing pass notification will be sent to the designated personnel; otherwise, based on coordinate unification and parameter optimization methods, specifically: Based on the corrected preset spatial mapping relationship, the vector boundary of the target forest area is converted into pixel coordinates of SAR data, specifically: Obtain the corresponding geographic reference value based on the geographic coordinates of all nodes of the vector boundary of the target area; The target area obtained after coordinate unification and parameter optimization is retrieved again, and the SAR overlap and boundary offset error are retrieved again. If the corresponding conditions are still not met, a SAR data processing anomaly prompt is sent to the preset personnel; otherwise, a SAR data processing qualified prompt is sent to the preset personnel.

9. A remote sensing data preprocessing system based on radiometric correction and topographic correction, wherein the remote sensing data preprocessing system based on radiometric correction and topographic correction is used to implement the remote sensing data preprocessing method based on radiometric correction and topographic correction as described in any one of claims 1-8, characterized in that, The system includes: a SAR image cloud cover assessment module, a SAR control point feature matching module, and a SAR data reliability analysis module; The SAR image cloud cover assessment module is used to assess the cloud cover status of SAR images and determine whether cloud and cloud cover shielding processing is required based on the assessment results. The SAR control point feature matching module is used to match the features of SAR control points after the SAR image evaluation is completed, and to determine whether to perform processing to improve the radiation distortion reduction effect based on the matching qualification. The SAR data reliability analysis module is used to analyze the reliability of SAR data after SAR control point matching is completed, so as to reflect the qualification of SAR data processing. Based on the analysis, it is used to determine whether to perform optimization to improve the cropping accuracy of target area.

Images