SAR radiation cross-calibration method and device based on building area feature weight distribution, equipment and storage medium

Abstract

The application discloses a SAR radiation cross-calibration method and device based on building area feature weight distribution, equipment and storage medium, relates to the technical field of remote sensing, and comprises the following steps: performing mask extraction on a target reference SAR image and a target to-be-calibrated SAR image based on a land cover type product to obtain target reference building area pixels and target to-be-calibrated building area pixels; performing data clustering on the target reference building area pixels and the target to-be-calibrated building area pixels through a target clustering algorithm to obtain a reference region set and a to-be-calibrated region set; performing feature calculation based on weight distribution on the reference region set and the to-be-calibrated region set to obtain reference stable features and to-be-calibrated stable features; and performing radiation correction on the target to-be-calibrated SAR image based on the reference stable features and the to-be-calibrated stable features to obtain a target calibrated SAR image. The target clustering algorithm and the weighted feature calculation are used for radiation correction on the SAR image, and the accuracy of the radiation correction is improved.

Description

Technical Field

[0001] This application relates to the field of remote sensing technology, and in particular to a SAR radiometric cross-calibration method, apparatus, equipment and storage medium based on the weight allocation of building area features. Background Technology

[0002] Synthetic Aperture Radar (SAR), as an important tool for active microwave remote sensing, has been widely used in fields such as Earth observation and quantitative remote sensing analysis. Traditional SAR cross-calibration methods typically rely on natural targets such as open water, dense forests, and deserts as pseudo-invariant features for radiation reference transfer. However, these natural targets are easily affected by environmental factors such as wind speed, soil moisture, and vegetation phenology, resulting in poor scattering stability. Especially when the time baseline is long or there are strong seasonal differences, significant radiation uncertainty is introduced, thereby reducing calibration accuracy and the reliability of the results.

[0003] In contrast, urban building regions, due to their stable geometric structure and small variations in dielectric properties, exhibit strong temporal stability and can serve as more reliable calibration feature sources. However, existing cross-calibration methods based on building regions still have significant limitations. The most prominent of these is treating the entire building region as a homogeneous target for feature extraction, neglecting its internal heterogeneity in spatial distribution, density, and scattering characteristics. This results in insufficient representativeness of the extracted features, making it difficult to fully reflect the true radiative stability of the building region, and consequently affecting the accuracy and robustness of the final calibration results.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this application is to provide a SAR radiometric cross-calibration method, apparatus, equipment, and storage medium based on the weight allocation of building area features, aiming to solve the technical problem of insufficient accuracy in SAR radiometric cross-calibration.

[0006] To achieve the above objectives, this application proposes a SAR radiometric cross-calibration method based on building area feature weight allocation, the method comprising:

[0007] Based on the land cover type product, mask extraction is performed on the target reference SAR image and the target uncalibrated SAR image to obtain the target reference building area pixels and the target uncalibrated building area pixels.

[0008] The target reference building region pixels and the target uncalibrated building region pixels are clustered using a target clustering algorithm to obtain a reference region set and an uncalibrated region set.

[0009] The reference region set and the region set to be calibrated are subjected to feature calculation based on weight allocation to obtain reference stable features and region set to be calibrated stable features.

[0010] Radiometric correction is performed on the target SAR image to be calibrated based on the reference stability features and the stability features to be calibrated, to obtain the target SAR image.

[0011] In one embodiment, before the step of performing mask extraction on the target reference SAR image and the target uncalibrated SAR image based on the land cover type product to obtain the target reference building area pixels and the target uncalibrated building area pixels, the method further includes:

[0012] Acquire the initial reference SAR image and the initial SAR image to be calibrated;

[0013] The initial reference SAR image and the initial uncalibrated SAR image are registered to obtain an intermediate reference SAR image and an intermediate uncalibrated image.

[0014] The intermediate reference SAR image and the intermediate image to be calibrated are preprocessed to obtain the target reference SAR image and the target image to be calibrated.

[0015] In one embodiment, the step of performing mask extraction on the target reference SAR image and the target uncalibrated SAR image based on the land cover type product to obtain the target reference building area pixels and the target uncalibrated building area pixels includes:

[0016] Building area mask extraction was performed on the target reference SAR image and the target uncalibrated SAR image using land cover type products to obtain the initial reference building area pixels and the initial uncalibrated building area pixels.

[0017] The pixels of the initial reference building area and the pixels of the initial building area to be calibrated are filtered based on a threshold to obtain the pixels of the target reference building area and the pixels of the target building area to be calibrated.

[0018] In one embodiment, the step of clustering the target reference building region pixels and the target uncalibrated building region pixels using a target clustering algorithm to obtain a reference region set and an uncalibrated region set includes:

[0019] Based on the preset neighborhood radius and minimum number of points, density connectivity analysis is performed on the pixels of the target reference building area and the pixels of the target uncalibrated building area using a target clustering algorithm to determine the pixel category attributes of the pixels of the target reference building area and the pixels of the target uncalibrated building area.

[0020] Based on the pixel category attributes, clustering and filtering are performed on the pixels of the target reference building area and the pixels of the target building area to be calibrated, respectively, to obtain the reference area set and the area to be calibrated set.

[0021] In one embodiment, the step of performing feature calculations based on weight allocation on the reference region set and the region set to be calibrated to obtain reference stable features and region set to be calibrated stable features includes:

[0022] Multi-factor feature quantization is performed on the reference region set and the region set to be calibrated, respectively, to obtain the reference weight coefficient and the weight coefficient to be calibrated.

[0023] The reference stability characteristics of the reference region set are obtained by performing a weighted calculation based on the reference weight coefficients and the reference region set.

[0024] The uncalibrated stability characteristics of the uncalibrated region set are obtained by performing a weighted calculation based on the uncalibrated weight coefficients and the uncalibrated region set.

[0025] In one embodiment, the step of performing multi-factor feature quantization on the reference region set and the region set to be calibrated respectively to obtain reference weight coefficients and calibrated weight coefficients includes:

[0026] Calculate the reference density factor, reference size factor, and reference stability factor of the reference region set;

[0027] The reference weight coefficients of the reference region set are obtained by fusing the reference density factor, the reference size factor, and the reference stability factor.

[0028] Calculate the uncalibrated density factor, uncalibrated density factor, and uncalibrated stability factor of the set of regions to be calibrated;

[0029] The uncalibrated weight coefficients of the uncalibrated region set are obtained by fusion calculation based on the uncalibrated density factor, the uncalibrated density factor and the uncalibrated stability factor.

[0030] In one embodiment, the step of performing radiometric correction on the target SAR image to be calibrated based on the reference stabilization features and the calibrated SAR features to obtain a target calibrated SAR image includes:

[0031] The calibration constant is calculated based on the reference stability feature and the stability feature to be calibrated.

[0032] The mean square error is calculated based on the calculated value of the calibration constant and the preset theoretical value of the calibration constant to obtain the calibration accuracy;

[0033] Based on the calibration accuracy, radiometric correction is performed on the SAR image of the target to be calibrated to obtain the SAR image of the target.

[0034] Furthermore, to achieve the above objectives, this application also proposes a SAR radiometric cross-calibration device based on building area feature weight allocation, wherein the SAR radiometric cross-calibration device based on building area feature weight allocation includes:

[0035] The mask extraction module is used to extract the mask from the target reference SAR image and the target uncalibrated SAR image based on the land cover type product, so as to obtain the target reference building area pixels and the target uncalibrated building area pixels.

[0036] The data clustering module is used to perform data clustering on the pixels of the target reference building area and the pixels of the target uncalibrated building area using a target clustering algorithm to obtain a reference area set and an uncalibrated area set.

[0037] The feature calculation module is used to perform feature calculation based on weight allocation on the reference region set and the region set to be calibrated, so as to obtain reference stable features and calibrated stable features.

[0038] The radiometric correction module is used to perform radiometric correction on the target SAR image to be calibrated based on the reference stability features and the stability features to be calibrated, so as to obtain the target calibrated SAR image.

[0039] Furthermore, to achieve the above objectives, this application also proposes a SAR radiometric cross-calibration device based on building area feature weight allocation. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the SAR radiometric cross-calibration method based on building area feature weight allocation as described above.

[0040] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the SAR radiation cross-calibration method based on building area feature weight allocation as described above.

[0041] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the SAR radiation cross-calibration method based on building area feature weight allocation as described above.

[0042] One or more technical solutions proposed in this application have at least the following technical effects:

[0043] This application proposes a SAR radiometric cross-calibration method, apparatus, device, and storage medium based on building area feature weight allocation. The method involves mask extraction of a target reference SAR image and a target uncalibrated SAR image based on land cover type products, yielding target reference building area pixels and target uncalibrated building area pixels. A target clustering algorithm is then used to cluster these pixels, resulting in a reference area set and a uncalibrated area set. Weighted feature calculations are performed on the reference area set and the uncalibrated area set to obtain reference stable features and uncalibrated stable features. Radiometric correction is then applied to the uncalibrated SAR image based on these features to obtain a calibrated SAR image. By using a target clustering algorithm and weighted feature calculation for radiometric correction of the SAR image, the accuracy of radiometric correction is improved. Attached Figure Description

[0044] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating an embodiment of the SAR radiometric cross-calibration method based on building area feature weight allocation in this application;

[0047] Figure 2 This is an example image showing the extraction of building masks and initial pixel screening of SAR images provided in an embodiment of this application.

[0048] Figure 3 An example diagram is provided for the clustering results obtained after clustering using the DBSCAN clustering algorithm in this application embodiment;

[0049] Figure 4 This is a schematic diagram of the module structure of the SAR radiation cross-calibration device based on building area feature weight allocation in an embodiment of this application;

[0050] Figure 5 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the SAR radiation cross-calibration method based on building area feature weight allocation in the embodiments of this application.

[0051] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0052] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0053] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0054] The main solution of this application embodiment is as follows: based on land cover type products, mask extraction is performed on the target reference SAR image and the target uncalibrated SAR image to obtain target reference building area pixels and target uncalibrated building area pixels; the target reference building area pixels and the target uncalibrated building area pixels are clustered using a target clustering algorithm to obtain a reference area set and a uncalibrated area set; feature calculation based on weight allocation is performed on the reference area set and the uncalibrated area set to obtain reference stable features and uncalibrated stable features; radiometric correction is performed on the target uncalibrated SAR image based on the reference stable features and the uncalibrated stable features to obtain a target calibrated SAR image.

[0055] In this embodiment, for ease of description, the SAR radiation cross-calibration device based on building area feature weight allocation will be used as the execution subject for the following description.

[0056] Synthetic Aperture Radar (SAR), as an important tool for active microwave remote sensing, has been widely used in fields such as Earth observation and quantitative remote sensing analysis. Traditional SAR cross-calibration methods typically rely on natural targets such as open water, dense forests, and deserts as pseudo-invariant features for radiation reference transfer. However, these natural targets are easily affected by environmental factors such as wind speed, soil moisture, and vegetation phenology, resulting in poor scattering stability. Especially when the time baseline is long or there are strong seasonal differences, significant radiation uncertainty is introduced, thereby reducing calibration accuracy and the reliability of the results.

[0057] In contrast, urban building regions, due to their stable geometric structure and small variations in dielectric properties, exhibit strong temporal stability and can serve as more reliable calibration feature sources. However, existing cross-calibration methods based on building regions still have significant limitations. The most prominent of these is treating the entire building region as a homogeneous target for feature extraction, neglecting its internal heterogeneity in spatial distribution, density, and scattering characteristics. This results in insufficient representativeness of the extracted features, making it difficult to fully reflect the true radiative stability of the building region, and consequently affecting the accuracy and robustness of the final calibration results.

[0058] This application provides a solution for radiometric correction of SAR images through target clustering algorithms and weighted feature calculations, thereby improving the accuracy of radiometric correction.

[0059] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a SAR radiation cross-calibration device based on building area feature weight allocation. The following description uses a SAR radiation cross-calibration device based on building area feature weight allocation as an example to illustrate this embodiment and the subsequent embodiments.

[0060] Based on this, the embodiments of this application provide a SAR radiation cross-calibration method based on building area feature weight allocation, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the SAR radiation cross-calibration method based on building area feature weight allocation in this application.

[0061] In this embodiment, the SAR radiometric cross-calibration method based on building area feature weight allocation includes steps S11~S14:

[0062] Step S11: Based on the land cover type product, perform mask extraction on the target reference SAR image and the target uncalibrated SAR image to obtain the target reference building area pixels and the target uncalibrated building area pixels.

[0063] It should be noted that land cover type products are data products that include classification information on different types of land cover (such as buildings, vegetation, water bodies, etc.). Their purpose is to provide a spatial distribution reference for built areas and help to accurately locate built areas in SAR images.

[0064] Additionally, it should be noted that the target reference SAR image is a SAR image that has undergone precise absolute calibration, serving as a radiometric calibration reference to provide reliable information on radiometric characteristics. The target SAR image to be calibrated is a SAR image that has not undergone radiometric calibration and is the object of this calibration process. Mask extraction is a technique that uses the classification identifiers of building areas in land cover type products to filter regions in SAR images. By constructing a building area mask, only pixels in the SAR image corresponding to the mask are retained, while pixels in non-building areas are removed.

[0065] Understandably, the purpose of this step is to separate building region pixels from two types of SAR images. Building regions are characterized by stable geometry, small changes in dielectric constant, and stable backscattering characteristics, making them suitable as pseudo-invariant features for cross-calibration. Utilizing the classification information of land cover type products, masking technology is used to achieve accurate segmentation between building and non-building regions, avoiding interference from unstable scatterers in non-building regions on subsequent calibration processes. This lays the foundation for extracting stable radiation features. This step focuses only on the initial separation of building regions and does not involve subsequent clustering and weight calculation operations.

[0066] Specifically, select a land cover type product with a resolution consistent with the SAR image to ensure spatial matching of the mask extraction; use either vector masks or raster masks for extraction. Vector masks are suitable for situations where building boundaries are clear, while raster masks are suitable for large-scale, continuously distributed building areas; perform a preliminary spatial consistency check on the extracted building area pixels, removing isolated pixels that significantly exceed the reasonable building area range. Using a high-resolution land cover type product alone can improve the positioning accuracy of building areas; combining vector masks and spatial consistency checks can reduce the residual pixels in non-building areas; using multiple methods in combination can effectively improve the accuracy and completeness of building area pixel extraction.

[0067] For example, a target reference SAR image and a target SAR image to be calibrated are acquired using Sentinel-1 C-band at a resolution of 10m. An ESA land cover type product (with the same resolution as the SAR image) based on Sentinel-1 and Sentinel-2 data is selected. Through grid masking technology, the grids in the ESA product that are marked as building areas are matched with the SAR image, and the building area pixels in the two types of SAR images are extracted to complete the masking extraction operation.

[0068] For better understanding, please refer to Figure 2 , Figure 2 This is an example image showing the results of extracting building areas from a SAR image after extracting building masks and performing initial pixel screening.

[0069] Step S12: Cluster the target reference building area pixels and the target uncalibrated building area pixels using a target clustering algorithm to obtain a reference area set and an uncalibrated area set.

[0070] It should be noted that target clustering algorithm is an algorithm used to group and classify data points. In this application, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm is used to divide the building area into different building groups based on the spatial distribution density and correlation of pixels in the building area.

[0071] Additionally, it should be noted that the target reference building region pixels are the set of building region pixels extracted from the target reference SAR image. The target uncalibrated building region pixels are the set of building region pixels extracted from the target uncalibrated SAR image. The reference region set is a set of multiple building groups obtained by clustering the target reference building region pixels, with each group having similar spatial distribution characteristics. The uncalibrated region set is a set of multiple building groups obtained by clustering the target uncalibrated building region pixels, and it corresponds to the reference region set.

[0072] Understandably, the purpose of this step is to address the structural heterogeneity within building areas by dividing scattered building area pixels into building groups with specific spatial characteristics. The principle is to use clustering algorithms to identify the density connectivity structure of building area pixels, grouping spatially connected pixels with similar densities into one category. This distinguishes different types of building groups, such as high-density urban areas and sparsely populated industrial areas, avoiding the problem of insufficient feature representativeness caused by treating the entire building area as a homogeneous target. This provides a basis for subsequent targeted weight allocation. This step only completes the division of building groups and does not involve weight calculation or radiation correction.

[0073] Specifically, the DBSCAN density clustering algorithm is used to identify density-connected regions by setting reasonable neighborhood radius Eps and minimum number of points MinPts. After clustering, each group is screened by size, removing small groups with too few pixels to reduce noise interference.

[0074] For example, the DBSCAN clustering algorithm is used on the extracted target reference building area pixels and target uncalibrated building area pixels. The neighborhood radius Eps=30 and the minimum number of points MinPts=50 are set. Density connectivity analysis is performed on the two types of pixels respectively. Pixels that meet the density connectivity conditions are grouped into one class, and noise points and small groups with fewer than 50 pixels are removed. Finally, a reference area set and an uncalibrated area set containing multiple effective building groups are obtained.

[0075] For better understanding, please refer to Figure 3 , Figure 3 To use the DBSCAN clustering algorithm to Figure 2 An example image showing the clustering results obtained after clustering the mask extraction results. Figure 3 The minimum number of points set in the algorithm is 800. After clustering, 4 effective clusters and 2010 noise points are obtained, and 42 small clusters are hidden.

[0076] Step S13: Perform feature calculation based on weight allocation on the reference region set and the region set to be calibrated to obtain reference stable features and region set to be calibrated stable features.

[0077] It should be noted that the reference stable feature is a stable and representative radiometric feature value obtained by weighted calculation of the reference region set, reflecting the reliable radiometric characteristics of the reference image. The calibrated stable feature is a radiometric feature value obtained by weighted calculation of the region set to be calibrated, used to compare with the reference stable feature to achieve calibration.

[0078] Understandably, the purpose of this step is to extract radiation features with high reliability and low uncertainty, providing an accurate basis for radiation correction. The principle is to consider the differences in the contribution of different building groups to the calibration model, assigning higher weights to high-density, large-scale, and highly stable building groups through weight allocation, reducing the influence of low-confidence groups. Then, based on the weighted group characteristics, the overall stability features are calculated, avoiding the calibration accuracy degradation caused by treating all building areas equally in traditional methods. This step only extracts the stability features and does not involve the final radiation correction operation.

[0079] Specifically, weight coefficients are constructed based on the number of pixels (scale factor), standard deviation of pixel values ​​(stability factor), and density of the building group (density factor). Multiple factors are fused to obtain the weight coefficients, which are then normalized to ensure the sum of the weights equals 1, guaranteeing the rationality of feature calculation. Finally, the pixel mean of each group is multiplied by its corresponding normalized weight to obtain stable features.

[0080] For example, the number of pixels (scale factor), standard deviation of pixel value (stability factor), and population density (density factor) based on the K-nearest neighbor algorithm are calculated for each building group in the reference region set and the region set to be calibrated. The three factors are fused to obtain the initial weights. After normalizing the initial weights, the pixel mean of each group is multiplied by the corresponding normalized weight, and the sum is obtained to obtain the reference stable features and the stable features to be calibrated.

[0081] Step S14: Perform radiometric correction on the target SAR image to be calibrated based on the reference stability features and the stability features to be calibrated, to obtain the target SAR image.

[0082] It should be noted that radiometric correction is the process of adjusting the radiometric values ​​of the SAR image to be calibrated so that they are consistent with the radiometric scale of the reference image. The target-calibrated SAR image is a SAR image whose radiometric values ​​are reliable and comparable after radiometric correction.

[0083] Understandably, the purpose of this step is to perform radiometric calibration on the SAR image to be calibrated, ensuring the accuracy and consistency of its radiometric values. The principle is to determine a calibration constant based on the difference between the reference stable characteristics and the stable characteristics of the image to be calibrated. This calibration constant is then used to adjust the radiometric values ​​of the image to be calibrated, eliminating radiometric bias between the two types of images. This ensures that the radiometric characteristics of the image to be calibrated are consistent with the reference image, meeting the needs of quantitative remote sensing applications and data fusion. This step is the final execution stage of the entire calibration process, completing radiometric correction based on the results of the preceding steps.

[0084] Specifically, the calibration constant is calculated using the difference method, i.e., calibration constant = reference stable feature - stable feature to be calibrated. The calculated value of the calibration constant is obtained and verified in conjunction with the calibration accuracy. If the accuracy does not meet the preset requirements, the parameters of the preceding steps (such as clustering parameters and weighting factors) are readjusted and the calibration constant is recalculated and corrected.

[0085] For example, the difference between the obtained reference stable feature and the stable feature to be calibrated is calculated to obtain the calibration constant value (e.g., 27.955dB). The mean square error of the calculated value is calculated with the preset theoretical value of the calibration constant to verify the calibration accuracy (0.197dB). After confirming that the accuracy meets the requirements, the radiometric value of the target SAR image to be calibrated is adjusted using the calibration constant to complete the radiometric correction and obtain the target calibrated SAR image.

[0086] This embodiment, through the above-described scheme, first extracts pixels from building areas, then clusters them into building groups, calculates stable features based on multi-factor weight allocation, and finally uses the differences in stable features for radiometric correction. This effectively solves the problems of natural targets being easily affected by environmental interference, outlier interference in building area calibration, and unreasonable weight allocation in traditional cross-calibration. By leveraging the stable characteristics of building areas and a scientific processing flow, the accuracy and stability of SAR radiometric cross-calibration are improved, enabling calibrated SAR images to have good comparability and consistency at the radiometric scale, providing reliable data support for subsequent quantitative remote sensing applications.

[0087] Based on the above implementation scheme, in one feasible implementation, the step of performing mask extraction on the target reference SAR image and the target uncalibrated SAR image based on the land cover type product to obtain the target reference building area pixels and the target uncalibrated building area pixels further includes S21~S23:

[0088] Step S21: Obtain the initial reference SAR image and the initial SAR image to be calibrated.

[0089] It should be noted that the initial reference SAR image is an absolutely calibrated SAR image that has not undergone registration or preprocessing. It serves as the raw reference data for subsequent processing, and its core function is to provide initial reliable radiometric information. The initial SAR image to be calibrated is a SAR image that has not undergone registration, preprocessing, or radiometric calibration. It is the raw processing object in this calibration process, and its function is to achieve radiometric calibration through a series of subsequent processing steps.

[0090] Understandably, the purpose of this step is to acquire the basic raw data required for calibration, providing data support for the entire cross-calibration process. The principle is to select two types of SAR images with consistent specific imaging conditions to ensure the feasibility of subsequent processing and the reliability of the calibration results. Because differences in imaging conditions can introduce additional radiometric biases that affect calibration accuracy, it is necessary to pay attention to the consistency of key parameters such as imaging mode, band, and polarization channel when acquiring images. This step only completes the acquisition of the raw images and does not involve subsequent registration, preprocessing, or other operations.

[0091] Specifically, SAR images can be acquired from open-source satellite data platforms (such as the Copernicus Data Space Ecosystem). These platforms provide SAR data from various satellites (such as Sentinel-1), offering broad coverage and high accessibility. Alternatively, customized SAR images can be obtained from commercial satellite data providers, suitable for scenarios with specific requirements for imaging time and area. When acquiring images, images with consistent imaging modes (such as IW interferometric wide-swath imaging), the same bands (such as C-band), and consistent polarization channels should be selected to reduce the impact of differences in imaging conditions.

[0092] For example, SAR image data from the Sentinel-1 satellite was obtained from the Copernicus Data Space Ecosystem. The initial reference SAR image and the initial uncalibrated SAR image were both C-band, GRD Level 1 products, IW interferometric wide-swath imaging mode, 10m resolution, and the imaging area was the same, ensuring that the basic imaging conditions of the two types of images were consistent, laying the foundation for subsequent processing.

[0093] Step S22: Register the initial reference SAR image and the initial SAR image to be calibrated to obtain an intermediate reference SAR image and an intermediate SAR image to be calibrated.

[0094] It should be noted that registration is a processing technique that makes different SAR images consistent in spatial location. By finding corresponding points between images, the spatial coordinates of the images are adjusted to achieve spatial correspondence of ground features. The intermediate reference SAR image is the image obtained after registration of the initial reference SAR image and the spatially corrected image. The intermediate image to be calibrated is the image obtained after registration of the initial image to be calibrated and the spatially corrected image.

[0095] Understandably, the purpose of this step is to eliminate spatial location discrepancies between the two types of SAR images, ensuring accurate spatial correspondence of the same ground feature in both types of images. Through image registration techniques, the initial reference SAR image and the initial SAR image to be calibrated are unified to the same projection coordinate system and resolution grid, enabling subsequently extracted building area pixels to accurately correspond to the same actual building area. This avoids building area mismatches caused by spatial location discrepancies, which could affect the reliability of calibration features. This step only corrects spatial location and does not involve radiometric preprocessing.

[0096] Specifically, a feature-point-based registration method is adopted, which establishes the correspondence between feature points of two types of images by extracting feature points such as corners and edges, and is suitable for areas with rich ground features. A region-based registration method is adopted, which achieves registration by calculating the gray-level correlation of local areas of the image, and is suitable for areas with uniform ground texture. After registration, spatial consistency verification is performed to calculate the registration error. If the error exceeds a preset threshold (e.g., 1 pixel), registration is repeated. The feature-point-based registration method has high registration accuracy; the region-based registration method has strong stability; and the combination of spatial consistency verification can ensure the registration effect.

[0097] For example, for the acquired initial reference SAR image and initial SAR image to be calibrated, a registration method based on SIFT (Scale-Invariant Feature Transform) feature points is adopted. First, SIFT feature points of the two types of images are extracted. Feature points are matched by Euclidean distance, and reliable corresponding points are selected. The spatial transformation matrix is ​​solved based on the corresponding points. The spatial coordinates of the initial SAR image to be calibrated are adjusted using the transformation matrix, so that the two types of images are unified under the same UTM (Universal Transverse Mercator Grid System) projection coordinate system and 10m resolution grid. After registration, the registration error is verified to be 0.5 pixels, which meets the spatial consistency requirement, and intermediate reference SAR image and intermediate SAR image to be calibrated are obtained.

[0098] Step S23: Preprocess the intermediate reference SAR image and the intermediate image to be calibrated to obtain the target reference SAR image and the target image to be calibrated.

[0099] It should be noted that preprocessing is a series of radiometric and geometric corrections performed on the registered SAR image, including orbit correction, thermal noise removal, filtering, terrain correction, and other operations.

[0100] Understandably, the purpose of this step is to eliminate the radiometric and geometric distortions inherent in the SAR image, thereby improving image quality and the stability of radiometric quantification. The principle is to address issues such as orbital deviations, thermal noise, speckle noise, and terrain distortion generated during SAR image imaging by eliminating or reducing them one by one through appropriate correction processing. This makes the image's radiometric values ​​closer to the true backscattering characteristics of ground objects, resulting in more accurate geometric locations and providing high-quality image data for subsequent building area extraction and feature calculation. In this step, the reference image also needs to be radiometrically calibrated, converting the DN (Digital Number) value into backscattering coefficients to provide a traceable radiometric benchmark for calibration.

[0101] Specifically, orbit correction: Precise ephemeris replacement of predicted orbits is used to calibrate sensor positions and improve geolocation accuracy. This operation can be performed based on the initial positioning accuracy of the image. Thermal noise removal: A noise estimation method based on statistical characteristics is used to remove additive thermal noise in the imaging chain, avoiding its systematic increase in the backscattering coefficient. Filtering: Lee filtering, Gamma filtering, and other methods are used to suppress multiplicative speckle noise and improve the image signal-to-noise ratio. Terrain correction: Based on a cosine correction model or a radiative transfer model, it aims to correct geometric distortions such as forward compression, overlay, and shadows caused by terrain undulations and reproject them onto map coordinates, thereby obtaining high-precision geolocation and consistency of radiance values.

[0102] For example, the Sentinel-1 satellite data processing software SNAP is used to preprocess the intermediate reference SAR image and the intermediate image to be calibrated: orbit correction, thermal noise removal, Lee filtering, terrain correction and radiometric calibration are performed on the intermediate reference SAR image to convert its DN value into backscattering coefficients; orbit correction, thermal noise removal, Gamma filtering and terrain correction are performed on the intermediate image to be calibrated to obtain the target reference SAR image and the target image to be calibrated to have optimized radiometric and geometric characteristics.

[0103] This embodiment, through the above-described scheme, eliminates spatial location deviations and optimizes the radiometric and geometric characteristics of the images by acquiring original images with consistent imaging conditions. This solves the problems of spatial location mismatch, self-radiometry, and geometric distortion affecting calibration accuracy in the original SAR images. The registration step ensures the spatial correspondence accuracy of the same ground feature in both types of images. The preprocessing step improves image quality and radiometric quantitative stability, providing high-quality basic data for subsequent building area extraction, clustering, feature calculation, and radiometric correction. This further enhances the accuracy and reliability of the entire SAR radiometric cross-calibration, making the calibration results more compliant with the stringent requirements of quantitative remote sensing applications.

[0104] Based on the above implementation scheme, in one feasible implementation, the step of performing mask extraction on the target reference SAR image and the target uncalibrated SAR image based on the land cover type product to obtain the target reference building area pixels and the target uncalibrated building area pixels includes S31~S32:

[0105] Step S31: Using land cover type products, extract building area masks from the target reference SAR image and the target uncalibrated SAR image to obtain the initial reference building area pixels and the initial uncalibrated building area pixels.

[0106] It should be noted that mask extraction utilizes building area identifiers from land cover type products to select regions in the SAR image. By constructing a binary mask, SAR pixels corresponding to building areas are retained, while pixels in non-building areas are discarded. The initial reference building area pixels are the set of building area pixels obtained after mask extraction from the target reference SAR image, containing a small number of unstable scatterer pixels. The initial uncalibrated building area pixels are the set of building area pixels obtained after mask extraction from the target uncalibrated SAR image, also containing a small number of interfering pixels.

[0107] Understandably, the purpose of this step is to initially separate building area pixels from the two types of SAR images, providing basic data for subsequent screening and clustering. The principle is to use the classification results of land cover type products to achieve preliminary segmentation of building and non-building areas through masking technology. It utilizes the stable backscattering characteristics of building areas to replace natural targets susceptible to environmental interference, solving the uncertainty problem of natural targets in traditional cross-calibration. This step only completes the preliminary extraction of building areas and does not handle outlier interference within the area.

[0108] Specifically, a land cover type product with the same resolution as the SAR image (such as ESA land cover product) is selected to ensure spatial matching of the mask extraction and avoid building area positioning deviations caused by resolution differences. A grid mask extraction method is used, setting the grid value of the building area in the land cover type product to 1 and the non-building area to 0, and multiplying it pixel by pixel with the SAR image to obtain the initial building area pixels. After extraction, the initial building area pixels are checked for spatial continuity, retaining continuously distributed pixel areas and removing isolated single pixels.

[0109] For example, an ESA (European Space Agency) land cover type product with the same resolution as the target reference SAR image and the target uncalibrated SAR image (10m resolution) is selected. This product is based on Sentinel-1 and Sentinel-2 data and contains accurate building area classification information. A raster mask is constructed, setting the raster values ​​of the buildings in the product to 1 and the remaining areas to 0. Pixel-by-pixel multiplication is performed with the two types of SAR images, retaining the pixels whose product result is 1, to obtain the initial reference building area pixels and the initial uncalibrated building area pixels. At the same time, isolated single pixels in the extraction result are removed to ensure the spatial continuity of the initial building area.

[0110] Step S32: Filter the initial reference building area pixels and the initial uncalibrated building area pixels based on a threshold to obtain the target reference building area pixels and the target uncalibrated building area pixels.

[0111] It should be noted that the threshold is a numerical limit set based on the initial distribution of pixel values ​​in the building area, used to distinguish between normal building area pixels and extreme value pixels.

[0112] Understandably, the purpose of this step is to eliminate outlier interference in the initial building area pixels and improve the purity of the building area pixels. The principle is that the initial building area may contain energy peak pixels generated by metal corner reflectors, rooftop equipment, etc. These extreme values ​​can contaminate the accuracy of feature statistics. By setting a threshold, outlier pixels exceeding a reasonable range are pruned, solving the outlier interference problem in building area calibration and providing cleaner building area pixel data for subsequent clustering processing. This step only filters out extreme value pixels and does not change the overall structure of the building area.

[0113] Specifically, in one embodiment of this application, a quantile threshold method is used to set the threshold, such as selecting the 98th quantile as the upper limit threshold. Extreme values ​​of pixel values ​​greater than this threshold are eliminated. This method does not require assumptions about the pixel value distribution type and is suitable for SAR image pixel data with complex distributions. In another embodiment of this application, a statistical distribution threshold method is used. Assuming that the pixel values ​​follow a normal distribution, the mean ± 3 times the standard deviation is calculated as the threshold range, and pixels exceeding this range are eliminated. This method is suitable for situations where the pixel value distribution is relatively concentrated. The threshold is adjusted in conjunction with visual interpretation. The results after threshold screening are visually inspected. If there are still obvious extreme values, the threshold is adjusted appropriately for re-screening.

[0114] For example, a threshold is set using the quantile thresholding method for the obtained initial reference building area pixels and initial target building area pixels. The two types of pixel values ​​are sorted in ascending order, and the 98th percentile value is calculated as the upper limit threshold. This removes extremely high-value pixels after the 98th percentile after sorting, avoiding interference from energy peaks introduced by metal corner reflectors, rooftop equipment, etc., on subsequent feature calculations, and finally obtaining clean target reference building area pixels and target target building area pixels.

[0115] This embodiment, through the above-described scheme, effectively solves the outlier interference problem in built-up area calibration by employing a two-stage process: first, preliminary mask extraction using land cover type products, and then extreme value filtering based on thresholds. Preliminary mask extraction separates built-up areas from non-built-up areas, utilizing the stable characteristics of built-up areas to replace natural targets, thus improving the stability of calibration features. Threshold filtering eliminates extreme value pixels, avoiding the impact of outliers on the accuracy of feature statistics, and providing high-purity built-up area pixel data for subsequent clustering and weight allocation, further improving the accuracy and reliability of SAR radiometric cross-calibration.

[0116] Based on the above implementation scheme, in a feasible implementation, the step of performing data clustering on the target reference building area pixels and the target uncalibrated building area pixels using a target clustering algorithm to obtain a reference area set and an uncalibrated area set includes S41~S42:

[0117] Step S41: Based on the preset neighborhood radius and minimum number of points, density connectivity analysis is performed on the pixels of the target reference building area and the pixels of the target uncalibrated building area using a target clustering algorithm to determine the pixel category attributes of the pixels of the target reference building area and the pixels of the target uncalibrated building area.

[0118] It should be noted that the neighborhood radius (Eps) is a parameter in the target clustering algorithm that defines the range of a pixel's neighborhood, used to determine the size of a pixel's neighborhood space. The minimum number of points (MinPts) is a parameter in the target clustering algorithm that defines a core point; that is, when the number of pixels contained within a pixel's neighborhood reaches this value, the pixel is determined to be a core point. Density connectivity analysis is an analytical method that determines whether pixels belong to the same density region by judging the neighborhood inclusion relationship and connectivity between pixels. Pixel category attributes refer to the category labels assigned to pixels after clustering, including core point category, boundary point category, and noise point category.

[0119] It is understandable that the purpose of this step is to preliminarily classify the categories of building area pixels through density connectivity analysis, laying a foundation for the formation of subsequent building groups. The principle is to use the density clustering algorithm to identify the spatial density distribution characteristics of building area pixels, group pixels that are closely connected in space and meet the density requirements into the same connected region, distinguish core points, boundary points and noise points, and solve the problem of feature homogenization caused by the traditional method of treating the entire building area as a homogeneous target. By identifying connected regions with different densities, it provides a basis for distinguishing different types of building groups in the subsequent stage. This step only completes the preliminary determination of pixel categories and does not form the final set of building groups.

[0120] Specifically, when presetting the neighborhood radius and the minimum number of points, refer to the SAR image resolution and the spatial distribution characteristics of the building area. For example, set Eps = 30 (corresponding to a radius of 30 scales), and MinPts = 50 (to ensure that the clustering group has a certain scale). Adaptive parameters can also be set to automatically adjust the Eps value according to the building density in different regions. For example, reduce Eps in dense urban areas to distinguish adjacent building blocks, and increase Eps in sparse areas to merge scattered buildings; use the multi-dimensional DBSCAN clustering algorithm, which not only considers spatial coordinates but also the backscattering intensity as a clustering dimension, and groups pixels that are spatially adjacent and have similar scattering characteristics into one category.

[0121] Furthermore, (1) input the neighborhood radius Eps and the minimum number of points MinPts; (2) in the data set S (reference area set and待定标区域集), select any point P, calculate the number of points within the neighborhood determined by the neighborhood radius of point P, denoted as N(P). If N(P) ≥ MinPts, then P is a core point, create a new class M, and start iteration with P as the initial point to find all points in S that are density-reachable or density-connected to P, and classify them into class M; otherwise, regard P as a noise point; (3) repeat step (2) until all points in S are classified (including noise points).

[0122] Exemplarily, for the obtained target reference building area pixels and target待定标建筑区域像素, use the DBSCAN clustering algorithm for density connectivity analysis. Preset the neighborhood radius Eps = 30 (corresponding to a radius of 30 scales), and the minimum number of points MinPts = 50. Traverse each pixel and calculate the number of pixels N(P) within its Eps neighborhood. If the number N(P) ≥ MinPts, then this pixel is a core point; if the number of pixels N(P) < MinPts within the Eps neighborhood of a certain pixel, but it falls within the Eps neighborhood of a core point, then it is a boundary point; if it is neither a core point nor a boundary point, then it is a noise point. Finally, assign the category attributes of core point, boundary point or noise point to each pixel.

[0123] Step S42: Based on the pixel category attributes, cluster the pixels of the target reference building area and the pixels of the target uncalibrated building area respectively to obtain the reference area set and the uncalibrated area set.

[0124] Understandably, the purpose of this step is to filter out effective building groups, forming a structured set of regions to provide specific processing units for subsequent weight allocation. The principle is based on pixel category attributes, eliminating noise points, and grouping core points and their connected boundary points into a building group. Each group represents a building area with similar density characteristics, distinguishing different types of building groups such as high-density urban areas and sparse industrial areas. This avoids treating the entire building area as a homogeneous target, solving the problem of feature homogenization. Simultaneously, the removal of noise points further reduces outlier interference. This step only completes the filtering and set construction of building groups and does not involve weight calculation.

[0125] Specifically, only the core points and their density-connected boundary points are retained, while noise points are removed, and each independent density-connected region is treated as a building group. The selected building groups are then filtered by size, and small groups with fewer than a preset threshold (e.g., 100 pixels) are removed to reduce the impact of small, low-reliability groups on subsequent processing. Each building group undergoes a spatial integrity check to ensure that the group boundaries are continuous and without obvious breaks. If breaks exist, adjacent density-connected regions are merged.

[0126] For example, based on the determined pixel category attributes, the pixels of the target reference building area and the target uncalibrated building area are filtered: noise points are removed, and the core points and their density-connected boundary points are grouped into a building group to form an initial group set; the group size threshold is set to 100 pixels, and small groups with fewer than 100 pixels in the initial set are removed; spatial integrity checks are performed on the remaining groups, and adjacent and spatially continuous groups are merged to finally obtain a reference area set and an uncalibrated area set containing multiple valid building groups, wherein the reference area set contains 4 valid building groups, and the uncalibrated area set contains the corresponding 4 valid building groups.

[0127] This embodiment, through the above-described scheme, achieves the transformation of building area pixels into structured building groups by density connectivity analysis based on preset parameters and clustering filtering based on category attributes. Density connectivity analysis accurately identifies the density distribution characteristics of building areas, and clustering filtering eliminates noise points and low-reliability small groups, effectively solving the problems of homogenization of building area features and outlier interference. It divides building areas of different densities and sizes into independent groups, providing a reasonable processing unit for subsequent weight allocation based on group characteristics, making weight allocation more targeted, and thus improving the accuracy and stability of SAR radiometric cross-calibration.

[0128] Based on the above implementation scheme, in one feasible implementation, the step of performing feature calculations based on weight allocation on the reference region set and the region set to be calibrated to obtain reference stable features and region set to be calibrated stable features includes S51~S53:

[0129] Step S51: Perform multi-factor feature quantization on the reference region set and the region set to be calibrated, respectively, to obtain the reference weight coefficient and the weight coefficient to be calibrated.

[0130] It should be noted that multi-factor feature quantification is the process of numerically representing multiple key features of a building complex (such as density, scale, and stability), and its purpose is to provide a quantitative basis for the calculation of weight coefficients. The reference weight coefficient is calculated based on the multi-factor features of the reference region set and reflects the numerical contribution of each building complex to the reference stability features. The uncalibrated weight coefficient is calculated based on the multi-factor features of the uncalibrated region set and reflects the numerical contribution of each building complex to the uncalibrated stability features.

[0131] The multiple factors include: density factor (the spatial clustering degree calculated based on the K-nearest neighbor algorithm, taking the reciprocal of the distance to the Kth nearest neighbor, and reducing hyperparameter sensitivity by averaging multiple K values), size factor (a statistical indicator characterizing the number of pixels in a building sub-region), and stability factor (an indicator characterizing the dispersion of the backscattering coefficient of a building sub-region, usually the reciprocal of the standard deviation or the reciprocal of the coefficient of variation). K-nearest neighbor density estimation is a nonparametric density estimation method that measures local density by calculating the reciprocal of the distance from a sample point to its Kth nearest neighbor. Averaging multiple K values ​​refers to calculating the density separately for multiple K values ​​(e.g., K=3, 5, 7) and then averaging the results to reduce the hyperparameter sensitivity caused by the selection of a single K value.

[0132] Understandably, the purpose of this step is to assign reasonable weights to different building groups, addressing the problem of unreasonable weight allocation caused by treating all building areas equally in traditional methods. The principle is that different building groups have different characteristics such as density, size, and stability, and their contributions to the calibration model also differ. Through multi-factor feature quantification, the confidence level of each group is comprehensively reflected, providing a scientific basis for subsequent weighted calculations. This allows high-confidence groups (high density, large scale, high stability) to receive higher weights, improving the estimation accuracy of stable features. This step only calculates the weight coefficients and does not synthesize stable features.

[0133] Specifically, density factor (population density), size factor (number of pixels), and stability factor (reciprocal of the standard deviation of pixel values) are calculated as multi-factor features. The density factor reflects the degree of spatial aggregation of the population, the size factor reflects the size of the population, and the stability factor reflects the stability of the population's radiation characteristics. Normalization is used to standardize each factor to eliminate dimensional differences and make different factors comparable. Then, the weight coefficients are calculated by fusion to obtain the weight coefficients of each population.

[0134] Step S52: Perform a weighted calculation based on the reference weight coefficients and the reference region set to obtain the reference stability characteristics of the reference region set.

[0135] Understandably, the purpose of this step is to synthesize stable radiometric features of the reference image, highlighting the contribution of high-confidence building clusters. The principle is based on reference weighting coefficients, where the pixel mean of each building cluster in the reference region is weighted and summed. The pixel mean of high-weight clusters accounts for a higher proportion of the overall features, while the influence of low-weight clusters is weakened, thus reducing the uncertainty brought by low-confidence clusters and obtaining more reliable stable reference features. This provides an accurate reference benchmark for subsequent calibration constant calculations. This step only completes the synthesis of stable reference features and does not involve the calculation of stable features to be calibrated.

[0136] Specifically, the pixel mean of each building group is calculated as the representative value of the radiation feature of that group; the reference stable feature is calculated using a weighted summation formula: Reference stable feature = Σ (Pixel mean of the i-th reference group × Reference weight coefficient of the i-th group); the validity of the calculation results is verified, and if the stable feature value exceeds the reasonable range (determined based on the radiometric calibration results of the reference image), the weight coefficient is readjusted and the calculation is repeated.

[0137] Furthermore, the formula for calculating the reference stability feature is as follows:

[0138]

[0139] in, For reference stability characteristics; The average pixel value of the reference group; The reference weight coefficient; N represents the number of clusters in the DBSCAN algorithm.

[0140] For example, a weighted calculation is performed on the four building groups in the reference area set: First, the pixel mean of each group is calculated using the arithmetic mean method, which are 0.2493, 0.2606, 0.2094, and 0.2948 respectively; then, combined with the reference weight coefficients [0.0994, 0.7126, 0.1094, 0.0785], a weighted sum is performed: 0.2493×0.0994+0.2606×0.7126+0.2094×0.1094+0.2948×0.0789=0.2567. Finally, the square root is applied according to the Sentinel-1 calibration formula, and the calculation formula is as follows: ≈0.5067, this value is the reference stability characteristic.

[0141] Step S53: Perform a weighted calculation based on the uncalibrated weight coefficients and the uncalibrated region set to obtain the uncalibrated stability characteristics of the uncalibrated region set.

[0142] Understandably, the purpose of this step is to extract reliable radiometric features from the image to be calibrated, establish a correspondence with the reference stable features, and provide core data for radiometric correction. Weighted calculations highlight the contribution of high-confidence building clusters concentrated in the area to be calibrated, ensuring that the stable features accurately reflect the radiometric characteristics of the image to be calibrated. This allows the difference between the two to accurately characterize the radiometric deviation between the two types of images, laying the foundation for subsequent calibration constant calculations. This step only extracts features from the set of areas to be calibrated.

[0143] Specifically, the pixel mean of each building group is calculated as the representative value of the radiation feature of that group; the stable feature to be calibrated is calculated using a weighted summation formula, which is: stable feature to be calibrated = Σ (pixel mean of the i-th group to be calibrated × weight coefficient of the i-th group to be calibrated); the validity of the calculation results is verified, and if the stable feature value exceeds the reasonable range (determined based on the radiation calibration results of the reference image), the weight coefficient is readjusted and the calculation is repeated.

[0144] For example, a weighted calculation is performed on the four building groups in the calibration region set: First, the arithmetic mean is used to calculate the pixel mean of each group, which are 10691, 9539.7, 8000.8, and 11620 respectively; then, combined with the calibration weight coefficients [0.4577, 0.1962, 0.2146, 0.1315], a weighted sum is performed: 106910×0.4577+95397×0.1962+80008×0.2146+116200×0.1315=100009.61. Finally, the square root is applied according to the Sentinel-1 calibration formula, and the calculation formula is as follows: ≈316.38, this value is the stability characteristic to be calibrated.

[0145] This embodiment, through the above-described scheme, achieves a scientific allocation of building group weights via multi-factor feature quantization, followed by weighted calculation to extract stable features. This effectively solves the problems of unreasonable weight allocation and insufficient feature representativeness in existing technologies. Multi-factor quantization comprehensively considers the key features of building groups, and weighted calculation highlights the contribution of high-confidence groups, making the extracted reference stable features and uncalibrated stable features more reliable and representative. This provides a precise basis for subsequent radiometric correction and significantly improves the accuracy and stability of SAR radiometric cross-calibration.

[0146] Based on the above implementation scheme, in one feasible implementation, the step of performing multi-factor feature quantization on the reference region set and the region set to be calibrated respectively to obtain the reference weight coefficient and the weight coefficient to be calibrated includes S61~S64:

[0147] Step S61: Calculate the reference density factor, reference size factor, and reference stability factor of the reference region set.

[0148] It should be noted that the reference density factor is a quantitative indicator characterizing the spatial clustering degree of a building group. It is calculated using a multi-K nearest neighbor algorithm, taking the average density across multiple K values ​​to reduce hyperparameter sensitivity. The calculation process involves randomly selecting a pixel within the group, calculating the reciprocal of the Kth distance from its coordinates as the local density, averaging this across all pixels, and then averaging the densities across multiple K values. The reference size factor is a quantitative indicator characterizing the size of a building group, expressed as the number of pixels in the group. The reference stability factor is a quantitative indicator characterizing the stability of the radiation characteristics of a building group, calculated as the reciprocal of the standard deviation of pixel values; a smaller standard deviation indicates stronger stability. The reference group pixel mean represents the average value of all pixel values ​​within the building group.

[0149] Understandably, the purpose of this step is to provide a quantitative basis for the fusion calculation of reference weight coefficients, comprehensively capturing the characteristic differences of different building groups concentrated in the reference area. The principle is that density, scale, and stability are the core dimensions determining the calibration value of a building group. High density means a more stable spatial structure, large scale means stronger sample representativeness, and high stability means more reliable radiation characteristics. By accurately calculating these three types of factors, comprehensive and objective input data can be provided for subsequent fusion calculations. This step only completes factor calculation and does not involve factor fusion.

[0150] Specifically, referring to the density factor calculation, multi-K nearest neighbor density estimation is used for each building sub-region. Each pixel in Calculate the distance to its Kth nearest neighbor pixel based on its coordinates. Take the reciprocal As a local density measure, to reduce the hyperparameter sensitivity of a single K value, multiple values ​​are taken. Calculate the local density separately and then take the average value. traversal All 100 pixels (where 100 pixels) (where N represents the number of clusters in the DBSCAN algorithm) to calculate the population density. This yields the reference density factor. The reference scale factor is calculated by directly statistically analyzing the building sub-regions. Quantity, calculation The number of pixels is used as a reference scaling factor. A reference stability factor is calculated. Standard deviation of backscattering coefficient values ​​of all pixels Take the reciprocal As a reference stability factor, another implementation method is to use adaptive K-value selection, based on the size of the building sub-region. The range of K values ​​is automatically adjusted (e.g., K=5, 7, 9 for large-scale areas and K=3, 5 for small-scale areas) to adapt to the spatial structure of building groups with different densities.

[0151] For example, for building sub-regions in the reference area set First, count its pixel count. =1500 as the scaling factor; calculate the standard deviation of the backscattering coefficient of each pixel in this region. =0.4, take the reciprocal to get the stability factor. =2.5; Using three nearest neighbor parameters K=3, 5, 7, the distances from each pixel in the region to its 3rd, 5th, and 7th nearest neighbors are calculated (e.g., d3=12m, d5=18m, d7=25m). The reciprocals (0.083, 0.056, 0.040) are taken and averaged to obtain a local density of 0.060. After traversing all pixels, the average is calculated to obtain the population density. =0.12 pixel distance.

[0152] Step S62: Based on the reference density factor, the reference size factor and the reference stability factor, a fusion calculation is performed to obtain the reference weight coefficient of the reference region set.

[0153] Understandably, the purpose of this step is to transform the three types of factors into unified weight coefficients, thereby achieving a comprehensive evaluation of the calibration value of each building group in the reference area set. The three types of factors reflect the calibration value of building groups from different dimensions. They first need to be normalized to make them comparable, and then weighted according to the importance of each dimension for fusion. The resulting weight coefficients accurately reflect the differences in contribution among different building groups, providing a scientific basis for subsequent stable feature calculations.

[0154] Specifically, this is achieved through a geometric fusion method, i.e., by using a comprehensive weight calculation formula:

[0155]

[0156] in, For comprehensive weighting; For reference size factor; As a reference stability factor; This is the reference density factor.

[0157] Furthermore, after calculating the overall weight, weight normalization is performed to obtain the weight coefficients. The formula for weight normalization is as follows:

[0158]

[0159] in, This refers to the weighting coefficient. It is the reference weighting coefficient when calculating the reference weighting coefficient; and the weighting coefficient to be calibrated when calculating the weighting coefficient to be calibrated.

[0160] Step S63: Calculate the uncalibrated density factor, uncalibrated density factor, and uncalibrated stability factor of the uncalibrated region set.

[0161] It should be noted that the definitions and calculation methods of the uncalibrated density factor, uncalibrated size factor, and uncalibrated stability factor are exactly the same as those of the reference density factor, reference size factor, and reference stability factor, respectively. The only difference is that the calculation object is the uncalibrated region set, which ensures that the factor calculations of the two types of region sets are consistent and comparable.

[0162] Understandably, the purpose of this step is to provide a quantitative basis for the fusion calculation of the weight coefficients to be calibrated, corresponding to the factor calculation of the reference area set, and ensuring the consistency of the calculation logic of the two types of weight coefficients. By accurately calculating the density, scale, and stability factors of the reference area set, the characteristic differences of different building groups are fully captured, providing comprehensive and objective input data for subsequent fusion calculations. This step only completes the factor calculation and does not involve factor fusion.

[0163] Specifically, the implementation method of this step is the same as that of step S61, including multi-K nearest neighbor density estimation, pixel number statistical scale, standard deviation inverse calculation stability, etc., to calculate the uncalibrated density factor, uncalibrated density factor and uncalibrated stability factor for the image to be calibrated.

[0164] Step S64: Based on the uncalibrated density factor, the uncalibrated density factor and the uncalibrated stability factor, perform a fusion calculation to obtain the uncalibrated weight coefficient of the uncalibrated region set.

[0165] Understandably, the purpose of this step is to transform the three types of factors in the dataset to be calibrated into unified weight coefficients, consistent with the fusion logic of the reference weight coefficients, ensuring the comparability of the calculations of the two types of stable features. Normalization eliminates dimensional differences, followed by linear weighted fusion based on the same weight percentage, and finally normalization yields the weight coefficients. This accurately reflects the differences in contribution among different building groups within the dataset to be calibrated, providing a scientific basis for subsequent stable feature calculations.

[0166] Specifically, the comprehensive weight is calculated using a geometric fusion method, that is, by using the formula for calculating the comprehensive weight.

[0167] This embodiment, through the above-described scheme and the scientific fusion of density, scale, and stability three-factor quantification, achieves refined evaluation and differentiated weight allocation of building sub-region quality. It effectively reduces hyperparameter sensitivity through multi-K nearest neighbor density estimation, and ensures the synergistic constraint of the three factors through a geometric fusion formula. This accurately identifies high-quality building groups that contribute most to calibration, effectively suppressing the negative impacts of low-density, small-scale, or unstable areas. The weight allocation is more consistent with the physical characteristics of building areas and remote sensing imaging patterns, significantly improving the representativeness of pseudo-invariant building features and the robustness of the calibration model. It provides a reliable statistical weight basis for obtaining high-precision, low-bias calibration constants, achieving high-precision SAR radiometric cross-calibration independent of corner reflectors.

[0168] Based on the above implementation scheme, in a feasible implementation, the step of performing radiometric correction on the target SAR image to be calibrated based on the reference stability feature and the target stability feature to be calibrated, to obtain the target calibrated SAR image, includes S71~S73:

[0169] Step S71: Calculate the calibration constant based on the reference stability feature and the stability feature to be calibrated.

[0170] It should be noted that the calculated value of the calibration constant is a value obtained by subtracting the values ​​of the two types of images, which reflects the radiometric deviation. As the core parameter of radiometric correction, it is used to adjust the radiometric value of the image to be calibrated.

[0171] Understandably, the purpose of this step is to determine the radiometric deviation between the two types of images, providing a quantitative basis for radiometric correction. The principle is that the reference image has undergone precise absolute calibration, resulting in accurate and reliable radiometric values. The radiometric values ​​of the image to be calibrated deviate from those of the reference image. By subtracting the stability characteristics of the two images, this deviation can be precisely quantified, yielding a calculated calibration constant. This constant reflects the radiometric offset of the image to be calibrated relative to the reference image, providing direct parameters for subsequent correction operations. This step only calculates the calibration constant and does not verify its accuracy.

[0172] Specifically, in one embodiment of this application, the calibration constant is calculated directly by subtraction: Calculated value of calibration constant = Reference stable feature - Stable feature to be calibrated. This method is simple and intuitive, and suitable for scenarios where the differences in stable features are linearly correlated. In another embodiment of this application, the calibration constant is calculated by weighted subtraction: If the reference stable feature and the stable feature to be calibrated consist of features from multiple bands or polarization channels, weights can be set according to the importance of each channel, and weighted subtraction can be performed. The formula is: Calculated value of calibration constant = Σ(Reference channel stable feature × Channel weight) - Σ(Stable feature to be calibrated channel × Channel weight), which is suitable for multi-channel SAR image calibration.

[0173] For example, the difference is calculated based on the obtained reference stable feature and the stable feature to be calibrated: the reference stable feature is 0.5067, and the stable feature to be calibrated is 316.38. Using the direct difference calculation method, the calibration constant is calculated as 10log10(316.68)-10log10(0.5067)=27.955dB. This value reflects the radiometric offset of the image to be calibrated relative to the reference image, providing core parameters for subsequent radiometric correction.

[0174] Step S72: Calculate the mean square error based on the calculated value of the calibration constant and the preset theoretical value of the calibration constant to obtain the calibration accuracy.

[0175] It should be noted that the preset theoretical values ​​of calibration constants are based on pre-determined standard calibration constant values ​​according to SAR sensor parameters, imaging geometry, etc., and serve as a benchmark for measuring calibration accuracy. A table of theoretical calibration constant values ​​is generated beforehand, and these values ​​are looked up when needed. Mean square error (MSE) calculation refers to the method of calculating the average of the squares of the deviations between the calculated and theoretical values ​​of the calibration constants. Calibration accuracy is an indicator reflecting the accuracy of the calibration constant calculation, obtained after calculating the MSE; the smaller the value, the more reliable the calibration result.

[0176] Understandably, the purpose of this step is to verify the accuracy of the calculated calibration constant and ensure the reliability of the radiometric correction. The principle is that the theoretical value of the calibration constant is a true radiometric deviation benchmark determined based on the inherent characteristics of the sensor and imaging conditions. By calculating the mean square error between the calculated and theoretical values ​​of the calibration constant, the degree of deviation in the calibration results can be quantified, and it can be determined whether there are errors in the calibration process. If the calibration accuracy meets the preset requirements (e.g., less than 0.3 dB), subsequent correction can be performed; if not, the parameters of the preceding steps need to be adjusted retrospectively. This step only verifies the calibration accuracy and does not perform radiometric correction.

[0177] Specifically, first, calculate the Mean Squared Error (MSE): the formula is MSE = (Calculated value of calibration constant - Theoretical value of calibration constant)², which is suitable for verifying the accuracy of a single calibration result. The Root Mean Squared Error (RMSE) can also be calculated: the formula is RMSE = (n is the number of calibrations), applicable to the comprehensive accuracy verification of multiple calibration results. The determined accuracy threshold can be set according to the actual application requirements, such as an absolute calibration accuracy threshold of 0.3dB and a relative calibration accuracy threshold of 0.3dB, used to determine whether the calibration results are qualified.

[0178] For example, the mean square error (MSE) is calculated based on the calculated value of the calibration constant (27.955 dB) and the preset theoretical value of the calibration constant (27.865 dB): Using the MSE calculation method, MSE = (27.955 - 27.865)² = 0.0081, corresponding to an absolute calibration accuracy of... dB (adjusted here based on example data; accuracy is actually represented by mean square error). This accuracy has been verified to meet the preset 0.3dB threshold requirement, and subsequent radiation correction can be performed.

[0179] Step S73: Perform radiometric correction on the target SAR image to be calibrated based on the calibration accuracy to obtain the target calibrated SAR image.

[0180] Understandably, the purpose of this step is to correct the radiometric values ​​of the image to be calibrated, ensuring it has a radiometric scale consistent with the reference image and meeting the requirements of quantitative remote sensing applications. The principle is to perform radiometric adjustments on each pixel value of the target SAR image to be calibrated based on the calculated calibration constant value that has passed calibration accuracy verification. This eliminates radiometric bias between the two types of images, enabling the radiometric characteristics of the target SAR image to accurately reflect the backscattering characteristics of ground objects and maintain good comparability with the reference image. This step is the final execution stage of the entire calibration process, completing the calibration objective.

[0181] Specifically, in one embodiment of this application, linear correction is employed: for each pixel value of the target SAR image to be calibrated, linear adjustment is performed according to the formula "corrected pixel value = original pixel value + calculated calibration constant value," which is suitable for scenarios where the radiation deviation is a constant value. In another embodiment of this application, segmented correction is employed: if there are differences in the deviation between different radiation intensity ranges, the image pixel values ​​are divided into multiple ranges, and a corresponding calculated calibration constant value is set for each range, and correction is performed separately, which is suitable for scenarios where the radiation deviation changes with the radiation intensity. After correction, verification is performed by verifying the radiation characteristics of the target calibrated SAR image and checking its radiation consistency with the reference image to ensure the correction effect.

[0182] For example, based on the verified calibration accuracy and the calculated calibration constant value (27.955dB), radiometric correction is performed on the target SAR image to be calibrated: a linear correction method is used to adjust the value of each pixel in the target SAR image to be calibrated one by one; after the correction is completed, the pixels of the building area in the target calibrated SAR image are extracted, its radiometric characteristics are calculated, and compared with the radiometric characteristics of the reference image to verify the radiometric consistency between the two, and finally a target calibrated SAR image with accurate radiometric values ​​and consistent radiometric scale with the reference image is obtained.

[0183] This embodiment, through the above-described scheme, ensures the accuracy and reliability of radiometric correction by employing a process of calculating calibration constants through difference, verifying calibration accuracy through mean square error, and performing radiometric correction based on the accuracy. The precise calculation of the calibration constants accurately quantifies the radiometric deviation between the two types of images. Verification of calibration accuracy provides quality assurance for the correction operation, avoiding correction failures caused by using inaccurate calibration constants. The radiometric correction step, based on verified parameters, accurately adjusts the radiometric values ​​of the image to be calibrated, effectively solving the problem of inaccurate radiometric deviation correction in traditional calibration. Ultimately, it yields a target-calibrated SAR image with reliable radiometric characteristics and good comparability to the reference image, meeting the needs of quantitative remote sensing applications and data fusion.

[0184] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the SAR radiation cross-calibration method based on building area feature weight allocation in this application. Any simple transformations based on this technical concept are within the protection scope of this application.

[0185] This application also provides a SAR radiation cross-calibration device based on building area feature weight allocation. Please refer to [reference needed]. Figure 4 The SAR radiation cross-calibration device based on building area feature weight allocation includes:

[0186] The mask extraction module 401 is used to extract the mask of the target reference SAR image and the target uncalibrated SAR image based on the land cover type product, so as to obtain the target reference building area pixels and the target uncalibrated building area pixels.

[0187] Data clustering module 402 is used to perform data clustering on the target reference building area pixels and the target uncalibrated building area pixels using a target clustering algorithm to obtain a reference area set and an uncalibrated area set;

[0188] Feature calculation module 403 is used to perform feature calculation based on weight allocation on the reference region set and the region set to be calibrated to obtain reference stable features and region set to be calibrated stable features;

[0189] Radiometric correction 404 is used to perform radiometric correction on the target SAR image to be calibrated based on the reference stability features and the stability features to be calibrated, so as to obtain the target calibrated SAR image.

[0190] The SAR radiation cross-calibration device based on building area feature weight allocation provided in this application has the same beneficial effects as the SAR radiation cross-calibration method based on building area feature weight allocation provided in the above embodiments. Furthermore, the other technical features in the SAR radiation cross-calibration device based on building area feature weight allocation are the same as those disclosed in the above embodiments, and will not be repeated here.

[0191] This application provides a SAR radiation cross-calibration device based on building area feature weight allocation. The SAR radiation cross-calibration device based on building area feature weight allocation includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the SAR radiation cross-calibration method based on building area feature weight allocation in the above embodiment 1.

[0192] The following is for reference. Figure 5 This document illustrates a structural schematic diagram of a SAR radiation cross-calibration device suitable for implementing the building area feature weight allocation embodiments of this application. The SAR radiation cross-calibration device based on building area feature weight allocation in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle terminals (e.g., vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The SAR radiation cross-calibration device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0193] like Figure 5As shown, the SAR radiometric cross-calibration device based on building area feature weight allocation may include a processing unit 1001 (e.g., a central processing unit, graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in read-only memory 1002 or a program loaded from storage device 1003 into random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the SAR radiometric cross-calibration device based on building area feature weight allocation. The processing unit 1001, read-only memory 1002, and random access memory 1004 are interconnected via bus 1005. Input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to input / output interface 1006: input devices 1007 including, for example, touch screens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays, speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the SAR radiometric cross-calibration device based on building area feature weight allocation to communicate wirelessly or wiredly with other devices to exchange data. While the figures show SAR radiometric cross-calibration devices based on building area feature weight allocation with various systems, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0194] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0195] The SAR radiometric cross-calibration device based on building area feature weight allocation provided in this application adopts the SAR radiometric cross-calibration method based on building area feature weight allocation in the above embodiments, which can solve the technical problem of insufficient accuracy in SAR radiometric cross-calibration. Compared with the prior art, the beneficial effects of the SAR radiometric cross-calibration device based on building area feature weight allocation provided in this application are the same as the beneficial effects of the SAR radiometric cross-calibration method based on building area feature weight allocation provided in the above embodiments, and other technical features in the SAR radiometric cross-calibration device based on building area feature weight allocation are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0196] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0197] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0198] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the SAR radiometric cross-calibration method based on building area feature weight allocation in the above embodiments.

[0199] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory or flash memory, optical fiber, portable compact disk read-only memory (CD-Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency, etc., or any suitable combination thereof.

[0200] The aforementioned computer-readable storage medium may be included in a SAR radiation cross-calibration device based on building area feature weight allocation; or it may exist independently and not assembled into a SAR radiation cross-calibration device based on building area feature weight allocation.

[0201] The aforementioned computer-readable storage medium carries one or more programs that, when executed by a SAR radiometric cross-calibration device based on building area feature weight allocation, cause the SAR radiometric cross-calibration device based on building area feature weight allocation to: perform mask extraction on a target reference SAR image and a target uncalibrated SAR image based on land cover type products to obtain target reference building area pixels and target uncalibrated building area pixels; perform data clustering on the target reference building area pixels and the target uncalibrated building area pixels using a target clustering algorithm to obtain a reference area set and a uncalibrated area set; perform feature calculation based on weight allocation on the reference area set and the uncalibrated area set to obtain reference stable features and uncalibrated stable features; and perform radiometric correction on the target uncalibrated SAR image based on the reference stable features and the uncalibrated stable features to obtain a target calibrated SAR image.

[0202] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0203] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0204] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0205] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described SAR radiometric cross-calibration method based on building area feature weight allocation, thereby solving the technical problem of insufficient accuracy in SAR radiometric cross-calibration. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the SAR radiometric cross-calibration method based on building area feature weight allocation provided in the above embodiments, and will not be repeated here.

[0206] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the SAR radiation cross-calibration method based on building area feature weight allocation as described above.

[0207] The computer program product provided in this application can solve the technical problem of insufficient accuracy in SAR radiometric cross-calibration. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the SAR radiometric cross-calibration method based on building area feature weight allocation provided in the above embodiments, and will not be repeated here.

[0208] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A SAR radiation cross-calibration method based on building area feature weight distribution, characterized in that, The SAR radiometric cross-calibration method based on building area feature weight allocation includes: Based on the land cover type product, mask extraction is performed on the target reference SAR image and the target uncalibrated SAR image to obtain the target reference building area pixels and the target uncalibrated building area pixels. The target reference building region pixels and the target uncalibrated building region pixels are clustered using a target clustering algorithm to obtain a reference region set and an uncalibrated region set. The reference region set and the region set to be calibrated are subjected to feature calculation based on weight allocation to obtain reference stable features and region set to be calibrated stable features. Radiometric correction is performed on the target SAR image to be calibrated based on the reference stability features and the stability features to be calibrated to obtain the target SAR image. The step of clustering the target reference building region pixels and the target uncalibrated building region pixels using a target clustering algorithm to obtain a reference region set and an uncalibrated region set includes: Based on a preset neighborhood radius and minimum number of points, density connectivity analysis is performed on the pixels of the target reference building area and the pixels of the target uncalibrated building area using a target clustering algorithm to determine the pixel category attributes of the pixels of the target reference building area and the pixels of the target uncalibrated building area. Based on the pixel category attributes, clustering and filtering are performed on the pixels of the target reference building area and the pixels of the target uncalibrated building area to obtain the reference area set and the uncalibrated area set; The step of performing feature calculations based on weight allocation on the reference region set and the region set to be calibrated to obtain reference stable features and region set to be calibrated stable features includes: Multi-factor feature quantization is performed on the reference region set and the region set to be calibrated, respectively, to obtain the reference weight coefficient and the weight coefficient to be calibrated. The reference stability characteristics of the reference region set are obtained by performing a weighted calculation based on the reference weight coefficients and the reference region set. The uncalibrated stability characteristics of the uncalibrated region set are obtained by performing a weighted calculation based on the uncalibrated weight coefficients and the uncalibrated region set.

2. The SAR radiation cross-calibration method based on building area feature weight distribution according to claim 1, wherein, Before the step of performing mask extraction on the target reference SAR image and the target uncalibrated SAR image based on the land cover type product to obtain the target reference building area pixels and the target uncalibrated building area pixels, the following is also included: Acquire the initial reference SAR image and the initial SAR image to be calibrated; The initial reference SAR image and the initial uncalibrated SAR image are registered to obtain an intermediate reference SAR image and an intermediate uncalibrated image. The intermediate reference SAR image and the intermediate image to be calibrated are preprocessed to obtain the target reference SAR image and the target image to be calibrated.

3. The SAR radiation cross-calibration method based on building area feature weight distribution according to claim 1, wherein, The step of performing mask extraction on the target reference SAR image and the target uncalibrated SAR image based on land cover type products to obtain the target reference building area pixels and the target uncalibrated building area pixels includes: Building area mask extraction was performed on the target reference SAR image and the target uncalibrated SAR image using land cover type products to obtain the initial reference building area pixels and the initial uncalibrated building area pixels. The pixels of the initial reference building area and the pixels of the initial building area to be calibrated are filtered based on a threshold to obtain the pixels of the target reference building area and the pixels of the target building area to be calibrated.

4. The SAR radiation cross-calibration method based on building area feature weight distribution according to claim 1, wherein, The step of performing multi-factor feature quantization on the reference region set and the region set to be calibrated respectively to obtain the reference weight coefficients and the weight coefficients to be calibrated includes: Calculate the reference density factor, reference size factor, and reference stability factor of the reference region set; The reference weight coefficients of the reference region set are obtained by fusing the reference density factor, the reference size factor, and the reference stability factor. Calculate the uncalibrated density factor, uncalibrated density factor, and uncalibrated stability factor of the set of regions to be calibrated; The uncalibrated weight coefficients of the uncalibrated region set are obtained by fusion calculation based on the uncalibrated density factor, the uncalibrated density factor and the uncalibrated stability factor.

5. The SAR radiation cross-calibration method based on building area feature weight distribution according to claim 1, wherein, The step of performing radiometric correction on the target SAR image to be calibrated based on the reference stability features and the target stability features to be calibrated, to obtain the target calibrated SAR image, includes: The calibration constant is calculated based on the reference stability feature and the stability feature to be calibrated. The mean square error is calculated based on the calculated value of the calibration constant and the preset theoretical value of the calibration constant to obtain the calibration accuracy; Based on the calibration accuracy, radiometric correction is performed on the SAR image of the target to be calibrated to obtain the SAR image of the target.

6. A SAR radiation cross-calibration device based on building area feature weight distribution, characterized in that, The SAR radiation cross-calibration device based on building area feature weight allocation includes: The mask extraction module is used to extract the mask from the target reference SAR image and the target uncalibrated SAR image based on the land cover type product, so as to obtain the target reference building area pixels and the target uncalibrated building area pixels. The data clustering module is used to perform data clustering on the pixels of the target reference building area and the pixels of the target uncalibrated building area using a target clustering algorithm to obtain a reference area set and an uncalibrated area set. The feature calculation module is used to perform feature calculation based on weight allocation on the reference region set and the region set to be calibrated, so as to obtain reference stable features and calibrated stable features. A radiometric correction module is used to perform radiometric correction on the target SAR image to be calibrated based on the reference stability features and the stability features to be calibrated, so as to obtain a target calibrated SAR image. The data clustering module is further configured to perform density connectivity analysis on the pixels of the target reference building area and the pixels of the target uncalibrated building area based on a preset neighborhood radius and minimum number of points, using a target clustering algorithm, to determine the pixel category attributes of the pixels of the target reference building area and the pixels of the target uncalibrated building area. Based on the pixel category attributes, clustering and filtering are performed on the pixels of the target reference building area and the pixels of the target uncalibrated building area to obtain the reference area set and the uncalibrated area set; The feature calculation module is further used to perform multi-factor feature quantization on the reference region set and the region set to be calibrated, respectively, to obtain reference weight coefficients and calibrated weight coefficients. The reference stability characteristics of the reference region set are obtained by performing a weighted calculation based on the reference weight coefficients and the reference region set. The uncalibrated stability characteristics of the uncalibrated region set are obtained by performing a weighted calculation based on the uncalibrated weight coefficients and the uncalibrated region set.

7. A SAR radiation cross-calibration device based on building area feature weight distribution, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the SAR radiometric cross-calibration method based on building area feature weight allocation as described in any one of claims 1 to 5.

8. A storage medium, characterized by The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the SAR radiation cross-calibration method based on building area feature weight allocation as described in any one of claims 1 to 5.

Images