A sar polarization calibration reference distributed target extraction method
By extracting polarization distortion parameters and orientation angles from uncalibrated fully polarimetric SAR images, and combining linear iterative clustering and random forest classifiers, the high threshold setting difficulty and erroneous extraction problems of existing reference target extraction methods are solved, achieving high-precision and automated distributed target extraction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA CENT FOR RESOURCES SATELLITE DATA & APPL
- Filing Date
- 2024-08-30
- Publication Date
- 2026-06-09
AI Technical Summary
Existing reference target extraction methods are difficult to set thresholds, prone to erroneous extraction, and difficult to achieve robust automated processing.
By extracting polarization distortion parameters and polarization orientation angles from uncalibrated fully polarimetric SAR images for threshold-based pixel classification, and combining this with a linear iterative clustering algorithm for superpixel segmentation, a random forest classifier is trained using the circular polarization basis correlation coefficient and reflection symmetry coefficient features to obtain a reference distributed target that meets the polarization calibration requirements.
It reduces reliance on complex prior information, improves the accuracy and precision of classification, and achieves automated, robust, and distributed reference target extraction across the entire map.
Smart Images
Figure CN119335484B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method, apparatus, device, and storage medium for extracting reference distributed targets, belonging to the field of synthetic aperture radar polarization processing. Background Technology
[0002] Polarimetric calibration plays a crucial role in the quantitative analysis and application of polarimetric synthetic aperture radar (PolSAR) images. Distributed target polarimetric calibration methods, such as the Van Zyl, Quegan, and Ainsworth algorithms, have become increasingly important due to their reduced reliance on manual calibrators and ability to cover larger calibration areas. However, Quegan-type algorithms rely on distributed targets dominated by volume scattering with reciprocity and reflection symmetry. When these characteristics are not met, they may encounter challenges in accurately estimating polarimetric distortion parameters, leading to decreased accuracy and stability.
[0003] In most images, both natural and man-made targets exist. Only a subset of natural targets exhibit reciprocity and reflection symmetry, making them directly usable as reference targets in the Quegan algorithm. To address the challenge of manually selecting reference targets, polarization features are needed for their extraction. Existing solutions utilize selected polarization features combined with thresholding for target extraction. However, threshold setting requires substantial prior information, lacks concrete physical meaning, is prone to erroneous target extraction, and is difficult to automate robustly. Summary of the Invention
[0004] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide a SAR polarization calibration reference distributed target extraction method, which solves the technical problems of high threshold setting difficulty, easy error extraction of reference targets, and difficulty in achieving robust automated processing in the existing reference target extraction methods.
[0005] The technical solution of this invention is: a SAR polarization calibration reference distributed target extraction method, comprising:
[0006] Using polarization distortion parameters and polarization orientation angles extracted from uncalibrated fully polarimetric SAR images, a threshold-based pixel classification method was used to obtain pixel-level high-confidence volume scattering-dominant targets and non-volume scattering-dominant targets.
[0007] Superpixel segmentation is performed on uncalibrated fully polarimetric SAR images to obtain superpixel units;
[0008] Two polarization features, circular polarization basis correlation coefficient and reflection symmetry coefficient, are extracted from the superpixel unit;
[0009] The pixel ratio threshold method is used to process the high-confidence volume scattering dominated and non-volume scattering dominated targets at the pixel level, and the high-confidence volume scattering dominated and non-volume scattering dominated targets of superpixel units are obtained. The superpixel unit positive and negative samples are obtained by combining the two polarization features, the circular polarization basis correlation coefficient and the reflection symmetry coefficient.
[0010] A random forest classifier is used to train the positive and negative samples of the superpixel units to obtain the full image classification result. The dominant target of the superpixel unit volume scattering in the full image classification result is the reference distributed target that meets the polarization calibration requirements.
[0011] The thresholding method for pixel classification, using polarization distortion parameters and polarization orientation angles extracted from uncalibrated fully polarimetric SAR images, yields pixel-level high-confidence volume scattering-dominant targets and non-volume scattering-dominant targets, including:
[0012] The reciprocity angle is calculated using uncalibrated fully polarimetric SAR images. The reciprocity angle is then compared with a reciprocity judgment threshold, and targets corresponding to reciprocity angles exceeding the threshold are removed.
[0013] The Quegan polarization calibration algorithm was used to process the uncalibrated fully polarimetric SAR image after the removal process to obtain crosstalk distortion parameters and channel imbalance distortion parameters.
[0014] Set the physical meaning thresholds for the phase standard deviation of the channel imbalance distortion parameter and the crosstalk distortion parameter. Determine whether the phase standard deviation of the channel imbalance distortion parameter and the crosstalk distortion parameter of each pixel exceed the two set thresholds. If either one exceeds the threshold, the classification result of the pixel is a high-confidence non-volume scattering dominant target.
[0015] For each pixel in the uncalibrated fully polarimetric SAR image after removal processing, the polarization orientation angle is calculated. A physical meaning threshold for the standard deviation of the polarization orientation angle is set. The classification result of the pixel corresponding to the standard deviation of the polarization orientation angle exceeding the physical meaning threshold of the standard deviation of the polarization orientation angle is the pixel-level high confidence volume scattering dominant target.
[0016] The physical threshold for the phase standard deviation of the channel imbalance distortion parameter is set to 140°, and the physical threshold for the crosstalk distortion parameter is set to -6dB.
[0017] The physical meaning threshold for the standard deviation of the polarization orientation angle is set to 29°.
[0018] The calculation of the polarization orientation angle of the uncalibrated fully polarimetric SAR image after the removal process is specifically as follows:
[0019]
[0020] Where POA represents the polarization orientation angle; <·> represents the spatial averaging operation; * represents the conjugate operation; H represents horizontal polarization; V represents vertical polarization; M ij Represents the elements of the observed polarization scattering matrix.
[0021] The uncalibrated fully polarimetric SAR image is segmented into superpixel units using a linear iterative clustering algorithm, including:
[0022] 1) Set the number of superpixels and seed points, and distribute the seed points evenly within the uncalibrated fully polarimetric SAR image;
[0023] 2) Process the uncalibrated fully polarimetric SAR image to obtain the Pauli pseudocolor image of the uncalibrated fully polarimetric SAR image. Calculate the gradient value of all pixels in the Pauli pseudocolor image within the 3×3 neighborhood of each seed point. Move the seed point to the place with the smallest gradient in the neighborhood to obtain the reselected initial seed point, i.e. the initial cluster center.
[0024] 3) Use the initial cluster centers to search for pixels within a 2S×2S neighborhood. For each searched pixel, calculate its color distance d to the initial cluster centers within a 2S×2S range. c Spatial distance d s The distance metric D is calculated; S represents the distance between two adjacent seed points.
[0025] 4) Sort the distance metrics between the pixels and each initial cluster center, and define the initial cluster center corresponding to the minimum distance metric D as the new cluster center of the pixel;
[0026] 5) Calculate the average spatial location of the pixels assigned to each new cluster center to obtain the spatial location of the new cluster center;
[0027] 6) Iterate through steps 3)-5) until the spatial location of the cluster centers no longer changes, and obtain the final clustering result; define the pixel of each cluster center as a superpixel unit.
[0028] The setting of the number of superpixels and the number of seed points includes: setting the number of superpixels = the number of seed points = the total number of pixels / (S×S).
[0029] The distance metric Where, N c This indicates the brightness range of the Lab color model.
[0030] The extraction of two polarization features—circular polarization basis correlation coefficient and reflection symmetry coefficient—from superpixel units includes:
[0031] Spatial averaging of superpixel units yields the second-order statistical covariance matrix;
[0032] The reflection symmetry coefficient is calculated using the second-order statistic covariance matrix.
[0033] A polarization basis transformation is performed on the uncalibrated fully polarimetric SAR image to convert it from a linear polarization basis to a circular polarization basis, that is, to convert the horizontal and vertical polarizations into left-hand circular polarizations and right-hand circular polarizations, thus obtaining a circular polarization basis image.
[0034] The correlation coefficient of the circular polarization basis is calculated using the obtained circular polarization basis image.
[0035] Reflection symmetry coefficient Q rs The calculation method is as follows:
[0036]
[0037] C ′ (θ)=U(θ)C ′ U T (θ)
[0038]
[0039] Where det(·) represents the determinant operation of a matrix, C' represents the second-order statistic covariance matrix, and C 11 Let C' be a 2×2 matrix with its top left corner, and C... 22 Let C' be a 1×1 matrix in the lower right corner.
[0040] The circular polarization basis correlation coefficient ρ is calculated using the obtained circular polarization basis image. RRLL ,include:
[0041]
[0042] Where R represents right-hand circular polarization, L represents left-hand circular polarization, and i represents the imaginary part.
[0043] The pixel proportion threshold method is used to process pixel-level high-confidence volume scattering-dominated and non-volume scattering-dominated targets to obtain superpixel units of high-confidence volume scattering-dominated and non-volume scattering-dominated targets. These targets are then combined with two polarization features—the correlation coefficient between the non-volume scattering-dominated target and the circular polarization basis, and the reflection symmetry coefficient—to obtain positive and negative samples of the superpixel units, including:
[0044] Using the pixel proportion threshold method, the proportion of high-confidence volume scattering dominant targets in each superpixel unit is calculated. A proportion threshold is set, and superpixel units with a proportion of high-confidence volume scattering dominant targets exceeding the proportion threshold are identified as volume scattering dominant targets. The positive samples of superpixel units are obtained by combining the characteristics of the circular polarization basis correlation coefficient and reflection symmetry coefficient of the superpixel unit.
[0045] Using the pixel proportion threshold method, the proportion of high-confidence non-volume scattering dominant targets in each superpixel unit is calculated. A proportion threshold is set, and superpixel units with a proportion of high-confidence non-volume scattering dominant targets exceeding the proportion threshold are identified as non-volume scattering dominant targets. The results are fused with the circular polarization basis correlation coefficient and reflection symmetry coefficient features to obtain the negative samples of the superpixel unit.
[0046] The ratio threshold is set to 50%.
[0047] The beneficial effects of this invention are as follows: In this invention, polarization distortion parameters and polarization orientation angles are extracted from uncalibrated fully polarimetric SAR images, and threshold-based pixel classification is performed to obtain pixel-level high-confidence targets dominated by volume scattering and those not dominated by volume scattering. The advantage of this method is that it reduces the dependence on complex prior information, enables more intuitive target differentiation, and improves classification accuracy. Furthermore, a linear iterative clustering algorithm is used to perform superpixel segmentation on the uncalibrated fully polarimetric SAR image to obtain distributed target extraction superpixel units. The advantage of superpixel segmentation is that it divides the image into smaller regions with similar features, reducing computational complexity while preserving the edge information of the target and improving the accuracy of target extraction. The superpixel units obtained in the previous step are subjected to circular polarization basis correlation coefficient and reflectance... Extracting the two polarization features of symmetry coefficients has the advantage of improving the feature representation capability of superpixels, providing richer and more accurate feature information for subsequent classification. Finally, using the pixel proportion threshold method, the high-confidence volume scattering dominated and non-volume scattering dominated targets at the pixel level are obtained as high-confidence volume scattering dominated and non-volume scattering dominated targets in superpixel units. These targets are then fused with the circular polarization basis correlation coefficient and reflection symmetry coefficient features to obtain positive and negative samples of superpixel units. A random forest classifier is used to train the samples to obtain the full-image classification result. The high-confidence volume scattering dominated targets in the superpixel units are the reference distributed targets that meet the polarization calibration requirements. This method has good generalization ability and can achieve automatic classification of the entire image, effectively extracting reference distributed targets that meet the polarization calibration requirements. Attached Figure Description
[0048] Figure 1 This is a flowchart of the method of the present invention.
[0049] Figure 2 To demonstrate the Q values of a built-up area target (A) dominated by non-volume scattering and a vegetation target (B) dominated by volume scattering. rs Feature sequence; Detailed Implementation
[0050] This invention provides a method for extracting distributed targets using SAR polarization calibration references, referring to... Figure 1 , Figure 1This is a flowchart illustrating the first embodiment of the SAR polarization calibration reference distributed target extraction method of the present invention.
[0051] In this embodiment, the SAR polarization calibration reference distributed target extraction method includes the following steps:
[0052] Step S10: Extract polarization distortion parameters and polarization orientation angles from the uncalibrated fully polarimetric SAR image, perform thresholding pixel classification, and obtain pixel-level high-confidence volume scattering-dominated and non-volume scattering-dominated targets.
[0053] It should be noted that ground feature distribution is often complex and non-uniform, making it impossible to satisfy the reciprocity assumption for simple, uniformly distributed targets. Existing polarization calibration algorithms all require reference targets to satisfy the reciprocity assumption. However, in actual acquired polarization data, many distributed targets often fail to satisfy reciprocity due to propagation effects, interactions with special objects, weak backscattering cross sections, etc. Therefore, non-reciprocal scatterers are first removed. This can be measured using the projection angle θ between the target and the subspace.
[0054]
[0055] in, The unit vector of the observed polarization scattering matrix:
[0056]
[0057] H = |M HH | 2 +|M HV | 2 +|M VH | 2 +|M VV | 2
[0058] Where H represents horizontal polarization, V represents vertical polarization, and M ij Represents the elements of the observed polarization scattering matrix.
[0059] The norm of reciprocal projection is expressed as:
[0060]
[0061] When θ = 0, the scattering matrices are strictly reciprocal; when θ = π / 2, the scattering matrices are not reciprocal at all. π / 4 can be considered as the reciprocity threshold; targets exceeding this threshold are removed, and the remaining targets are processed further.
[0062] It should be noted that the polarization distortion model can be approximated by the following formula:
[0063] M = R·S·T + N
[0064] Where M is the observed polarization scattering matrix, R and T represent the receive and transmit channel distortion matrices, respectively, S is the theoretical polarization scattering matrix, and N is the system noise matrix. Ignoring system noise, the above equation can be rewritten as:
[0065]
[0066] Where H represents horizontal polarization, V represents vertical polarization, and S ij Represents the elements of the theoretical polarization scattering matrix, r ij Represents the elements of the received channel distortion matrix, t ij This represents the elements of the transmit channel distortion matrix.
[0067] After vectorizing the scattering matrix S, the following equation is obtained:
[0068] m = D·s
[0069] m = [M HH M HV M VH M VV ] T ,s=[Y HH ,S HV ,S VH ,S VV ] T
[0070] Where the superscript T denotes the transpose operation, and D denotes the redefined distortion matrix, which is composed of the distortion matrices R and T, and is defined as follows:
[0071]
[0072] in, Let Y be the Kronecker product, Y be the system gain, α be the cross-polarization channel imbalance distortion, u, v, w, z be the crosstalk distortion, and k be the co-polarization channel imbalance distortion. These parameters are defined as follows:
[0073]
[0074] Where N is the total number of pixels within the sliding window, α i Let α and μ be the values of the i-th pixel within the sliding window. α Let α be the mean value of α within the sliding window.
[0075] In built-up areas and other areas with artificial features, the crosstalk amplitude |u| estimated by the Quegan algorithm is much higher than the actual system distortion, typically ranging from -5dB to -20dB. In water areas, the cross-polarity channel imbalance phase Std(arg(α)) estimated by the Quegan algorithm is high, fluctuating between 20° and 150°. Therefore, physical thresholds slightly lower than the upper limits of the above ranges are set for the two parameters |u| and Std(arg(α)) estimated by the Quegan algorithm: -6dB and 140°, respectively. Pixel classification is then performed, and the classification results exceeding the thresholds are combined to obtain high-confidence artificial structures and other non-volume scattering-dominated targets. These two parameters have clear physical meanings, and the polarization distortion range of the SAR system is easy to estimate roughly, providing prior information for threshold setting and reducing the difficulty of threshold setting.
[0076] The polarization orientation angle (POA) represents the angular relationship between the target plane and the radar line of sight. Vegetated areas are typically dominated by volume scattering, with Std(POA) values generally ranging from 10° to 30°. Other objects have more stable Std(POA) values, typically within 5°. Therefore, a physical Std(POA) threshold of 29°, slightly below the upper limit of this range, is chosen for pixel classification. Targets exceeding this threshold are classified as high-confidence volume scattering targets such as forests, shrubs, and grasslands. The POA estimation method is as follows:
[0077]
[0078] Where <·> represents spatial averaging, Re(·) represents the real part operation, * represents the conjugate operation, N is the total number of pixels in the sliding window, and POA i Let μ be the POA of the i-th pixel within the sliding window. POA This represents the mean of POA within the sliding window.
[0079] By applying the aforementioned polarization feature parameters and combining them with thresholds that have clear physical meaning, high-confidence targets dominated by volume scattering and those not dominated by volume scattering can be extracted. The advantage of this method is that it reduces reliance on complex prior information, enables more intuitive target differentiation, and improves classification accuracy.
[0080] Step S20: Use a linear iterative clustering algorithm to perform superpixel segmentation on the uncalibrated fully polarimetric SAR image to obtain superpixel units for distributed target extraction.
[0081] In the specific implementation, the number of pixels contained in each superpixel is first set to 10×10. The number of superpixels is calculated in combination with the image size: total number of pixels / (10×10). Seed points are evenly distributed in the image with a step size of S = 10. The number of seed points is the total number of pixels / (10×10).
[0082] Then, the uncalibrated fully polarimetric SAR image is processed to obtain the Pauli pseudocolor image. The gradient values of all pixels in the 3×3 neighborhood of the seed point are calculated. The seed point is moved to the place with the smallest gradient in the neighborhood to obtain the reselected initial seed point, i.e. the initial cluster center.
[0083] Using the initial cluster centers obtained in the previous step, a pixel search is performed within a 2S×2S neighborhood. For each searched pixel, its color and spatial distance to the surrounding cluster centers within a 2S×2S range are calculated, and a distance metric is synthesized.
[0084] Color distance d c The calculation method is as follows:
[0085]
[0086] Where l, a, and b represent brightness, the range from magenta to green, and the range from yellow to blue in the Lab color model, respectively.
[0087] Spatial distance d s The calculation method is as follows:
[0088]
[0089] Where x and y are the horizontal and vertical coordinates of the pixel, respectively.
[0090] The distance metric synthesis method is as follows:
[0091]
[0092] Where, N c This represents the maximum color distance, with a value range of [1, 40], and is typically set to 10.
[0093] Sort the distance metrics between the pixels from the previous step and each cluster center, and assign the cluster center with the smallest distance as the new cluster center for that pixel.
[0094] Based on the newly assigned pixels to each cluster center, calculate the average spatial location of these pixels to obtain the new spatial location of the cluster centers.
[0095] The above steps are iterated until the spatial location of the cluster centers no longer changes, thus obtaining the final clustering result.
[0096] The final clustering results obtained from the above steps are processed by reassigning discontinuous or excessively small superpixels to their nearest neighbors, resulting in the final superpixel segmentation result. The advantage of superpixel segmentation is that it divides the image into smaller regions with similar features, reducing computational complexity while preserving the edge information of the target and improving the accuracy of target extraction.
[0097] Step S30: Extract two polarization features, the circular polarization basis correlation coefficient and the reflection symmetry coefficient, from the superpixel units obtained in the previous step.
[0098] It should be noted that in order to extract targets with volume scattering dominated by reflection symmetry from PolSAR images, appropriate polarization features need to be selected, including the reflection symmetry coefficient Q. rs Correlation coefficient ρ with circular polarization base RRLL .
[0099] Spatial averaging is performed on the superpixel units obtained in the previous step to obtain the second-order statistic covariance matrix C. ′ :
[0100]
[0101] The reflection symmetry coefficient Q is calculated using the superpixel covariance matrix. rs The calculation method is as follows:
[0102]
[0103] Where det(·) represents the determinant operation of a matrix, C' represents the covariance matrix, and C 11 Let C' be a 2×2 matrix with its top left corner, and C... 22 The lower right corner is a 1×1 matrix. When reflection symmetry is satisfied and the influence of polarization distortion is small, C' approaches the theoretical value C. ′ t Therefore, the stronger the reflection symmetry, the higher the Q. rs The smaller the value.
[0104] The polarization matrix observed from the same target in different directions can be significantly different. The diversity of target scattering directions contains rich information, which can be used to reveal the polarization characteristics of different targets. By transforming the polarization basis, the reflection symmetry coefficients can be extended to the rotation domain, allowing for classification based on the variation patterns of surface cover under different polarization bases. The POA rotation method can be used to transform the scalar feature Q... rs Expanded into vector Q rs This provides a more accurate and robust target description by leveraging the target's characteristics as a function of the Point of Orientation (POA). Q for different orientation angles θ rs The calculation method is as follows:
[0105]
[0106] C ′ (θ)=U(θ)C ′ U T (θ)
[0107]
[0108] Among them, Q rs (θ) represents Q under different POAs rs By sampling θ at equal intervals within the range [0, 90°), the scalar features of the target can be expanded into a vector Q. rs . Figure 2 The Q values of built-up area targets dominated by non-volume scattering and vegetation targets dominated by volume scattering are shown. rs The feature sequence shown demonstrates that the features can effectively distinguish between these two types of targets.
[0109] For uncalibrated fully polarimetric SAR images, a polarization basis transformation is performed, converting linear polarization to circular polarization, that is, converting horizontal and vertical polarization into left-handed and right-handed circular polarization:
[0110]
[0111] Where R represents right-hand circular polarization, L represents left-hand circular polarization, and i represents an imaginary number.
[0112] Using the circularly polarized basis image obtained in the previous step, the superpixel circularly polarized correlation coefficient ρ is calculated. RRKL Calculation:
[0113]
[0114] The polarization features obtained in the above steps are used to construct feature vectors, thereby achieving an accurate description of superpixels. The advantage of this step is that it improves the feature representation capability of superpixels, providing richer and more accurate feature information for subsequent classification.
[0115] Step S40: Using the pixel proportion threshold method, high-confidence volume scattering dominated and non-volume scattering dominated targets at the pixel level are used to obtain high-confidence volume scattering dominated and non-volume scattering dominated targets in superpixel units. These targets are then fused with the circular polarization basis correlation coefficient and reflection symmetry coefficient features to obtain positive and negative samples of superpixel units. A random forest classifier is used to train the samples to obtain the full image classification result. The high-confidence volume scattering dominated targets in superpixel units are the reference distributed targets that meet the polarization calibration requirements.
[0116] In the specific implementation, the pixel proportion threshold method is used to calculate the proportion of pixel-level high-confidence volume scattering dominant targets within each superpixel unit. A threshold of 50% is set. Superpixel units where the proportion of high-confidence volume scattering dominant targets exceeds this threshold are classified as volume scattering dominant targets. This is then fused with the circular polarization basis correlation coefficient and reflection symmetry coefficient features to obtain positive superpixel unit samples. Similarly, the pixel proportion threshold method is used to calculate the proportion of pixel-level high-confidence non-volume scattering dominant targets within each superpixel unit. A threshold of 50% is set. Superpixel units where the proportion of high-confidence non-volume scattering dominant targets exceeds this threshold are classified as non-volume scattering dominant targets. This is then fused with the circular polarization basis correlation coefficient and reflection symmetry coefficient features to obtain negative superpixel unit samples. A random forest classifier is used to train the samples obtained in the above steps to obtain the full-image classification results. The high-confidence volume scattering dominant targets in the superpixel units are the reference distributed targets that meet the polarization calibration requirements. This method has good generalization ability, can achieve automatic classification of the entire image, and effectively extracts reference distributed targets that meet the polarization calibration requirements.
[0117] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.
Claims
1. A method for extracting distributed targets using SAR polarization calibration references, characterized in that, include: Using polarization distortion parameters and polarization orientation angles extracted from uncalibrated fully polarimetric SAR images, a threshold-based pixel classification method was used to obtain pixel-level high-confidence volume scattering-dominant targets and non-volume scattering-dominant targets. Superpixel segmentation is performed on uncalibrated fully polarimetric SAR images to obtain superpixel units; Two polarization features, circular polarization basis correlation coefficient and reflection symmetry coefficient, are extracted from the superpixel unit; The pixel ratio threshold method is used to process the high-confidence volume scattering dominated and non-volume scattering dominated targets at the pixel level, and the high-confidence volume scattering dominated and non-volume scattering dominated targets of superpixel units are obtained. The superpixel unit positive and negative samples are obtained by combining the two polarization features, the circular polarization basis correlation coefficient and the reflection symmetry coefficient. A random forest classifier is used to train the positive and negative samples of the superpixel unit to obtain the full image classification result. The superpixel unit volume scattering dominant target in the full image classification result is the reference distributed target that meets the polarization calibration requirements. The thresholding method for pixel classification, using polarization distortion parameters and polarization orientation angles extracted from uncalibrated fully polarimetric SAR images, yields pixel-level high-confidence volume scattering-dominant targets and non-volume scattering-dominant targets, including: The reciprocity angle is calculated using uncalibrated fully polarimetric SAR images. The reciprocity angle is then compared with a reciprocity judgment threshold, and targets corresponding to reciprocity angles exceeding the threshold are removed. The Quegan polarization calibration algorithm was used to process the uncalibrated fully polarimetric SAR image after the removal process to obtain crosstalk distortion parameters and channel imbalance distortion parameters. Set the physical meaning thresholds for the phase standard deviation of the channel imbalance distortion parameter and the crosstalk distortion parameter. Determine whether the phase standard deviation of the channel imbalance distortion parameter and the crosstalk distortion parameter of each pixel exceed the two set thresholds. If either one exceeds the threshold, the classification result of the pixel is a high-confidence non-volume scattering dominant target. For each pixel in the uncalibrated fully polarimetric SAR image after removal processing, the polarization orientation angle is calculated. A physical meaning threshold for the standard deviation of the polarization orientation angle is set. The classification result of the pixel corresponding to the standard deviation of the polarization orientation angle exceeding the physical meaning threshold of the standard deviation of the polarization orientation angle is the pixel-level high confidence volume scattering dominant target.
2. The SAR polarization calibration reference distributed target extraction method according to claim 1, characterized in that, The physical threshold for the phase standard deviation of the channel imbalance distortion parameter is set to 140°, and the physical threshold for the crosstalk distortion parameter is set to -6dB.
3. The SAR polarization calibration reference distributed target extraction method according to claim 1, characterized in that, The physical meaning threshold for the standard deviation of the polarization orientation angle is set to 29°.
4. The SAR polarization calibration reference distributed target extraction method according to claim 1, characterized in that, The calculation of the polarization orientation angle of the uncalibrated fully polarimetric SAR image after the removal process is specifically as follows: in, This indicates a spatial averaging operation. Indicates conjugate operation; H Indicates horizontal polarization. V Indicates vertical polarization. Represents the elements of the observed polarization scattering matrix.
5. The SAR polarization calibration reference distributed target extraction method according to claim 4, characterized in that, The uncalibrated fully polarimetric SAR image is segmented into superpixel units using a linear iterative clustering algorithm, including: 1) Set the number of superpixels and seed points, and distribute seed points evenly within the uncalibrated fully polarimetric SAR image; 2) Process the uncalibrated fully polarimetric SAR image to obtain the Pauli pseudocolor image of the uncalibrated fully polarimetric SAR image. Calculate the gradient value of all pixels in the Pauli pseudocolor image within the 3×3 neighborhood of each seed point. Move the seed point to the place with the smallest gradient in the neighborhood to obtain the reselected initial seed point, i.e. the initial cluster center. 3) Use the initial cluster centers to search for pixels within a 2S×2S neighborhood. For each searched pixel, calculate its color distance to the initial cluster centers within a 2S×2S neighborhood. Spatial distance And calculate the distance metric. S represents the distance between two adjacent seed points; 4) Sort the distance metrics between pixels and each initial cluster center, and select the smallest distance metric. The corresponding initial cluster center is defined as the new cluster center for that pixel; 5) Calculate the average spatial location of the pixels assigned to each new cluster center to obtain the spatial location of the new cluster center; 6) Iterate through steps 3)-5) until the spatial location of the cluster centers no longer changes, and obtain the final clustering result; define the pixel of each cluster center as a superpixel unit.
6. The SAR polarization calibration reference distributed target extraction method according to claim 5, characterized in that, The setting of the number of superpixels and the number of seed points includes: Setting the number of superpixels = Setting the number of seed points = .
7. The SAR polarization calibration reference distributed target extraction method according to claim 5, characterized in that, The distance metric ,in, This indicates the brightness range of the Lab color model.
8. The SAR polarization calibration reference distributed target extraction method according to claim 5, characterized in that, The extraction of two polarization features—circular polarization basis correlation coefficient and reflection symmetry coefficient—from superpixel units includes: Spatial averaging of superpixel units yields the second-order statistical covariance matrix; The reflection symmetry coefficient is calculated using the second-order statistic covariance matrix. A polarization basis transformation is performed on the uncalibrated fully polarimetric SAR image to convert it from a linear polarization basis to a circular polarization basis, that is, to convert the horizontal and vertical polarizations into left-hand circular polarizations and right-hand circular polarizations, thus obtaining a circular polarization basis image. The correlation coefficient of the circular polarization basis is calculated using the obtained circular polarization basis image.
9. A SAR polarization calibration reference distributed target extraction method according to claim 8, characterized in that, Reflection symmetry coefficient The calculation method is as follows: in, Represents the determinant operation of a matrix. Represents the covariance matrix of second-order statistics. for The top left corner 2×2 matrix, for The bottom right corner is a 1×1 matrix.
10. A SAR polarization calibration reference distributed target extraction method according to claim 9, characterized in that, The correlation coefficient of the circular polarization basis is calculated using the obtained circular polarization basis image. ,include: Where R represents right-hand circular polarization and L represents left-hand circular polarization; 11. A SAR polarization calibration reference distributed target extraction method according to any one of claims 1-10, characterized in that, The pixel proportion threshold method is used to process pixel-level high-confidence volume scattering-dominated and non-volume scattering-dominated targets to obtain superpixel units of high-confidence volume scattering-dominated and non-volume scattering-dominated targets. These targets are then combined with two polarization features—the correlation coefficient between the non-volume scattering-dominated target and the circular polarization basis, and the reflection symmetry coefficient—to obtain positive and negative samples of the superpixel units, including: Using the pixel proportion threshold method, the proportion of high-confidence volume scattering dominant targets in each superpixel unit is calculated. A proportion threshold is set, and superpixel units with a proportion of high-confidence volume scattering dominant targets exceeding the proportion threshold are identified as volume scattering dominant targets. The positive samples of superpixel units are obtained by combining the characteristics of the circular polarization basis correlation coefficient and reflection symmetry coefficient of the superpixel unit. Using the pixel proportion threshold method, the proportion of high-confidence non-volume scattering dominant targets in each superpixel unit is calculated. A proportion threshold is set, and superpixel units with a proportion of high-confidence non-volume scattering dominant targets exceeding the proportion threshold are identified as non-volume scattering dominant targets. The results are fused with the circular polarization basis correlation coefficient and reflection symmetry coefficient features to obtain the negative samples of the superpixel unit.
12. The SAR polarization calibration reference distributed target extraction method according to claim 11, characterized in that, The ratio threshold is set to 50%.