Non-rotation region selection method based on full-polarization SAR channel imbalance calibration

Abstract

The invention relates to a non-rotation region selection method based on full-polarization SAR channel imbalance calibration. The method comprises the following steps: reading full-polarization SAR image data and carrying out multi-view extraction processing to obtain distorted back scattering data; the back scattering data with distortion are expressed in a polarization distortion model form for describing the distortion condition, and a covariance matrix and a coherence matrix for describing the distributed target are obtained through calculation; according to the covariance matrix and the coherence matrix, a scattering entropy H and a scattering angle # imgabs0 # describing target scattering characteristics are obtained through calculation, multi-dimensional scattering characteristic representation of distortion data is constructed based on the scattering entropy H and the scattering angle # imgabs1 #, and a scattering characteristic space of the distortion data is constructed; a scattering feature space of distortion data is optimized, and the influence of left-right movement of points on a # imgabs2 # plane in the feature space on extraction of a non-rotation region when a channel is unbalanced, the phase is unchanged and the amplitude is changed is eliminated; and a threshold adjustment mechanism and an iterative segmentation method are used to identify an area meeting a non-rotation condition in the feature space.

Description

Technical Field

[0001] The present invention relates to the field of radar detection, in particular to the field of polarization calibration of a polarization synthetic aperture radar (SAR). Background Art

[0002] Fully polarimetric SAR is an advanced remote sensing technology that can provide richer target information than traditional single-polarization SAR by capturing and analyzing radar signals in different polarization states. This technology has important applications in a variety of fields, including environmental monitoring, land use planning, disaster assessment, etc. For example, in forest monitoring, fully polarimetric SAR can distinguish different types of trees and surface materials, helping scientists and policymakers better understand changes in ecosystems. Due to the high resolution and multi-dimensional information it provides, fully polarimetric SAR has become an indispensable part of remote sensing technology.

[0003] However, in order to ensure the quality of polarimetric SAR images and the reliability of data, its measurement channels must be accurately calibrated. Channel imbalance can cause systematic errors in the acquisition of data by the full polarimetric SAR system, thus affecting the clarity and accuracy of the final image. Channel imbalance is mainly manifested in the inconsistency of the gain, phase and polarization characteristics of the receiving and transmitting channels. These problems may distort the reflection information of the target object in different polarization states. In order to overcome these problems, effective calibration methods are needed to adjust and compensate for these imbalances to ensure that the full polarimetric SAR images truly reflect the characteristics of the target.

[0004] In the full-polarization SAR channel imbalance calibration algorithm based on distributed targets, one of the key steps is to prepare the distributed targets with the scattering characteristics used in the distorted image. Existing technical solutions mainly focus on setting thresholds for polarization parameters and extracting distributed targets based on deep learning methods. However, both of the above solutions have problems when processing a large amount of co-track data: 1) The first solution does not consider the impact of channel imbalance on polarization parameters, and sets different thresholds for polarization parameters in each image, so such algorithms cannot be used in large amounts of data; 2) The second solution takes a lot of time to train the model in deep learning. Therefore, there is an urgent need for a region selection algorithm that can be used in multi-track distorted data to improve the accuracy of target extraction when processing large amounts of data. Summary of the invention

[0005] The technical problem solved by the present invention is to overcome the shortcomings of the prior art and provide a method for selecting an irrotational region based on full-polarization SAR channel imbalance calibration, and to use multi-dimensional scattering features to accurately extract the irrotational region for calibrating channel imbalance while ensuring that antenna isolation meets the index.

[0006] The solution to the technical problem of the present invention is: a method for selecting an irrotational region based on unbalanced calibration of a fully polarized SAR channel, the method comprising the following steps:

[0007] Read the fully polarized SAR image data and perform multi-view extraction processing to obtain distorted backscatter data;

[0008] The distorted backscattered data is expressed as a polarization distortion model that describes the distortion, and the covariance matrix and coherence matrix describing the distributed targets are calculated;

[0009] According to the covariance matrix and coherence matrix, the scattering entropy H and scattering angle describing the target scattering characteristics are calculated. Based on the scattering entropy H and scattering angle Constructing a multi-dimensional scattering feature representation of the distorted data and constructing a scattering feature space of the distorted data;

[0010] Optimize the scattering feature space of distorted data to eliminate channel imbalance. When the phase remains unchanged and the amplitude changes, the feature space The effect of left-right movement of a point on a plane on the extraction of irrotational regions;

[0011] Using a threshold adjustment mechanism and an iterative segmentation method, regions that meet the irrotational condition are identified in the feature space.

[0012] Preferably, the number of multi-views ML in the multi-view extraction process is N for:

[0013]

[0014] Wherein, [·] represents the rounding operator, Azimuth_spacing is the azimuth resolution of the fully polarized SAR image data, and Range_spacing is the range resolution of the fully polarized SAR image data.

[0015] Preferably, the multi-dimensional scattering characteristics of the distorted data are expressed as scattering entropy H and scattering angle data in different formats.

[0016] Preferably, the scattering entropy H and the scattering angle is defined as:

[0017]

[0018]

[0019]

[0020] Among them, λ iis the eigenvalue of the covariance matrix, α i Represents the arccosine value of the first parameter of the coherence matrix eigenvector. The value range of i is 1-4.

[0021] Preferably, the polarization distortion model describing the distortion situation is:

[0022] [M]=[F][S]

[0023] Right now:

[0024]

[0025] Among them, S pq and M pq represents the undistorted and distorted backscatter coefficients, p = h, q = v; [S] and [M] represent the undistorted and distorted backscatter matrices; k and χ represent the co-polarization and cross-polarization channel imbalances; [F] represents the channel imbalance matrix.

[0026] Preferably, the covariance matrix is ​​obtained by the following method:

[0027]

[0028] in, <·> represents the spatial average operator of the matrix, Represents the conjugate transpose operator of a matrix.

[0029] Preferably, the coherence matrix <t>for:

[0030]

[0031]

[0032] Preferably, the method for optimizing the scattering feature space of the distorted data is:

[0033] Get The maximum value H in the region of the plane containing only Bragg scattering max ;

[0034] Set the scattering entropy H to a value range of 0 to H max interval;

[0035] The multi-dimensional scattering feature representation of the distorted data is kept unchanged, and the optimized scattering feature space of the distorted data is obtained based on the updated value range of the scattering entropy H.

[0036] Preferably, the specific steps of using the threshold adjustment mechanism and the iterative segmentation method are as follows:

[0037] S1. Add a preset channel imbalance phase to the distorted backscatter data so that the points in the irrotational region move toward the undistorted state on the plane. The preset channel imbalance phase is ±45° / 2. n ,n is the number of iterations, initialized to 0;

[0038] S2. According to a preset threshold, extracting the irrotational region from the scattering feature space of the optimized distortion data and calculating the number of irrotational points;

[0039] S3, add 1 to n to update n, and repeat steps S1-S2;

[0040] S4. If the points in the irrotational region have moved to an undistorted state on the plane or the number of iterations has reached a preset cutoff number, the process ends; otherwise, step S3 is repeated.

[0041] Preferably, the criterion for determining whether a point in the irrotational region has moved to an undistorted state on the plane is:

[0042] The number of irrotational points in the nth iteration is compared with the number of irrotational points in the n-1th iteration. If the two are equal, it means that the points in the irrotational region have moved to an undistorted state on the plane; otherwise, it means that the irrotational points are still moving on the plane, which means that phase still needs to be added for further movement.

[0043] The beneficial effects of the present invention compared with the prior art are:

[0044] (1) The present invention accurately identifies the irrotational region by extracting multi-dimensional scattering features and optimizing the feature space, thereby significantly improving the calibration accuracy of polarization channel imbalance.

[0045] (2) The present invention uses threshold adjustment and iterative segmentation technology to automatically identify areas that meet the irrotational condition, thereby improving processing efficiency and consistency.

[0046] (3) The present invention enhances the adaptability of the method and simplifies the data processing process while ensuring antenna isolation. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 The present invention is a flowchart of a method for selecting an irrotational region based on full-polarization SAR channel imbalance calibration according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the protection scope of the present invention.

[0049] According to one embodiment of the present invention, the importance analysis of the irrotational region selection method based on the full polarization SAR channel unbalanced calibration is as follows:

[0050] The importance of the present invention lies in the following two aspects:

[0051] 1. In-depth analysis of multi-dimensional scattering characteristics: This invention analyzes the mathematical model of polarization distortion, obtains the multi-dimensional scattering characteristic representation of channel imbalance distortion data and constructs the feature space, which can more comprehensively analyze and understand the complex scattering characteristics in the data. This helps to extract more useful information from the data and improve the detection and recognition capabilities of targets in complex environments.

[0052] 2. Optimize feature space and effectively identify target areas: The present invention optimizes and limits feature space, especially enhances the distinguishability of irrotational areas in distorted data, and combines threshold adjustment mechanism and iterative segmentation technology to accurately identify target areas that meet irrotational conditions in feature space. This optimization and identification method effectively improves the distinguishability of feature space, enabling the system to more accurately extract and identify targets in distorted images, thereby significantly improving the target detection and classification capabilities of full-polarization SAR systems in practical applications.

[0053] In summary, the present invention significantly improves the detection accuracy and reliability of the full-polarimetric SAR system by systematically processing the distortion problem in the full-polarimetric SAR data, deeply analyzing the scattering characteristics, optimizing the feature space, and adopting advanced region recognition technology. It has broad application prospects and important practical significance.

[0054] Furthermore, the process of selecting the irrotational region based on the unbalanced calibration of the full polarization SAR channel is as follows: Figure 1 As shown, specifically:

[0055] Firstly, the auxiliary data of the distorted image is used to obtain the size and resolution of the data. Then, the fully polarimetric SAR image data is read and multi-view extraction is performed to reduce the data complexity. Finally, a polarimetric distortion model is constructed for the polarimetric data with channel imbalance. Then, the obtained distorted data is used to construct a variety of scattering features based on polarization entropy and scattering angle to construct a feature space. In the multidimensional scattering feature space, a suitable feature subspace is determined by optimization to effectively remove some non-target areas, thereby enhancing the distinguishability of the target signal. The threshold adjustment mechanism is used to ensure the segmentation standard. Through the iterative segmentation method, the candidate areas in the feature space are gradually refined, and the non-target areas are eliminated until convergence to the optimal solution, ensuring that the target area can be accurately distinguished from other areas in different scenarios, thereby achieving high-precision irrotational area recognition.

[0056] Based on the above analysis, according to one embodiment of the present invention, a method for evaluating the performance of a fully polarized SAR system based on a calibrator group is proposed. The method comprises the following steps:

[0057] (1) Read the fully polarized SAR image data and perform multi-view extraction processing to obtain distorted backscatter data;

[0058] First, the size and resolution information of the fully polarimetric SAR image data is obtained using the auxiliary data of the distorted image. This step involves parsing the original image data to determine the actual size of the image (such as the number of points in azimuth and range) and its resolution (including azimuth resolution Azimuth_spacing and range resolution Range_spacing). Next, read the image data and perform multi-view extraction processing on it to reduce the size of the data. N It can be expressed as

[0059]

[0060] Wherein, [·] represents the rounding operator, Azimuth_spacing is the azimuth resolution of the fully polarized SAR image data, and Range_spacing is the range resolution of the fully polarized SAR image data.

[0061] (2) Representing the distorted backscatter data in the form of a polarization distortion model that describes the distortion, and calculating the covariance matrix and coherence matrix that describe the distributed target;

[0062] In the present invention, it is necessary to consider the distortion differences between different channels and to describe these distortions by establishing a mathematical model so as to better perform data correction and analysis. The mathematical model can be expressed as

[0063] [M]=[F][S]

[0064] Right now:

[0065]

[0066] Among them, S pq and M pq represents the undistorted and distorted backscatter coefficients, p = h, q = v; [S] and [M] represent the undistorted and distorted backscatter matrices; k and χ represent the co-polarization and cross-polarization channel imbalances; [F] represents the channel imbalance matrix.

[0067] When using distributed targets for polarization calibration, the covariance matrix is ​​often used to describe the characteristics of the distributed targets. The covariance matrix is ​​obtained by the following method:

[0068]

[0069] in, <·> represents the spatial average operator of the matrix, Represents the conjugate transpose operator of a matrix.

[0070] In addition to the covariance matrix, you can also use the coherence matrix <t>Describe the characteristics of distributed targets, coherence matrix <t>The relationship with the covariance matrix is:

[0071]

[0072]

[0073] The use of the above model can more directly analyze and obtain the irrotational region in the data, thereby correcting the polarization distortion in the data and improving the quality and reliability of the final data.

[0074] (3) According to the covariance matrix and coherence matrix, the scattering entropy H and scattering angle describing the target scattering characteristics are calculated. Based on the scattering entropy H and scattering angle Constructing a multi-dimensional scattering feature representation of the distorted data and constructing a scattering feature space of the distorted data;

[0075] Before selecting the irrotational region, it is necessary to obtain a multi-dimensional scattering feature representation of the distorted data and construct a scattering feature space of the distorted data, and use mathematical means to describe the relationship between the scattering features, and then extract the irrotational region. Common natural objects that exhibit irrotation include bare soil, deserts, and forests with uniform scattering. Polarization decomposition is a process that converts multi-polarization SAR data into a set of parameters with clear physical meanings. This method can be used to extract the scattering characteristics and fine structure information of the target from the multi-polarization data. In polarization decomposition, the features commonly used to describe the above-mentioned objects include Freeman two / three-component decomposition, Cloude-Pottier decomposition (also known as decomposition), Yamaguchi four-component decomposition, etc. Among them, The decomposition includes two parameters: scattering entropy H and scattering angle Compared with the above two / three components and four components, H and The natural objects can be divided into eight categories more finely, and the characteristics of natural objects can be described more fully. is defined as:

[0076]

[0077]

[0078]

[0079] Among them, λ i is the eigenvalue of the covariance matrix, α i Represents the arccosine value of the first parameter of the coherence matrix eigenvector. The value range of i is 1-4.

[0080] In the typical irrotational region such as the bare soil Bragg scattering region, H < 0.5, α < 42.5°. In addition, due to the All have clear mathematical meanings and have upper and lower limits on their values. Some mathematical variants also have mathematical meanings, including <N 11 >、(1-H)α、 Based on the above analysis, in the present invention, the multi-dimensional scattering characteristics of the distorted data are represented by the scattering entropy H and the scattering angle The data is transformed into H and As the "center", its mathematical transformation is the "vector direction" to construct the scattering feature space.

[0081] (4) Optimize the scattering feature space of the distorted data to eliminate channel imbalance. When the phase remains unchanged and the amplitude changes, the feature space The effect of left-right movement of a point on a plane on the extraction of irrotational regions;

[0082] In H and In the feature space formed by The plane can intuitively obtain the scattering characteristics of each target. Therefore, the impact of channel imbalance on the scattering characteristics of ground objects can be seen in It can be analyzed intuitively on the plane. Considering that k and χ are both complex numbers, including amplitude and phase, these two types of influence can be considered separately, and then the feature space can be optimized and restricted. When the amplitude of the two variables remains unchanged and the phase changes, the points on the entire plane only change up and down. Furthermore, the change period changes with It changes in a 180° cycle. When the phase of the two channels is unbalanced and the amplitude changes, the point on the plane moves left and right. When the amplitude range of the channel imbalance is limited, H has a maximum value H only in the area containing Bragg scattering. max Therefore, when selecting the irrotational region, set H from 0 to H max This interval can effectively avoid the extraction effect caused by the left and right movement of points on the plane.

[0083] The method for optimizing the scattering feature space of distorted data is:

[0084] Get The maximum value H in the region of the plane containing only Bragg scattering max ;

[0085] Set the scattering entropy H to a value range of 0 to H max interval;

[0086] The multi-dimensional scattering feature representation of the distorted data is kept unchanged, and the optimized scattering feature space of the distorted data is obtained based on the updated value range of the scattering entropy H.

[0087] (5) Using the threshold adjustment mechanism and iterative segmentation method, the regions that meet the irrotational condition are identified in the feature space.

[0088] In this step, the main problem is the effect of the up and down movement of points on the plane on the selected area. Since the up and down movement of points on the plane is mainly caused by the distorted phase, an additional channel imbalance phase can be added to the distorted data to move the points in the irrotational area to an undistorted state on the plane. It should be noted that adding the phase moves the points on the plane as a whole, rather than moving a single area.

[0089] Considering that the point change period in the plane is 180°, and the change is monotonically increasing and monotonically decreasing, in the actual process, we can start from ±45° / 2 0 Start the iteration, then set the threshold, calculate the number of irrotational points fixed in a certain area, and then add ±45° / 2 1 , calculate the number of irrotational points in this stage and compare it with the number of irrotational points in the previous stage. If they are equal, it means that the points in the irrotational region have moved to an undistorted state on the plane; if there is a change, that is, the irrotational points are still moving on the plane, then it means that phase still needs to be added to make it move further until it remains unchanged.

[0090] Therefore, the specific steps of using the threshold adjustment mechanism and iterative segmentation method are as follows:

[0091] S1. Add a preset channel imbalance phase to the distorted backscatter data so that the points in the irrotational region move toward the undistorted state on the plane. The preset channel imbalance phase is ±45° / 2. n ,n is the number of iterations, initialized to 0;

[0092] S2. According to a preset threshold, extracting the irrotational region from the scattering feature space of the optimized distortion data and calculating the number of irrotational points;

[0093] S3, add 1 to n to update n, and repeat steps S1-S2;

[0094] S4. If the points in the irrotational region have moved to an undistorted state on the plane or the number of iterations has reached a preset cutoff number, the process ends; otherwise, step S3 is repeated.

[0095] The criterion for determining whether a point in the irrotational region has moved to an undistorted state on the plane is:

[0096] The number of irrotational points in the nth iteration is compared with the number of irrotational points in the n-1th iteration. If the two are equal, it means that the points in the irrotational region have moved to an undistorted state on the plane; otherwise, it means that the irrotational points are still moving on the plane, which means that phase still needs to be added for further movement.

[0097] It should be noted that some images may have fewer irrotational points, resulting in larger changes. Therefore, a cutoff number is set in the present invention so that even if the above situation is not reached, iteration will be stopped to release computing resources.

[0098] Although the present invention has been disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications to the technical solution of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope of the technical solution of the present invention.< / t> < / t> < / t>

Claims

1. A method for selecting irrotational regions based on unbalanced calibration of fully polarized SAR channels, characterized by The steps include: Read the fully polarized SAR image data and perform multi-view extraction processing to obtain distorted backscatter data; The distorted backscattered data is expressed as a polarization distortion model that describes the distortion, and the covariance matrix and coherence matrix describing the distributed targets are calculated; According to the covariance matrix and coherence matrix, the scattering entropy H and scattering angle describing the target scattering characteristics are calculated. Based on the scattering entropy H and scattering angle Constructing a multi-dimensional scattering feature representation of the distorted data and constructing a scattering feature space of the distorted data; Optimize the scattering feature space of distorted data to eliminate channel imbalance. When the phase remains unchanged and the amplitude changes, the feature space The effect of left-right movement of a point on a plane on the extraction of irrotational regions; Using a threshold adjustment mechanism and an iterative segmentation method, regions that meet the irrotational condition are identified in the feature space.

2. The method for selecting the irrotational region based on the full polarization SAR channel unbalanced calibration according to claim 1, characterized in that: The number of multi-views ML in the multi-view extraction process N for: Wherein, [·] represents the rounding operator, Azimuth_spacing is the azimuth resolution of the fully polarized SAR image data, and Range_spacing is the range resolution of the fully polarized SAR image data.

3. The method for selecting the irrotational region based on the full polarization SAR channel unbalanced calibration according to claim 1, characterized in that: The multidimensional scattering characteristics of the distorted data are expressed as scattering entropy H and scattering angle data in different formats.

4. The method for selecting the irrotational region based on the full polarization SAR channel unbalanced calibration according to claim 1, characterized in that: The scattering entropy H and the scattering angle is defined as: Among them, λ i is the eigenvalue of the covariance matrix, α i Represents the arccosine value of the first parameter of the coherence matrix eigenvector. The value range of i is 1-4.

5. The method for selecting the irrotational region based on the full polarization SAR channel unbalanced calibration according to claim 1, characterized in that: The polarization distortion model describing the distortion situation is: [M]=[F][S] Right now: Among them, S pq and M pq represents the undistorted and distorted backscatter coefficients, p = h, q = v; [S] and [M] represent the undistorted and distorted backscatter matrices; k and χ represent the co-polarization and cross-polarization channel imbalances; [F] represents the channel imbalance matrix.

6. The method for selecting irrotational regions based on full polarization SAR channel unbalanced calibration according to claim 5, characterized in that: The covariance matrix is ​​obtained as follows: in, <·> represents the spatial average operator of the matrix, Represents the conjugate transpose operator of a matrix.

7. The method for selecting irrotational regions based on full polarization SAR channel unbalanced calibration according to claim 6, characterized in that: Coherence Matrix < T > for:

8. The method for selecting irrotational regions based on full polarization SAR channel unbalanced calibration according to claim 1, characterized in that: The method for optimizing the scattering feature space of distorted data is: Get The maximum value H in the region of the plane containing only Bragg scattering max ; Set the scattering entropy H to a value range of 0 to H max interval; The multi-dimensional scattering feature representation of the distorted data is kept unchanged, and the optimized scattering feature space of the distorted data is obtained based on the updated value range of the scattering entropy H.

9. The method for selecting irrotational regions based on full polarization SAR channel unbalanced calibration according to claim 1, characterized in that: The specific steps of using the threshold adjustment mechanism and iterative segmentation method are as follows: S1. Add a preset channel imbalance phase to the distorted backscatter data so that the points in the irrotational region move toward the undistorted state on the plane. The preset channel imbalance phase is ±45° / 2. n ,n is the number of iterations, initialized to 0; S2. According to a preset threshold, extracting the irrotational region from the scattering feature space of the optimized distortion data and calculating the number of irrotational points; S3, add 1 to n to update n, and repeat steps S1-S2; S4. If the points in the irrotational region have moved to an undistorted state on the plane or the number of iterations has reached a preset cutoff number, the process ends; otherwise, step S3 is repeated.

10. The method for selecting irrotational regions based on full polarization SAR channel unbalanced calibration according to claim 9, characterized in that: The criterion for determining whether a point in the irrotational region has moved to an undistorted state on the plane is: The number of irrotational points in the nth iteration is compared with the number of irrotational points in the n-1th iteration. If the two are equal, it means that the points in the irrotational region have moved to an undistorted state on the plane; otherwise, it means that the irrotational points are still moving on the plane, which means that phase still needs to be added for further movement.

Images