SAR polarimetric measurement method and system based on polarimetric autocorrelation matrix

Abstract

The invention belongs to the field of radar monitoring technology, and discloses a SAR polarization measurement method and system based on a polarization autocorrelation matrix. The polarization autocorrelation matrix can be used to capture hidden Doppler information in the frequency domain, and a high-order statistical method can be used to encode it into a phase. The matrix can conveniently extract and analyze polarization and Doppler information and features between polarization channels from polarization SAR (PolSAR) data through matrix analysis. The invention proposes that the polarization autocorrelation matrix can be decomposed into a Doppler matrix and a high-order covariance matrix. The amplitude and phase of these matrices can characterize the Doppler frequency shift and the variance of the backscatter coefficient between different polarizations. The invention shows that the Doppler frequency shift is not only affected by the radial velocity of the target, but also by the radar polarization mode. New descriptors and new parameters can detect artificial targets and sea ice, reduce the ambiguity caused by moving targets, and indicate the motion state of ship targets and ocean currents.

Description

Technical Field

[0001] The invention belongs to the technical field of radar monitoring, and in particular relates to a SAR polarization measurement method and system based on a polarization autocorrelation matrix. Background Art

[0002] Compared with traditional SAR systems, Polarimetric Synthetic Aperture Radar (PolSAR) not only provides a unique opportunity to observe and monitor the Earth's surface from space day and night in any weather conditions, but also more comprehensively describes the polarization scattering mechanism of the target. With the advent of radar polarimetric technology, its application in remote sensing has increased significantly. Compared with single-polarization and dual-polarization SAR sensors, PolSAR data contains richer echo information, which greatly promotes the interpretation, modeling and application of target scattering mechanisms. At present, PolSAR is considered to be the most effective tool for target feature recognition and is widely used in land cover classification, sea ice type classification, ice thickness inversion, detection and identification of artificial targets, oil pollution detection, disaster assessment, marine environment monitoring, and detection of glaciers and icebergs.

[0003] The most widely used method for PolSAR data analysis is the target decomposition (TD) technique. TD technology aims to provide a physical interpretation of the scene by representing radar echoes as a combination of echoes of simpler (canonical) scattering elements. TD techniques are gradually divided into two categories: coherent target decomposition (CTD) and incoherent target decomposition (ICTD). CTD technology is developed based on the Sinclair scattering matrix [S]. Typical CTD methods include Pauli decomposition, Krogager decomposition, Cameron decomposition, Huynen-Euler decomposition and symmetric scattering characterization methods. They regard [S] as a coherent superposition of pure scatterers and are therefore very suitable for analyzing deterministic (artificial) scatterers. The use of CTD for ship detection and identification has therefore received a lot of attention. For example, the sub-view analysis method of the joint spectrum, the polarization scattering characterization and contrast enhancement method, and the ship classification method disclosed in the prior art are all such techniques.

[0004] However, the characteristics of distributed scatterers show significant changes in the observation angle and polarization direction. To solve this problem, the ICTD method specifically for distributed scatterers is proposed. Within the ICTD framework, the characteristic description of distributed scatterers requires the use of second-order polarization descriptors, such as the covariance matrix [C] and the coherence matrix [T].

[0005] In the past, many incoherent target decomposition (ICTD) methods have been proposed, which can be divided into two subcategories. The first category is model-based decomposition. The concept of model-based decomposition was first introduced by Freeman's three-component decomposition method. In addition, the prior art proposes a four-component decomposition method. In recent years, six-component and seven-component decomposition algorithms have been proposed to represent more and more detailed scattering types. The other category is eigenvector or eigenvalue-based decomposition, which performs eigenvalue decomposition on the coherence matrix [T] of PolSAR data and uses eigenvalues ​​and eigenvectors to analyze the scattering process. A typical algorithm closely related to it is Cloude decomposition, and the widely recognized polarization parameters such as entropy, anisotropy and α can be derived from this decomposition.

[0006] In summary, the scattering matrix [S], covariance matrix [C] and coherence matrix [T] are the most important target polarization descriptors in the current methods used to explain the target scattering mechanism. They play an important role as key parameters in polarization data analysis, providing qualitative and quantitative information about the scattering behavior. However, these matrices have the following problems:

[0007] (1) Neither the first-order nor the second-order polarization matrices have available methods to obtain relative frequency or Doppler information related to the target polarization. Doppler, especially the Doppler shift (also called the Doppler center frequency, DC), is considered a reliable tool for monitoring and investigating the radial velocity of moving targets (wind, waves, currents, ships, and vehicles). In today's applications, Doppler and micro-Doppler signals are widely used to extract ocean current and wind speeds, as well as to detect ships, coastlines, and drones. However, it is usually assumed that the Doppler frequency depends mainly on the radar frequency and the radial velocity of the target, and has little sensitivity to polarization. Therefore, the difference in Doppler frequency between different polarizations is usually ignored in existing matrices (such as [S], [C], or [T]).

[0008] (2) Current methods are unable to characterize the Doppler differences between different polarizations in PolSAR data. Currently, there are two main types of Doppler polarimetric measurement methods: the first is the differential Doppler velocity (DDV) method and its extended algorithm, which are usually used to retrieve meteorological parameters such as clouds and rain; the second is the two-dimensional time-frequency method. However, the DDV method is limited to ground-based meteorological radars and cannot be applied to spaceborne SAR sensors. On the other hand, the two-dimensional time-frequency method can only obtain the non-stationary characteristics of the observed scene at different orientations and angles, and does not have the ability to extract Doppler information between different polarizations.

[0009] Furthermore, in terms of accurately quantifying the effect of radar polarization on Doppler shift or DC:

[0010] (1) Recent studies have confirmed the effect of radar polarization on the Doppler shift or DC. For example, by studying the Doppler frequency caused by sea breeze, it was found that the DC of HH polarization is always higher than that of VV polarization at the same wind speed. A similar phenomenon can also be observed in meteorological remote sensing, where the Doppler velocity of raindrops is greater under horizontal polarization than vertical polarization. This is because the dominant direction of elliptical raindrops is parallel to the horizontal principal axis of the sphere under inertia. Therefore, a higher proportion of energy is contributed to the horizontal polarization. In addition, by studying the micro-motion characteristics of the target, researchers have also observed the polarization sensitivity of the micro-Doppler signal. In their related research, the prior art studied the micro-Doppler characteristics of drones and found that the resonance caused by the rotation of the rotor blades has a great influence on the Doppler frequency. Since the rotor blades are horizontally oriented, the frequency range of the resonance effect observed in HH polarization is generally lower than that of VV polarization.

[0011] (2) If the Doppler characteristics between different polarization channels and the traditional polarization characteristics of the scatterer can be obtained, it will be greatly helpful to interpret the scattering characteristics of ground objects. The prior art shows that the use of polarization residual Doppler frequency (PRDF, i.e., the difference between VV and HH DC) and normalized radar cross section (NRCS) can enhance the ability to invert sea surface wind speed from SAR. PRDF not only has the potential to eliminate the bias in VV and HH Doppler estimates, but also can effectively reduce the ambiguity of wind speed inversion. The use of polarization-merged Doppler images can also enhance target detection and classification. The prior art found that the modified color mapping based on micro-Doppler characteristics and velocity maps can be used for classification problems of different azimuths and micro-UAV types.

[0012] In summary, although polarization decomposition and Doppler techniques have proven their effectiveness in characterizing targets or scatterers, there is a lack of quantitative tools to understand the differences in Doppler characteristics between different polarization channels. Summary of the invention

[0013] In order to overcome the problems existing in the related art, the present invention discloses a SAR polarization measurement method and system based on a polarization autocorrelation matrix.

[0014] The technical solution is as follows: A SAR polarimetric measurement method based on a polarimetric autocorrelation matrix comprises:

[0015] S1, obtain the polarization autocorrelation matrix from the polarization covariance matrix through autocorrelation operation;

[0016] S2, captures the hidden Doppler information in the frequency domain by acquiring the polarization autocorrelation matrix and encodes the Doppler information into the phase using high-order statistical methods;

[0017] S3, based on the polarization autocorrelation matrix, matrix analysis is used to simultaneously extract polarization and Doppler features from PolSAR data.

[0018] In step S1, a polarization autocorrelation matrix is ​​obtained from a polarization covariance matrix through an autocorrelation operation, including:

[0019] The first step is to define the polarization autocorrelation matrix;

[0020] The second step is to obtain the elements of the polarization autocorrelation matrix;

[0021] The third step is to obtain the polarization autocorrelation matrix.

[0022] In the first step, the polarization autocorrelation matrix is ​​defined, including:

[0023] In the PolSAR system, the Sinclair scattering matrix [S] is expressed as:

[0024]

[0025] Where S is the Sinclair scattering matrix, S HH is the scattering coefficient from horizontal polarization to horizontal polarization, S HV is the scattering coefficient from horizontal polarization to vertical polarization, S VH is the scattering coefficient from vertical polarization to horizontal polarization, S VV is the scattering coefficient from vertical polarization to vertical polarization;

[0026] Under the condition of single-station SAR, formula (1) satisfies the reciprocity theorem, that is, S HV =S VH ;

[0027] The covariance matrix C is obtained by calculating the outer product of the scattering vector of the Sinclair scattering matrix [S] and its conjugate transpose;

[0028]

[0029] Where C is the covariance matrix, * represents the conjugate transpose, k is the scattering vector, is the complex conjugate transpose of the scattering vector;

[0030] Calculate the autocorrelation function of each element in the covariance matrix C. The symbol R represents the autocorrelation operation performed on each element in the matrix. The polarization autocorrelation matrix [A] is defined as:

[0031]

[0032] Where A and [A] are polarization autocorrelation matrices, and R() is the autocorrelation operation.

[0033] In the second step, the elements of the polarization autocorrelation matrix are obtained, including:

[0034] By rearranging formula (3), we obtain:

[0035]

[0036] Where Aij represents the element in the polarization autocorrelation matrix, i,j = 1, 2, 3; A 11 ,A 22 ,A 33 is a real number, and the other data are complex numbers; A 12 and A 21 ,A 13 and A 31 ,A 23 and A 32 are complex conjugates of each other;

[0037]

[0038]

[0039]

[0040]

[0041]

[0042]

[0043]

[0044]

[0045]

[0046] Where R() and E() represent autocorrelation operation and expectation, respectively; in PolSAR images, m is the number of rows in the azimuth direction, k is the interval between two samples along the azimuth direction, * represents conjugate transposition, and S HH is the scattering coefficient from horizontal polarization to horizontal polarization, S HV is the scattering coefficient from horizontal polarization to vertical polarization, S VH is the scattering coefficient from vertical polarization to horizontal polarization, S VV is the scattering coefficient from vertical polarization to vertical polarization;

[0047] A 11 and A 13 Derived, obtain the elements of the polarization autocorrelation matrix, A 11 The derivation formula is expressed as:

[0048]

[0049] A 11 It consists of two parts:

[0050]

[0051] in:

[0052]

[0053]

[0054] A 13 Decomposed into the sum of two parts, the expression is:

[0055]

[0056]

[0057] In the formula, COV() represents covariance, A 11 and A 13 are the elements in the polarization autocorrelation matrix; Respectively represent the first element of the polarization autocorrelation matrix and its corresponding Dop term and Cov term;

[0058] in:

[0059]

[0060]

[0061] By reference A 11 and A 13 , establish the expression for all elements in the matrix [A]:

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071] In the formula, is the ensemble average of the backscatter coefficients of the three polarizations, is the standard deviation of the backscatter coefficients for the three polarizations, are the Doppler centroid (DC) of HH, VV and HV polarizations, respectively, yes and The correlation coefficient between and The definition of The same; e is the natural logarithm, j is the imaginary unit, and T is the pulse repetition period.

[0072] In the third step, the polarization autocorrelation matrix is ​​obtained, including:

[0073] According to formula (8), each element of the polarization autocorrelation matrix is ​​divided into a Dop term and a Cov term; therefore, the polarization autocorrelation matrix is ​​written as the Doppler correlation matrix [A Dop ] and the covariance correlation matrix [A Cov ], as follows:

[0074]

[0075] in,

[0076]

[0077]

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088] in, are the elements in the Doppler correlation matrix, are the elements of the covariance correlation matrix, i,j=1,2,3;

[0089] Each element of the polarization autocorrelation matrix is ​​divided into a Dop term and a Cov term; therefore, the polarization autocorrelation matrix is ​​written as the Doppler correlation matrix [A Dop ] and the covariance correlation matrix [A Cov ] and; use and The degree of deviation of Doppler shift and spectrum broadening represents the phase value.

[0090] In step S2, the hidden Doppler information is captured in the frequency domain by the acquired polarization autocorrelation matrix, and the Doppler information is encoded into the phase using a high-order statistical method, including:

[0091] Step 1: Generate a 3×3 matrix from a 2×2 Sinclair scattering matrix S

[0092] Step 2: Calculate in a sliding window and The set average of; set k to 1; set the sliding window size to a common value;

[0093] Step 3: In the sliding window of the same size, use formula (3) to Apply an autocorrelation operation to each element in to obtain the polarization autocorrelation matrix [A];

[0094] Step 4: Calculate the Doppler term [A] of the polarization autocorrelation matrix [A] by formula (10). Dop ];

[0095] Step 5: Calculate the high-order covariance term [A] of the polarization autocorrelation matrix [A] by formula (11): Cov ].

[0096] In step 4, the Doppler correlation matrix [A Dop ] has a rank of 1 and is represented as the product of two one-dimensional vectors:

[0097]

[0098] Among them, k Dop is defined as:

[0099]

[0100] Covariance correlation matrix [A Cov] by subtracting the Doppler correlation matrix [A] from the polarization autocorrelation matrix [A] Dop ] is calculated, and the expression is:

[0101] A Cov =[A Cov ]=AA Dop =[A]-[A Dop ](11).

[0102] In step 4, the Doppler correlation matrix [A Dop ], A Dop The phase of the off-diagonal elements in is controlled by the DC frequencies of the HH, VV, and HV polarizations. The DC frequencies of HH, VV, and HV / VH are: and

[0103] The Doppler shift or DC frequency of the radar echo is affected by multiple contributing sources, including platform motion, radar attitude, and moving targets on the surface, and is expressed as:

[0104] f DC =f geom +f elec +f phys +Δf (12);

[0105] Among them, f geom is the Doppler term generated by satellite geometry and attitude, f elec is the Doppler term generated by antenna pointing; Δf is the Doppler residual caused by non-geophysical terms and unknown deviations; f phys is the Doppler term generated by the ground moving target, f phys Commonly used to calculate the radial Doppler velocity v of ground moving targets d ;

[0106]

[0107] Among them, k e is the electromagnetic wave number of the radar, θ i is the radar incident angle; v d A positive value indicates that the target is approaching the radar direction, and a negative value indicates that the target is moving away from the radar direction;

[0108] For spaceborne SAR systems, the f geom ,f elec , and Δf are the same, but f phys It varies for different polarizations; therefore, this method can obtain the Doppler velocity difference of moving targets under different polarizations. and

[0109] In step 5, the covariance correlation matrix [A Cov The amplitude of the elements in ] is calculated using the radar backscatter coefficient and The standard deviation of represents a higher-order term, using and The degree of deviation of Doppler shift and spectrum broadening represents the phase value.

[0110] In step S3, based on the polarization autocorrelation matrix, matrix analysis is used to simultaneously extract polarization and Doppler features from the PolSAR data, including:

[0111] Step a, the scale factor of SPAN(q), for the covariance matrix [C], the total backscatter power SPAN is defined as the sum of the three diagonal elements; the SPAN of the polarization autocorrelation matrix [A] is defined as:

[0112]

[0113] Among them, SPAN C and SPAN A SPAN represents the covariance matrix [C] and the autocorrelation matrix [A] respectively; are the eigenvalues ​​of the covariance matrix [C] and the polarization autocorrelation matrix [A], respectively; Tr([C]) is the SPAN of the covariance matrix [C], and Tr([A] is the SPAN of the polarization autocorrelation matrix [A];

[0114] The elements of the polarization autocorrelation matrices [A] and [C] have a quadratic relationship in amplitude, making the SPAN of [A] smaller than the SPAN of the covariance matrix [C]. To enhance readability and avoid the values ​​of the elements in [A] being too small, the value of the polarization autocorrelation matrix [A] is amplified using a scaling factor q, which is calculated as follows:

[0115]

[0116] Among them, q is the scaling factor Tr() represents finding the trace of the matrix;

[0117] After scaling, the autocorrelation matrix is ​​recorded as A scale =qA; scaled matrix A Dop and A Cov They are and

[0118] Step b, polarization Doppler difference ΔF;

[0119] The Doppler differences between different polarizations are obtained from the matrix A Dop Calculated from the off-diagonal elements of

[0120]

[0121] Where, ΔF ij is the Doppler shift difference between polarization channel i and polarization channel j, Arg() represents the phase of the complex number in brackets;

[0122]

[0123] Step c, the ratio Z of the high-order and zero-order terms of the backscatter coefficient;

[0124] The ratio of the high-order to zero-order terms of the backscatter coefficient Z is calculated using the matrix A Dop and A Cov Calculation of diagonal elements of ;

[0125]

[0126]

[0127]

[0128] Where the denominator of Z represents the average backscattering of the non-sloped surface, while the numerator of Z corresponds to the change in NRCS caused by the surface slope;

[0129] Step d, polarization correlation coefficient of radar backscattering high-order terms

[0130] [A Cov ], is the polarization correlation coefficient Standard deviation of radar backscatter coefficient The product of The calculation is as follows:

[0131]

[0132]

[0133]

[0134]

[0135] in, represents the polarization correlation coefficient of the high-order terms of the radar echo, is less than 1, which measures the degree of depolarization caused by high-order scattering. The phase is equivalent to The phase of the zero-order term is given by [A Dop ] matrix, the corresponding correlation coefficient of the zero-order term is 1; represents the polarization correlation coefficient Standard deviation of radar backscatter coefficient The product of , i, j = 1 (HH), 2 (HV / VH), 3 (VV) and i ≠ j; * indicates conjugate transpose; is the ensemble average of the backscatter coefficients of the three polarizations, is the standard deviation of the backscatter coefficients for the three polarizations, are the Doppler centroids DCs of HH, VV and HV polarizations, yes and The correlation coefficient between and The definition of same.

[0136] Another object of the present invention is to provide a SAR polarization measurement system based on a polarization autocorrelation matrix, the system is implemented by the SAR polarization measurement method based on the polarization autocorrelation matrix, and the system comprises:

[0137] A polarization autocorrelation matrix acquisition module is used to acquire a polarization autocorrelation matrix from a polarization covariance matrix through an autocorrelation operation;

[0138] Phase encoding module, used to capture hidden Doppler information in the frequency domain through the acquired polarization autocorrelation matrix and encode the Doppler information into the phase using high-order statistical methods;

[0139] The polarization and Doppler feature extraction module is used to simultaneously extract polarization and Doppler features from PolSAR data using matrix analysis based on the polarization autocorrelation matrix.

[0140] Combined with all the above technical solutions, the beneficial effects of the present invention are as follows: the present invention proposes a new method for polarimetric measurement of synthetic aperture radar (SAR), and proposes a new descriptor called polarimetric autocorrelation matrix. Unlike polarimetric covariance and correlation matrices, polarimetric autocorrelation matrix can capture hidden Doppler information in the frequency domain and encode it into the phase using high-order statistical methods. The matrix conveniently extracts and analyzes polarization and Doppler information from PolSAR data through matrix analysis. The present invention proposes that the polarimetric autocorrelation matrix can be decomposed into matrices related to Doppler and covariance. The amplitude and phase of these matrices can characterize the Doppler shift and the variance of the backscatter coefficient between different polarizations. The research of the present invention shows that the Doppler shift is not only affected by the radial velocity of the target, but also by the polarization mode of the radar. The present invention uses two fully polarized RADARSAT-2 images for testing, and derives a set of new features and parameters from the polarimetric autocorrelation matrix. The new descriptor and new parameters can detect artificial targets and sea ice, reduce the blur caused by moving targets, and indicate the motion state of ship targets and ocean currents.

[0141] As auxiliary evidence of the inventiveness of the present invention, it is also reflected in the following important aspects:

[0142] (1) The technical solution of the present invention can be used to improve the accuracy and reliability of target detection and recognition systems; reduce the blur caused by moving targets and improve target tracking capabilities; enhance the effectiveness of military and security applications, etc., and can be converted into commercial value in the above-mentioned technical solutions or systems.

[0143] (2) The technical solution of the present invention combines polarization information and Doppler information for parameter inversion, which were previously considered separately.

[0144] (3) The technical solution of the present invention solves the problem that the existing technology lacks quantitative tools to understand the differences between Doppler characteristics between different technical channels, and develops a new tool that combines the target polarization characteristics and the Doppler scattering mechanism.

[0145] (4) The technical solution of the present invention overcomes the technical bias of being unable to combine polarization information with Doppler characteristics for parameter inversion. BRIEF DESCRIPTION OF THE DRAWINGS

[0146] The accompanying drawings herein are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description, serve to explain the principles of the present disclosure;

[0147] Figure 1 It is a flow chart of a SAR polarization measurement method based on a polarization autocorrelation matrix provided by an embodiment of the present invention;

[0148] Figure 2The embodiment of the present invention provides a method for calculating matrices A and A from the scattering matrix S. Dop and A Cov Flowchart of

[0149] Figure 3 This is the scenario 1 provided in the embodiment of the present invention. and (C 11 ) 2 Compare the schematic diagram;

[0150] Figure 4 This is the scenario 1 provided in the embodiment of the present invention. and (C 22 ) 2 Compare the schematic diagram;

[0151] Figure 5 This is the scenario 1 provided in the embodiment of the present invention. and (C 33 ) 2 Compare the schematic diagram;

[0152] Figure 6 This is the scenario 1 provided in the embodiment of the present invention. and Compare the schematic diagrams;

[0153] Figure 7 This is the scenario 1 provided in the embodiment of the present invention. and Compare the schematic diagrams;

[0154] Figure 8 This is the scenario 1 provided in the embodiment of the present invention. and Compare the schematic diagrams;

[0155] Fig. 9 is the high-order term and zero-order term of the radar backscatter coefficient in scenario 1 provided by the embodiment of the present invention for different polarization direction ratios Z HH Histogram;

[0156] Fig.10 is the high-order term and zero-order term of the radar backscatter coefficient in scenario 1 provided by the embodiment of the present invention for different polarization direction ratios Z HV Histogram;

[0157] Fig.11 The high-order and zero-order terms of the radar backscatter coefficient in scenario 1 are calculated for different polarization direction ratios. VV Histogram;

[0158] Fig.12 The field is the co-polarization correlation coefficient |ρ in scene 1 provided in the embodiment of the present invention HH-VV | and Schematic diagram of the correlation between them;

[0159] Fig.13 It is the prior art C in scenario 2 provided by the embodiment of the present invention. 13 Amplitude image;

[0160] Fig.14 This is the present invention A in scenario 2 provided by the embodiment of the present invention 13 Amplitude image;

[0161] Fig.15 This is the present invention in scenario 2 provided by the embodiment of the present invention. Amplitude image;

[0162] Fig.16 This is the present invention in scenario 2 provided by the embodiment of the present invention. The magnitude image of ;

[0163] Fig.17 It is the prior art C in scenario 2 provided by the embodiment of the present invention. 13 Phase image;

[0164] Fig.18 This is the present invention A in scenario 2 provided by the embodiment of the present invention 13 Phase image;

[0165] Fig.19 This is the present invention in scenario 2 provided by the embodiment of the present invention. Phase image;

[0166] Fig. 20 This is the present invention in scenario 2 provided by the embodiment of the present invention. Phase image of

[0167] Fig.21 is the ratio of the high-order and zero-order terms of the backscatter coefficients in different polarization directions in scenario 2 provided by the embodiment of the present invention. HH Histogram.

[0168] Fig. 22 is the ratio of the high-order and zero-order terms of the backscatter coefficients in different polarization directions in scenario 2 provided by the embodiment of the present invention. HV Histogram;

[0169] Fig.23 is the ratio of the high-order and zero-order terms of the backscatter coefficients in different polarization directions in scenario 2 provided by the embodiment of the present invention. VV Histogram;

[0170] Fig.24 This is the scenario 2 provided by the embodiment of the present invention. HH-VV |Co-polarization correlation coefficient plot;

[0171] Fig.25 This is the scenario 2 provided in the embodiment of the present invention. Co-polarization correlation coefficient plot;

[0172] Fig.26 This is the scenario 2 provided in the embodiment of the present invention. Comparison of co-polarization correlation coefficients between . DETAILED DESCRIPTION

[0173] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific implementation disclosed below.

[0174] The innovation of the present invention is that the present invention derives a polarization autocorrelation matrix [A] from the C matrix through autocorrelation operation, and the matrix can effectively capture Doppler information hidden in the frequency domain, encode it into the phase, and invert the polarization information and the Doppler information together. In addition, the present invention derives two new descriptors based on the polarization autocorrelation matrix.

[0175] The present invention proposes an innovative target polarization and Doppler descriptor, namely the polarization autocorrelation matrix, denoted as the [A] matrix. The main innovations and contributions of the proposed new matrix are as follows:

[0176] (1) The descriptor is derived from the polarization covariance matrix through an autocorrelation operation. Compared with classical polarization matrices (such as [S], [C], and [T] matrices), the newly proposed [A] matrix can effectively capture the Doppler information hidden in the frequency domain and encode it into the phase due to the use of high-order statistical methods.

[0177] (2) Based on [A], the present invention can use matrix analysis to simultaneously extract polarization and Doppler features from PolSAR data, which is a capability that the widely used [C] and [T] matrices do not have.

[0178] Tested using real fully polarimetric RADARSAT-2 images, the new descriptors, along with new parameters, demonstrate that the Doppler shift or DC in synthetic aperture radar returns is not only sensitive to the radial velocity of the target, as traditionally understood, but also to the radar polarization, an aspect that has not been fully explored in the past.

[0179] The descriptor proposed in the present invention is a qualitative change from the traditional polarization analysis method. The traditional method mainly focuses on the relative polarization phase of polarization scattering and target scattering, while this descriptor can deeply understand the properties of mobile targets and scattering phenomena.

[0180] To facilitate understanding of the relevant technologies of the present invention, general concepts about statistics and detailed derivation of polarization autocorrelation matrix elements are provided, and the correlation properties of complex backscatter coefficients and relevant contents of the DC estimation method based on the time domain are explained.

[0181] A. The expected value and covariance of two random variables.

[0182] In probability theory and statistics, for two random variables X and Y, each with finite variance (var[X] and var[Y]) and expected value (E[X] and E[Y]), the covariance COV between the two random variables is defined as:

[0183] COV(X,Y)=E[XE[X]]E[YE[Y]]

[0184] =E[XY]-E[X]E[Y] (X1);

[0185] In the formula, COV() represents the covariance, E() represents the variance, and X and Y represent random variables; from (X1), the present invention can obtain the expected value of XY as:

[0186] E[XY]=E[X]E[Y]+COV(X,Y) (X2);

[0187] Where E[XY] is the correlation between two random variables X and Y. COV(X,Y) can be calculated by the correlation coefficient of these two random variables, the correlation coefficient ρ XY Defined as:

[0188]

[0189] in, and are the standard deviations of random variables X and Y, respectively. Substituting formula (X3) into formula (X2), the present invention obtains:

[0190] E[XY]=E[X]E[Y]+ρ XY ΔXΔY (X4);

[0191] Additionally, it should be noted that the covariance COV(X,Y) of random variables X and Y has the following properties:

[0192]

[0193] B. and Detailed derivation of .

[0194] (1) Derivation of .

[0195] According to formula (15), It can be expressed as:

[0196]

[0197] Because E{|S HH (k+m)| 2} and E{|S HH (m)| 2} are all backscatter coefficients |S HH | 2 The ensemble average of , so:

[0198]

[0199] (2) Derivation of .

[0200] According to (X3), It can be expressed as:

[0201]

[0202] Among them, ρ HH-HH are two random variables and The correlation coefficient between Δ HH+ k represents the sliding window |S HH (k+m)| 2 The standard deviation of HH Indicates the same window |S HH (m)| 2 Here, |S HH | 2 The standard deviation of Obviously, |S HH (k+m)| 2 and |S HH (m)| 2 There is a strong correlation, and Δ HH+ k≈Δ HH So ρ HH-HH ≈1, And the present invention can obtain:

[0203]

[0204] By combining formulas (X5), (X7) and (15), A11 It can be expressed as:

[0205]

[0206] C. A 13 Detailed derivation of .

[0207] (1) Derivation of .

[0208] Referring to formula (5), the present invention introduces two variables:

[0209]

[0210] According to formula (3), and The average backscattering coefficient and Related:

[0211]

[0212] According to formula (8), if the moving target has a Doppler frequency shift f relative to the radar DC , then the autocorrelation coefficient of the moving target can be expressed as follows:

[0213]

[0214] Substituting (X12) into (X13):

[0215]

[0216] To calculate and The conjugate ( and ), the present invention first analyzes the conjugate property of Fourier transform. The conjugate property of Fourier transform is stated as follows:

[0217] For the function R0(k) in the time domain, its Fourier transform is So The Fourier transform of (complex conjugate) is

[0218] Based on this property, a Doppler frequency shift f DC The moving target, its autocorrelation coefficient and its Fourier transform of autocorrelation coefficient are:

[0219]

[0220] By combining formula (7), (8) and (X15), we get and The expression is as follows:

[0221]

[0222] Substituting (X14) and (X16) into formula (16) yields:

[0223]

[0224] As mentioned above, for the PolSAR system, k=1, T=1 / PRF.

[0225] (2) Derivation of .

[0226] In formula (16), Defined as:

[0227]

[0228] According to (X3), It can also be written as:

[0229]

[0230] in, yes and The correlation coefficient between them. Within the sliding window, Δ CHH and Δ CVV Respectively represent and Obviously, is a plural number, and and are all positive real numbers. Among them, among them, and are the standard deviations of the backscatter coefficients for HH and VV polarizations, respectively.

[0231] By combining formulas (X17), (X19) and (16), A 13 It can be expressed as:

[0232]

[0233] D. Correlation properties of the complex backscattering coefficient.

[0234] Radar echoes are considered as composite reflections because one radar resolution cell contains multiple scatterers. A practical model can be established by assuming that the scene consists of a set of point scatterers and that the amplitude and phase of the composite reflection need to be calculated. According to the prior art, the complex scattering coefficient per unit area, σ(r), can be expressed as:

[0235]

[0236] Among them, σ l is the radar cross section (RCS) of scatterer l, θ l is the total phase shift of the scatterer l due to reflection and slant range delay, r l is the position of the scatterer l, and δ(r) is the two-dimensional Dirac δ function. From (X21), the related properties of the complex backscattering coefficient are similar to the average scattering coefficient σ 0 (r) Relevant to:

[0237] R σ (r1,r2)=E{σ(r1)σ * (r2)}=δ(r1-r2)σ 0 (r2) (X22);

[0238] Where * represents conjugation and E{·} represents the ensemble mean or mathematical expectation. For homogeneous scenes, it has the following properties:

[0239] R σ (r1,r2)=R σ (r1-r2)=σ 0 δ(r1-r2) (X23);

[0240] By combining (X22) and (X23), the autocorrelation of the radar target scene can be rewritten as:

[0241] R σ (r+Δr,r)=E{σ(r+Δr)σ * (r)}=σ 0 (r)δ(Δr) (X24);

[0242] E. DC estimation based on time domain method.

[0243] A DC estimation algorithm based on the time domain is called the Correlation Doppler Estimator (CDE). The basic idea of ​​the time domain technique is to use the Fourier relationship between the signal power spectrum and its correlation function. Therefore, moving the signal in the frequency domain will produce a linear phase shift in the correlation coefficient, thus directly measuring the frequency shift.

[0244] If h0(m) is a random process, its autocorrelation function separated by k samples is given by:

[0245]

[0246] The power spectrum is calculated as:

[0247]

[0248] Where T is the inverse of the radar pulse repetition frequency (PRF), that is, (T = 1 / PRF). F{·} is the Fourier transform. If there is a Doppler frequency shift f in the power spectrum DC , then:

[0249] S h (f) = S0(ff DC )(X27);

[0250] At the same time, the autocorrelation function is rewritten as:

[0251]

[0252] Formula (X28) reveals the direct relationship between the phase and DC offset in the correlation function. Therefore, by estimating the correlation function Rh(k), an estimate of DC can be obtained:

[0253]

[0254] Here, arg{·} represents the phase angle. Here, a widely used k value is generally recommended to be k=1, because the sampling frequency of most SAR systems is only slightly higher than the Nyquist rate.

[0255] Embodiment 1: The SAR polarization measurement method based on the polarization autocorrelation matrix provided by the embodiment of the present invention includes:

[0256] S1, obtain the polarization autocorrelation matrix from the polarization covariance matrix through autocorrelation operation;

[0257] S2, captures the hidden Doppler information in the frequency domain by acquiring the polarization autocorrelation matrix and encodes the Doppler information into the phase using high-order statistical methods;

[0258] S3, based on the polarization autocorrelation matrix, matrix analysis is used to simultaneously extract polarization and Doppler features from PolSAR data.

[0259] In the embodiments of the present invention, the newly proposed [A] matrix can capture the hidden Doppler information in the frequency domain and encode it in the phase. The new matrix can simultaneously extract polarization and Doppler information in different polarization channels in PolSAR data through matrix analysis, which is a significant improvement over the traditional polarization analysis method that only focuses on target polarization scattering and relative polarization phase. The proposed descriptor can provide a deep explanation of target characteristics and scattering phenomena, and provides unique insights into target characteristics and scattering mechanisms in SAR polarization.

[0260] Similar to the covariance and correlation matrices, the polarization autocorrelation matrix [A] is a 3×3 positive semidefinite Hermitian matrix. It can be decomposed into a Doppler-related matrix [ADop ] and a matrix associated with the covariance [A Cov] , they are both 3×3 positive semidefinite Hermitian matrices. [A] and [A Cov ] are all matrices of rank 3, and [A Dop ] is a matrix of rank 1. Dop ] is related to the ensemble average radar scattering coefficient (zero-order term of NRCS), while [A Dop The phase of the non-diagonal elements in ] represents the Doppler frequency shift difference between different polarizations (referred to as polarization Doppler difference in the present invention). Dop ] compared to the matrix [A Cov ] computes the covariance of the matrix [A]. Cov The amplitude of the elements in [A] is determined by the standard deviation of the scattering coefficient (corresponding to the higher-order terms of the NRCS), while Cov The phase of ] is related to the degree of deviation from the polarization Doppler difference and spectrum broadening. In addition, a set of new features and metrics are derived from the polarization autocorrelation matrix, such as the scale factor q, the polarization Doppler difference ΔF, the ratio of the high-order term to the zero-order term of the backscatter coefficient Z, and the polarization correlation coefficient of radar backscatter All of these help to understand and characterize the scattering mechanisms.

[0261] Embodiment 2: The construction of the polarization autocorrelation matrix provided by the embodiment of the present invention includes: the correlation properties of the complex backscatter coefficient and the DC estimation method based on the time domain (see the above-mentioned parts D and E).

[0262] The first step is to define the polarization autocorrelation matrix.

[0263] In the PolSAR system, the Sinclair scattering matrix [S] can be expressed as:

[0264]

[0265] Where S is the Sinclair scattering matrix, S HH is the scattering coefficient from horizontal polarization to horizontal polarization, S HV is the scattering coefficient from horizontal polarization to vertical polarization, S VH is the scattering coefficient from vertical polarization to horizontal polarization, S VV is the scattering coefficient from vertical polarization to vertical polarization;

[0266] Under the condition of single-station SAR, formula (1) satisfies the reciprocity theorem, that is, S HV =S VH ;

[0267] The covariance matrix C is obtained by calculating the outer product of the scattering vector of the Sinclair scattering matrix [S] and its conjugate transpose;

[0268]

[0269] Where C is the covariance matrix, * represents the conjugate transpose, k is the scattering vector, is the complex conjugate transpose of the scattering vector

[0270] Calculate the autocorrelation function of each element in the covariance matrix C. The symbol R represents the autocorrelation operation performed on each element in the matrix. The polarization autocorrelation matrix [A] is defined as:

[0271]

[0272] Where A is the polarization autocorrelation matrix, [A] is the polarization autocorrelation matrix, and R() is the autocorrelation operation.

[0273] The present invention innovatively proposes formula (3), and the technical function is to extract and invert the polarization information and Doppler information by integrating them. Before the present invention proposes the polarization autocorrelation matrix, the two are separated and considered separately, while the present invention directly performs polarization autocorrelation processing on the C matrix and then obtains a new result; the detailed steps of calculating the autocorrelation operation are shown in the second step.

[0274] The second step is to polarize the elements of the autocorrelation matrix.

[0275] By rearranging formula (3), the present invention obtains the following results:

[0276]

[0277] Where Aij represents the element in the polarization autocorrelation matrix, i,j = 1, 2, 3; A 11 ,A 22 ,A 33 is a real number, A 12 and A 21 ,A 13 and A 31 ,A 23 and A 32 are complex conjugates of each other;

[0278]

[0279]

[0280]

[0281]

[0282]

[0283]

[0284]

[0285]

[0286]

[0287] Where R() and E() represent autocorrelation operation and expectation, respectively; in PolSAR images, m is the number of rows in the azimuth direction, k is the interval between two samples along the azimuth direction, * represents conjugate transposition, and S HH is the scattering coefficient from horizontal polarization to horizontal polarization, S HV is the scattering coefficient from horizontal polarization to vertical polarization, S VH is the scattering coefficient from vertical polarization to horizontal polarization, S VV is the scattering coefficient from vertical polarization to vertical polarization;

[0288] A 11 and A 13 Derived, obtain the elements of the polarization autocorrelation matrix, A 11 The derivation formula is expressed as:

[0289]

[0290] A 11 It consists of two parts:

[0291]

[0292] in:

[0293]

[0294]

[0295] A 13 Decomposed into the sum of two parts, the expression is:

[0296]

[0297]

[0298] In the formula, COV() represents covariance, A 11 and A 13 are the elements in the polarization autocorrelation matrix; Represent the first element of the polarization autocorrelation matrix and its corresponding Dop and Cov respectively;

[0299] in:

[0300]

[0301]

[0302] By reference A 11 and A 13 , establish the expression for all elements in the matrix [A]:

[0303]

[0304]

[0305]

[0306]

[0307]

[0308]

[0309]

[0310]

[0311]

[0312] In the formula, is the ensemble average of the backscatter coefficients of the three polarizations, is the standard deviation of the backscatter coefficients for the three polarizations, are the Doppler centroid (DC) of HH, VV and HV polarizations, respectively, yes and The correlation coefficient between and The definition of The same; e is the natural logarithm, j is the imaginary unit, and T is the pulse repetition period.

[0313] The present invention innovatively proposes formula (5) to formula (8), and its technical function is to provide a detailed mathematical derivation process and results of polarization autocorrelation matrix elements. The present invention directly performs polarization autocorrelation processing on the C matrix, and then directly obtains a new result.

[0314] The third step is to obtain the polarization autocorrelation matrix.

[0315] According to formula (8), each element of the polarization autocorrelation matrix [A] can be divided into two parts, namely the "Dop" term and the "Cov" term. Therefore, [A] can be written as the Doppler correlation matrix [A Dop ] and the covariance correlation matrix [A Cov ], as follows:

[0316]

[0317] in,

[0318]

[0319]

[0320]

[0321]

[0322]

[0323]

[0324]

[0325]

[0326]

[0327]

[0328]

[0329]

[0330] in, are the elements in the Doppler correlation matrix, is the element in the covariance correlation matrix, i,j=1,2,3;

[0331] Each element of the polarization autocorrelation matrix is ​​divided into a Dop term and a Cov term; therefore, the polarization autocorrelation matrix is ​​written as the Doppler correlation matrix [A Dop ] and the covariance correlation matrix [A Cov ] and; use and The degree of deviation of Doppler shift and spectrum broadening represents the phase value.

[0332] The invention innovatively proposes formula (9), whose technical function is to provide an expression and calculation method of the polarization autocorrelation matrix. The invention directly performs polarization autocorrelation processing on the C matrix, and then directly obtains a new result.

[0333] Embodiment 3, based on the polarization autocorrelation matrix obtained in Embodiment 2, captures hidden Doppler information in the frequency domain through the acquired polarization autocorrelation matrix, and encodes the Doppler information into the phase using a high-order statistical method, specifically including:

[0334] Physical interpretation of the polarization autocorrelation matrix.

[0335] According to formula (4) and formula (8), it is easy to know that the polarization autocorrelation matrix [A] is a 3×3 positive semidefinite Hermitian matrix, similar to the covariance and correlation matrices. Therefore, [A] has three real non-negative eigenvalues ​​and orthogonal eigenvectors. Compared with [A], [A Dop ] and [A Cov ] are all 3×3 positive semidefinite Hermitian matrices. However, [A Dop ] is a rank 1 matrix, and [A Cov ] is a rank 3 matrix. This means that [A Dop ] has only one real non-negative eigenvalue, and [A Cov ] has three real nonnegative eigenvalues.

[0336] For a rank 1 matrix [A Dop ], it can always be expressed as the product of two one-dimensional vectors:

[0337]

[0338] Among them, k Dop is defined as:

[0339]

[0340] Covariance correlation matrix [A Cov ] by subtracting the Doppler correlation matrix [A] from the polarization autocorrelation matrix [A] Dop ] is calculated, and the expression is:

[0341] A Cov =[A Cov ]=AA Dop =[A]-[A Dop ](11).

[0342] The following are matrices A, A Dop and A Cov A brief summary of the calculation method of Figure 2 Calculate the matrices A and A from the scattering matrix S Dop and A Cov Flowchart of the process.

[0343] Specifically include:

[0344] Step 1: derive a 3×3 matrix from a 2×2 Sinclair scattering matrix S

[0345] Step 2: Calculate in a sliding window and Set k to 1. The sliding window size is customizable and can be set to common values ​​such as 3, 5, 7, 9, 11, or other sizes.

[0346] Step 3: In the sliding window of the same size, use formula (3) to Apply the autocorrelation operation to each element in to obtain the [A] matrix.

[0347] Step 4: Calculate the Doppler term [A] of the polarization autocorrelation matrix [A] by formula (10). Dop ]matrix.

[0348] Step 5: Calculate the high-order covariance term [A] of the polarization autocorrelation matrix [A] by formula (11): Cov ]matrix.

[0349] In the embodiment of the present invention, A Dop The physical explanation of is as follows:

[0350] In [A Dop ], its amplitude is related to the ensemble average of the radar backscatter coefficients, while A Dop The phase of A indicates the difference in Doppler shift between different polarizations. In particular, Dop The phase of the off-diagonal elements in is determined by the DC frequency of the HH, VV, and HV polarizations ( and )control.

[0351] However, the Doppler shift or DC (f DC ) is affected by multiple contributing sources, including platform motion, radar attitude, and surface moving targets, and is expressed as:

[0352] f DC =f geom +f elec +f phys +Δf (12);

[0353] Among them, f geom is the Doppler term generated by satellite geometry and attitude, f elec is the Doppler term generated by antenna pointing; Δf is the Doppler residual caused by non-geophysical terms and unknown deviations; f phys is the Doppler term generated by the ground moving target, f physCommonly used to calculate the radial Doppler velocity v of ground moving targets d ;

[0354]

[0355] Among them, k e is the electromagnetic wave number of the radar, θ i is the radar incident angle; v d A positive value indicates that the target is approaching the radar direction, and a negative value indicates that the target is moving away from the radar direction;

[0356] For spaceborne SAR systems, the f geom ,f elec , and Δf are the same, but f phys It varies for different polarizations; therefore, this method can obtain the Doppler velocity difference of moving targets under different polarizations. and

[0357] In step 5, the covariance correlation matrix [A Cov The amplitude of the elements in ] is calculated using the radar backscatter coefficient and The standard deviation of represents a higher-order term, using and The degree of deviation of Doppler shift and spectrum broadening represents the phase value.

[0358] For spaceborne SAR systems, f geom ,f elec , and Δf are the same, but f phys For different polarizations, it is different. Therefore, this method can obtain the Doppler velocity difference of moving targets under different polarizations. and

[0359] In short, the matrix [A Dop ] has a significant advantage in intuitively and easily representing the hidden Doppler shift in the radar power spectrum through its phase component.

[0360] In the embodiment of the present invention, A Cov The physical explanation of is as follows:

[0361] With the matrix [A Dop ] compared to the matrix [A Cov ] Computes the higher-order covariance of the matrix [A]. The matrix A Cov The amplitude of the element depends on the backscattering coefficient and Typically, the radar backscatter coefficient is defined as the sum of the zeroth, first, and second order scattering.

[0362] This assumption is often used in marine applications. According to the existing technology, the sea surface radar scattering cross section (σ 0 ) can be expressed as the sum of zero-order terms and higher-order terms. The zero-order term refers to the undisturbed zero-order component, which is obtained by eliminating the slope and elevation in the direction parallel to the radar observation direction and the normal direction, respectively. The higher-order term is the sum of the second-order terms, which depends on the mean square surface slope parallel to the radar observation direction and the normal direction. In the prior art, the total σ from the ocean surface 0 It can be represented by the sum of Bragg scattering (low-order terms) and the non-polarized radar backscatter coefficient from the waves (high-order terms).

[0363] Among them, [A Dop ] is the average of the radar backscatter coefficients corresponding to the zero-order term in the overall NRCS. and At the same time, [A Cov ] in the radar backscatter coefficient and The standard deviation of represents the higher-order terms. Extracting the zero-order and higher-order terms of NRCS is crucial for understanding and characterizing the scattering mechanism and target characteristics. In particular, it is very important for analyzing and evaluating the impact of target Doppler motion on radar electromagnetic scattering characteristics.

[0364] [A Cov The physical meaning of the phase of the off-diagonal elements in ] is not as clear as its magnitude. From a mathematical point of view, the phase seems to be the same as the Doppler shift ( and ) is related to the degree of deviation of the spectrum broadening. Specifically, a higher degree of deviation or a wider spectrum corresponds to a larger phase value.

[0365] It can be understood that the technical function of the innovative formula (10) to formula (13) proposed by the present invention is to explain the physical meaning of the polarization autocorrelation matrix. The present invention can directly perform polarization autocorrelation processing on the C matrix, and then directly obtain a new result.

[0366] Embodiment 4, based on the polarization autocorrelation matrix, using matrix analysis to simultaneously extract polarization and Doppler features from PolSAR data includes: extracting new features and new measurable variables.

[0367] Due to the similarity between [A] and [C], both of them are 3×3 positive semidefinite Hermitian matrices, so the common features in [C] can also be applied to [A]. For example, the decomposition based on eigenvalue-eigenvector, such as Cloude-Pottier decomposition, can be applied to the polarization autocorrelation matrix. The polarization autocorrelation matrix proposed in the present invention combines polarization information and Doppler information to extract inversion. In addition, new features and parameters are extracted.

[0368] The present invention focuses on discovering and extracting new features and new measurable variables from the polarization autocorrelation matrix. Specifically, it includes:

[0369] Step a, scaling factor of SPAN(q).

[0370] For the covariance matrix [C], the total backscattered power (SPAN) is defined as the sum of the three diagonal elements. Similarly, the SPAN of [A] can be defined in the same way:

[0371]

[0372] Among them, SPAN C and SPAN A SPAN represents the covariance matrix [C] and the autocorrelation matrix [A] respectively; are the eigenvalues ​​of the covariance matrix [C] and the polarization autocorrelation matrix [A], respectively; Tr([C]) is the SPAN of the covariance matrix [C], and Tr([A] is the SPAN of the polarization autocorrelation matrix [A];

[0373] The elements of the polarization autocorrelation matrices [A] and [C] have a quadratic relationship in amplitude, making the SPAN of [A] smaller than the SPAN of the covariance matrix [C]. To enhance readability and avoid the values ​​of the elements in [A] being too small, the value of the polarization autocorrelation matrix [A] is amplified using a scaling factor q, which is calculated as follows:

[0374]

[0375] Among them, q is the scaling factor Tr() represents finding the trace of the matrix;

[0376] After scaling, the autocorrelation matrix is ​​recorded as A scale =qA; scaled matrix A Dop and A Cov They are and

[0377] Step b, polarization Doppler difference ΔF;

[0378] The Doppler differences between different polarizations are obtained from the matrix A Dop Calculated from the off-diagonal elements of

[0379]

[0380] Where, ΔF ij is the Doppler shift difference between polarization channel i and polarization channel j, Arg() represents the phase of the complex number in brackets;

[0381]

[0382] Step c, the ratio Z of the high-order and zero-order terms of the backscatter coefficient;

[0383] The ratio of the high-order to zero-order terms of the backscatter coefficient Z is calculated using the matrix A Dop and A Cov Calculation of diagonal elements of ;

[0384]

[0385]

[0386]

[0387] Where the denominator of Z represents the average backscattering of the non-sloped surface, while the numerator of Z corresponds to the change in NRCS caused by the surface slope;

[0388] Step d, polarization correlation coefficient of radar backscattering high-order terms

[0389] [A Cov ], is the polarization correlation coefficient Standard deviation of radar backscatter coefficient The product of The calculation is as follows:

[0390]

[0391] in, represents the polarization correlation coefficient of the high-order terms of the radar echo, is less than 1, which measures the degree of depolarization caused by high-order scattering. The phase is equivalent to The phase of the zero-order term is given by Dop Derived from the matrix, the corresponding correlation coefficient of the zero-order term is 1; represents the polarization correlation coefficient Standard deviation of radar backscatter coefficient The product of , i, j = 1 (HH), 2 (HV / VH), 3 (VV) and i ≠ j; * indicates conjugate transpose; is the ensemble average of the backscatter coefficients of the three polarizations, is the standard deviation of the backscatter coefficients for the three polarizations, are the Doppler centroid (DC) of HH, VV and HV polarizations, respectively, yes and The correlation coefficient between and The definition of same.

[0392] This conclusion has been confirmed in the prior art, where the zero-order term corresponds to the scattering of small perturbations without tilting or "pure" Bragg scattering, resulting in the inability to describe the depolarization effect. Therefore, the correlation coefficient of the zero-order term is always equal to 1. In contrast, the higher-order terms originate from the modulation of the radar echo by the surface slope, which also controls the degree of depolarization. Therefore, is less than 1, and it measures the degree of depolarization caused by higher-order scattering. Finally, as a complex number, The phase is equivalent to phase.

[0393] It can be understood that the innovative formula (14) to (19) proposed in the present invention has the technical effect of defining new features and new parameters derived from the polarization autocorrelation matrix, the mathematical derivation process and the resulting expression. The present invention can directly perform polarization autocorrelation processing on the C matrix, and then directly obtain a new result.

[0394] Embodiment 5, the present invention also provides a synthetic aperture radar polarization measurement system based on polarization autocorrelation matrix, the system comprising:

[0395] A polarization autocorrelation matrix acquisition module is used to acquire a polarization autocorrelation matrix from a polarization covariance matrix through an autocorrelation operation;

[0396] Phase encoding module, used to capture hidden Doppler information in the frequency domain through the acquired polarization autocorrelation matrix and encode the Doppler information into the phase using high-order statistical methods;

[0397] The polarization and Doppler feature extraction module is used to simultaneously extract polarization and Doppler features from PolSAR data using matrix analysis based on the polarization autocorrelation matrix.

[0398] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0399] Since the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present invention, their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0400] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated in one processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment.

[0401] An embodiment of the present invention further provides a computer device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor implements the steps in any of the above-mentioned method embodiments when executing the computer program.

[0402] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.

[0403] An embodiment of the present invention further provides an information data processing terminal, which, when executed on an electronic device, provides a user input interface to implement the steps in the above method embodiments. The information data processing terminal is not limited to mobile phones, computers, and switches.

[0404] An embodiment of the present invention further provides a server, which is used to provide a user input interface to implement the steps in the above method embodiments when executed on an electronic device.

[0405] An embodiment of the present invention provides a computer program product. When the computer program product runs on an electronic device, the electronic device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0406] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the camera / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk.

[0407] To further illustrate the effects of the embodiments of the present invention, the following experiment was conducted.

[0408] SAR data validation and analysis.

[0409] The present invention applies the new matrix to full-polarization RADARSAT-2 data covering different scenarios, demonstrating the potential of the new descriptor.

[0410] Step 1: Dataset and preprocessing.

[0411] The present invention uses two fully polarized RADARSAT-2C band images, and these data are provided in SLC slant range format. Table 1 provides basic information of the data. The present invention can also obtain Pauli RGB composite images of RADARSAT-2 fully polarized SAR images and Google Earth images through a certain national space agency, covering: a certain sea area 1 and a certain sea area 2. These data sets provide the present invention with a diversity of time series and incident angles.

[0412] Table 1 Main parameters of RADARSAT-2 quad-polarization SAR data

[0413]

[0414] Specifically, an image of a certain sea area was collected on December 9, 2008, in fine-quad-pol mode, corresponding to the FQ9 beam (descending orbit). The image was acquired at a medium incidence angle (about 30°) in the southern waters of a certain sea area, and a Google Earth image of the same area is also shown in an image of a certain sea area. The upper right part of the image of a certain sea area shows several different areas, and the middle waters are a certain channel 1 and a certain channel 2. An island is located in the upper left part of the image. The lower part of the SAR image is an open water area, which contains some islands, ship targets and suspected oil pollution caused by ship leakage.

[0415] A RADARSAT-2 image of a sea area covers an offshore wind farm located 45 km north of an island in a certain country. This SAR image was acquired in standard full polarization mode, using the SQ27 beam, with a large radar incidence angle (about 46°). Twelve wind turbines are visible in the lower part of the RADARSAT-2 image of a sea area. A large dark spot area can be clearly seen in the upper part compared to the surrounding background. According to ECMWF data records, at the time of radar imaging, the wind speed and significant wave height were 3.2 m / s and 0.22 m, respectively, indicating low sea conditions. Low wind speeds result in a flat sea surface, so these areas appear as dark spots on the image.

[0416] The preprocessing of the RADARSAT-2 dataset includes the following steps: First, the scattering matrix [S] is extracted from the RADARSAT-2 SLC image using PolSARpro 6.0.2 software. Then, an 11×11 sliding window is used to Figure 2 The flowchart presented in generates the polarization autocorrelation matrix [A].

[0417] In addition, the ship targets have polarization and motion characteristics, making them important objects for analysis in the present invention. In the present invention, AIS data is obtained to mark the ship targets in the SAR image. The method of matching SAR and AIS is described in detail in the prior art.

[0418] Step 2: Visualization and analysis of the polarization autocorrelation matrix.

[0419] According to the definitions of the polarization covariance matrix and the polarization autocorrelation matrix, their mathematical expressions are very similar. Therefore, the present invention can directly compare the polarization autocorrelation matrix with the polarization covariance matrix to illustrate the difference between the polarization autocorrelation matrix and the conventional second-order statistical matrix in describing the target scattering characteristics.

[0420] Study scenario 1: land-sea scenario.

[0421] In a test scenario in a certain sea area, which is a typical land-sea scenario, the present invention first verifies the consistency of the diagonal elements between the [A] matrix and the [C] matrix.

[0422] According to formula (18) and the physical interpretation of the polarization autocorrelation matrix, [A Dop ] and [A Cov The diagonal elements of ] are the zero-order term and the square of the higher-order terms of the radar backscatter coefficient. Figure 3-Figure 8 The [A Dop ] and [A Cov ] and the relationship between the diagonal elements and the radar backscatter coefficient. Figure 3-Figure 8 It can be seen that [A Dop ] and [A Cov The correlations between the diagonal elements of ] and the radar backscatter coefficient are all higher than 0.96, which provides substantial support for the theory proposed in this invention. Figure 3 for VS and (C 11 ) 2 Schematic diagram; Figure 4 for VS and (C 22 ) 2 Schematic diagram; Figure 5 for VS and (C 33 ) 2 Schematic diagram; Figure 6 for and Compare the schematic diagrams; Figure 7 for and Compare the schematic diagrams; Figure 8 for and Compare the schematic diagram.

[0423] The present invention can also obtain C 13 , A 13 , as well as As well as the magnitude and phase of . Compared to the [C] matrix, the matrix [A] more effectively represents the difference between moving and stationary objects. In the magnitude image, A 13 , as well as The land-sea boundary ratio C 13 In addition, referring to the Pauli-RGB image of a certain sea area, it can be known that the low wind area in the upper half of the Pauli-RGB image is in A13 as well as It is easier to identify in the amplitude image.

[0424] The effects due to motion are more pronounced in the phase of the polarization autocorrelation matrix, especially in the phase related to Doppler. Figure 3-Figure 8 , in order to verify the correlation between the diagonal elements of [A] and the radar backscatter coefficient using data from a test site in a certain sea area. Figure 3-Figure 5 Shows [A Dop ] with the diagonal elements of [C] (representing the zeroth-order terms of the backscatter coefficient). Figure 6-Figure 8 Shows [A Cov ] and the correlation between the diagonal elements of the backscattering coefficient and the higher-order terms.

[0425] Existing technology C in a certain sea area 13 , Prior Art A 13 The present invention And the present invention As well as the amplitude and phase, we can know that C 13 The phase difference between the two co-polarized channels effectively identifies even-order scattering. 13 However, C 13 The phase of C cannot distinguish between bulk scattering and odd-spot scattering. 13 In the phase of C 13 In contrast, the present invention and The phases differ significantly between ocean and land. In particular, The phase shows a clear sea / land boundary.

[0426] Among them, in the present invention In the phase diagram, The phase of the wave shows an obvious negative phase in one channel, while it shows a very prominent positive phase in another channel. This phenomenon is caused by the difference in the co-polarization Doppler frequency shift of the sea surface (hereinafter referred to as the co-polarization frequency difference).

[0427] In order to verify the above effect, the DC anomaly method (DCA) was used to calculate the DC anomaly of HH and VV polarizations, where the grid size was 2.0km×2.0km, and the results can be obtained from the NRCS and Doppler shift of HH polarization derived from DCA. These include the Doppler shift of HH, the Doppler shift of VV, and the difference in the co-polarization Doppler shift (HH-VV); the calculated DC anomaly has been corrected to eliminate satellite geometry and attitude terms, antenna pointing errors, and non-geophysical deviations. The Doppler shift diagram of HH shows that there is a difference between the DC of HH polarization and the DC of VV polarization (HH minus VV), and the maximum frequency difference can reach 100Hz. By comparing the existing technology C in a certain sea area, the DC anomaly of HH is corrected to eliminate the satellite geometry and attitude terms, antenna pointing errors, and non-geophysical deviations. 13 , the present invention A 13 The present invention And the present invention As well as the amplitude and phase of , and the NRCS and Doppler shift of HH polarization derived from DCA, the present invention can see The spatial pattern of the phase of is very similar to the DCA-derived co-polarization frequency difference (the difference in co-polarization Doppler shift (HH-VV)), except that The phase has a higher spatial resolution (about 100 meters). In addition, the ΔF calculated using formula (16) HH-VV The co-polarization frequency differences show strong numerical agreement.

[0428] The measurable variable q, used as a scaling factor to estimate the power ratio of [C] to [A], shows the same properties as predicted, with lower values ​​in land areas and higher values ​​in ocean areas (see the scaling factor q calculated from a test site in a certain sea area). This creates a clear contrast between ocean and land, making the oil film clearly visible.

[0429] like Fig. 9 The high-order and zero-order terms of the radar backscatter coefficient are calculated for different polarization direction ratios. HH Histogram; e.g. Fig.10 The high-order and zero-order terms of the radar backscatter coefficient are calculated for different polarization direction ratios. HV Histogram; e.g. Fig.11 The high-order and zero-order terms of the radar backscatter coefficient are calculated for different polarization direction ratios. VV Histogram; these data are from a test site in a certain sea area. Regarding the variable Z, it can characterize the ratio of the high-order and zero-order terms of the backscatter coefficient. Figure 9-11 It can be seen that the percentage of high-order scattering in cross-polarization is higher than that in co-polarization. In addition, the Z value of cross-polarization has a clearer land boundary and has the performance and ability to extract the coastline.

[0430] Furthermore, the present invention further analyzes the co-polarization correlation coefficient |v HH-VV | and Comparison between. HH-VV | is calculated from the covariance matrix [C], and It is from [A Cov ] calculated in . Despite their different definitions, they show a strong correlation (see Fig.12 ), which indicates that the high-order terms in the co-polarization echo make a significant contribution. In order to provide a more detailed comparison, the present invention analyzes the calculation With |ρ HH-VV |(i.e. ). It can be seen that |ρ HH-VV |Usually greater than There are exceptions for marine vessels and some land areas. Comparison with optical images from Google Earth shows that these exceptions correspond to man-made objects with significant even-order scattering contributions, such as urban buildings. For man-made objects, higher The value can be explained by the strong higher-order scattering components of the target with significant secondary scattering. The ratio of can obtain the result of threshold segmentation, and the empirical threshold is set to 1.2. Typical secondary scattering artificial structures can be easily detected.

[0431] (2) Study scenario 2: pure ocean scenario.

[0432] The existing technology C of a certain sea area in the NS test site is given. 13 , the present invention A 13 , the present invention And the present invention Compared with a test site in a certain sea area, the NS test site is an image of a pure ocean.

[0433] like Fig.13 Prior Art C 13 Amplitude image, Fig.14 The present invention A 13 Amplitude image, Fig.15 The present invention Amplitude images and Fig.16 The present invention The magnitude image of ;

[0434] Analyzing these four amplitude images, the present invention can observe that they have similar patterns in texture, with only slight differences in the distribution of intensity values: low wind speed areas show lower amplitudes, while wind turbines have higher amplitudes, which is in line with the expectations of the present invention.

[0435] like Fig.17 Prior Art C13 Phase image, Fig.18 The present invention A 13 Phase image, Fig.19 The present invention Phase image and Fig. 20 The present invention Phase image of .

[0436] However, among the four phase images, only and C 13 have similar characteristics, and and A 13 With C 13 Different significant Doppler features. For example, near a wind farm, such as Figure 17-Figure 20 As shown in the black box in the figure, two co-polarized frequency difference signals with opposite directions can be observed. Different images are also shown, indicating the direction of the radar observation.

[0437] For the convenience of comparison, the present invention provides the NRCS and Doppler shift of VV polarization derived from DCA, as well as the co-polarization frequency difference; including the Doppler shift of HH, the Doppler shift of VV, and the difference in co-polarization Doppler shift (HH-VV). Among them, the co-polarization frequency difference is calculated using the DCA method, and the grid size is 2.0km×2.0km. It can be seen from the NRCS and Doppler shift of VV polarization derived from DCA that both HH and VV polarizations show negative Doppler shifts as a whole, and their significant differences are mainly concentrated in the wind farm area. Outside the wind farm, the DC of HH and VV polarizations is almost the same, and the frequency difference of the co-polarization signals is negligible. Near the wind farm, the negative co-polarization frequency difference is in the opposite direction to the positive co-polarization frequency difference. Despite the opposite directions, these co-polarization frequency differences are symmetrical. According to calculations, a co-polarization frequency difference of 90Hz corresponds to a phase difference of 0.2 radians, which is consistent with Fig.19 This example shows that the polarimetric autocorrelation matrix can provide more detailed Doppler information while maintaining polarimetric properties.

[0438] Figure 21-23 It is shown in Figure 2 that in a certain sea area, cross polarization has a significant proportion of high-order terms. In addition, the Z value of HH polarization is larger than that of VV polarization, indicating that the contribution of Bragg waves in HH polarization is weaker (compared to larger and faster waves) relative to VV polarization. In addition, this also means that the Doppler shift strongly depends on the strength of the tilt modulation. For With |ρ HH-VV The ratio of Fig.24 |ρ HH-VV |Co-polarization correlation coefficient plot, Fig.25 Co-polarization correlation coefficient plot, Fig.26 Comparison of co-polarization correlation coefficients between )

[0439] Since the area is an open sea, in addition to marine structures such as ship targets and wind turbines, The value of is less than 1 for all features, which is consistent with Fig.12 Consistent.

[0440] It can be seen from the above embodiments that in the prior art, polarization and Doppler techniques have been proven to be effective for characterizing targets or scatterers, but the study of joint analysis of polarization and Doppler applied to mobile targets has been neglected. In particular, there is a lack of quantitative tools to understand the differences in Doppler characteristics between different polarization channels. To address this challenge, the present invention introduces a novel polarization autocorrelation matrix, labeled [A], which is derived from the polarization covariance matrix through an autocorrelation operation. Compared with second-order statistical matrices such as [C] and [T], the newly proposed [A] matrix can capture hidden Doppler information in the frequency domain and encode it in the phase due to the use of high-order statistical methods. The new matrix can simultaneously extract polarization and Doppler information in PolSAR data through matrix analysis, which is a significant improvement compared to traditional polarization analysis methods that only focus on target polarization scattering and relative polarization phase. The proposed descriptor can contribute to a deeper understanding of target characteristics and scattering phenomena, and provides unique insights into target characteristics and scattering mechanisms in SAR polarization.

[0441] Similar to the covariance and correlation matrices, the polarization autocorrelation matrix [A] is a 3×3 positive semidefinite Hermitian matrix. It can be decomposed into a Doppler-related matrix [A Dop ] and a matrix associated with the covariance [A Cov] , they are both 3×3 positive semidefinite Hermitian matrices. [A] and [A Cov ] are all matrices of rank 3, and [A Dop ] is a matrix of rank 1. Dop ] is related to the ensemble mean of the radar backscatter coefficients (the zeroth order term of the NRCS), while [A Dop The phase of the non-diagonal elements in ] represents the Doppler frequency shift difference between different polarizations (referred to as polarization Doppler difference in the present invention). Dop ] compared to the matrix [A Cov ] computes the covariance of the matrix [A]. Cov The amplitude of the elements in [A] is determined by the standard deviation of the backscattering coefficient (corresponding to the higher-order terms of the NRCS), while CovThe phase of ] is related to the degree of deviation from the polarization Doppler difference and the spectrum broadening. In addition, a set of new features and metrics are derived from the polarization autocorrelation matrix, such as the scale factor q, the polarization Doppler difference ΔF, the ratio of the high-order term to the zero-order term Z, and the polarization correlation coefficient of the backscattering All of these help to understand and characterize the scattering mechanisms.

[0442] The present invention uses two fully polarized RADARSAT-2 images for testing, and the new descriptors and new parameters prove that in the SAR echo, the Doppler shift or DC component is sensitive to the radial velocity of the target, which is traditionally understood, but is also sensitive to the radar polarization, which is an aspect that has not been fully explored in the past. In addition, the present invention's observations show that the polarization autocorrelation matrix enhances the sensitivity to radial velocity changes between different polarizations, effectively distinguishing between stationary and moving targets, and distinguishing between different scattering mechanisms. Especially for ocean scenes, the present invention's data found that the Doppler difference between HH polarization and VV polarization caused by sea surface motion can be as high as 100 Hz.

[0443] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered by the protection scope of the present invention.

Claims

1. A SAR polarimetric measurement method based on polarimetric autocorrelation matrix, characterized in that: The method includes: S1, obtain the polarization autocorrelation matrix from the polarization covariance matrix through autocorrelation operation; S2, captures the hidden Doppler information in the frequency domain by acquiring the polarization autocorrelation matrix and encodes the Doppler information into the phase using high-order statistical methods; S3, based on the polarization autocorrelation matrix, matrix analysis is used to simultaneously extract polarization and Doppler features from PolSAR data; In step S1, a polarization autocorrelation matrix is ​​obtained from a polarization covariance matrix through an autocorrelation operation, including: The first step is to define the polarization autocorrelation matrix; The second step is to obtain the elements of the polarization autocorrelation matrix; The third step is to obtain the polarization autocorrelation matrix; In the first step, the polarization autocorrelation matrix is ​​defined, including: In the PolSAR system, the Sinclair scattering matrix It is expressed as: ； In the formula, is the Sinclair scattering matrix, is the scattering coefficient from horizontal polarization to horizontal polarization, is the scattering coefficient from horizontal polarization to vertical polarization, is the scattering coefficient from vertical polarization to horizontal polarization, is the scattering coefficient from vertical polarization to vertical polarization; Under the condition of single-station SAR, formula (1) satisfies the reciprocity theorem, that is, ; Covariance matrix By using the Sinclair scattering matrix [ S ]'s scattering vector Calculate the outer product with its conjugate transpose and you get; （2）； In the formula, is the covariance matrix, represents the conjugate transpose, is the scattering vector, is the complex conjugate transpose of the scattering vector; Calculate the covariance matrix The autocorrelation function of each element in is Represents the autocorrelation operation performed on each element in the matrix, the polarized autocorrelation matrix [ A ] is defined as: （3）； In the formula, A and Both represent the polarization autocorrelation matrix, is the autocorrelation operation; In the second step, the elements of the polarization autocorrelation matrix are obtained, including: By rearranging formula (3), we obtain: （4）； In the formula, represents the elements in the polarization autocorrelation matrix, ; is a real number, are complex conjugates of each other; ； In the formula, and Respectively represent autocorrelation operation and expectation; in PolSAR images, is the number of rows in the azimuth direction, is the interval between two samples along the azimuth direction, represents the conjugate transpose; A 11 and A 13 Take this as an example to derive the formula in detail; A 11 Rewritten as: （5）； A 11 It is divided into the following two parts: （6）； in: ； Similarly, A 13 Decomposed into the sum of two parts, the expression is: ； （7）； In the formula, represents the covariance, A 11 and A 13 are the elements in the polarization autocorrelation matrix; Represent the first element of the polarization autocorrelation matrix and its corresponding Dop term and Cov respectively; in: ； By reference A 11 and A 13 , build the matrix [ A The expression for all elements in ]: （8）； In the formula, is the ensemble average of the backscatter coefficients of the three polarizations, is the standard deviation of the backscatter coefficients for the three polarizations, are the Doppler centroid (DC) of HH, VV and HV polarizations, yes and The correlation coefficient between and The definition of same; is the natural logarithm, is an imaginary unit, is the pulse repetition period; In the third step, the polarization autocorrelation matrix is ​​obtained, including: According to formula (8), each element of the polarization autocorrelation matrix is ​​divided into a Dop term and a Cov term; therefore, the polarization autocorrelation matrix is ​​written as the Doppler correlation matrix [A Dop ] and the covariance correlation matrix [A Cov ], as follows: （9）； in, ； in, are the elements in the Doppler correlation matrix, is the element in the covariance correlation matrix, ; Each element of the polarization autocorrelation matrix is ​​divided into a Dop term and a Cov term; therefore, the polarization autocorrelation matrix is ​​written as the Doppler correlation matrix [A Dop ] and the covariance correlation matrix [A Cov ] and; , ,and The degree of deviation of Doppler shift and spectrum broadening indicates the phase value; In step S2, the hidden Doppler information is captured in the frequency domain by the acquired polarization autocorrelation matrix, and the Doppler information is encoded into the phase using a high-order statistical method, including: Step 1, generate a 3×3 matrix from a 2×2 Sinclair scattering matrix S: ； Step 2: Calculate in a sliding window , ,and The set average of; set k to 1; set the sliding window size to a common value; Step 3: In the sliding window of the same size, use formula (3) to Apply an autocorrelation operation to each element in to obtain the polarization autocorrelation matrix [A]; Step 4: Calculate the polarization autocorrelation matrix using formula (10): A ] of the Doppler term [ A Dop ]; Step 5: Calculate the polarization autocorrelation matrix [ A ]'s higher-order covariance terms[ A Cov ]; In step 4, the Doppler correlation matrix [ A Dop ] has a rank of 1 and is represented as the product of two one-dimensional vectors: （10）； Among them, k Dop is defined as: ； Covariance correlation matrix [A Cov ] by getting the polarization autocorrelation matrix [ A ] minus the Doppler correlation matrix [ A Dop ] is calculated, and the expression is: 。 2. The SAR polarimetric measurement method based on polarimetric autocorrelation matrix according to claim 1, characterized in that: In step 4, in the Doppler correlation matrix [ A Dop ]middle, A Dop The phase of the off-diagonal elements in is controlled by the DC frequencies of the HH, VV, and HV polarizations. The DC frequencies of HH, VV, and HV / VH are: and ; The Doppler shift or DC frequency of the radar echo is affected by multiple contributing sources, including platform motion, radar attitude, and moving targets on the surface, and is expressed as: （12）； in, f geom is the Doppler term resulting from satellite geometry and attitude, f elec is the Doppler term generated by the antenna pointing; Δ f It is the Doppler residual caused by non-geophysical terms and unknown biases; f phys is the Doppler term generated by the ground moving target, f phys Commonly used to calculate the radial Doppler velocity of ground moving targets v d ; （13）； in, k e is the electromagnetic wave number of the radar, θ i is the radar incidence angle; v d A positive value indicates that the target is approaching the radar direction, and a negative value indicates that the target is moving away from the radar direction; For spaceborne SAR systems, different polarization f geom , f elec , and Δ f are the same, and f phys It varies for different polarizations; therefore, this method obtains the Doppler velocity difference of moving targets under different polarizations. , and ; In step 5, the covariance correlation matrix [A Cov The amplitude of the elements in ] is calculated using the radar backscatter coefficient , ,and The standard deviation of represents a higher-order term, using , and The degree of deviation of Doppler shift and spectrum broadening represents the phase value.

3. The SAR polarimetric measurement method based on polarimetric autocorrelation matrix according to claim 1, characterized in that: In step S3, based on the polarization autocorrelation matrix, matrix analysis is used to simultaneously extract polarization and Doppler features from the PolSAR data, including: Step a, SPAN(q) The scale factor for the covariance matrix [ C ], the total backscattered power SPAN is defined as the sum of the three diagonal elements; the polarization autocorrelation matrix [ A ]of SPAN Defined as: （14）； in, SPAN C and SPAN A Represent the covariance matrix [ C ] and the autocorrelation matrix [ A ]of SPAN ; , i=1, 2, 3, are the eigenvalues ​​of the covariance matrix [C] and the polarization autocorrelation matrix [A] respectively; Tr ([ C ]) is the covariance matrix[ C ]of SPAN , Tr ([ A ] is the polarization autocorrelation matrix [ A ]of SPAN ; Polarization autocorrelation matrix[ A ]and[ C ] has a quadratic relationship in magnitude, making [ A ]of SPAN is smaller than the covariance matrix [ C ]of SPAN ; To enhance readability and avoid [ A ] is too small, use the scaling factor q Amplified polarization autocorrelation matrix[ A ] value, the scaling factor q The calculation formula is as follows: （15）； in, q is the scale factor Tr ( ) means finding the trace of the matrix; After scaling, the autocorrelation matrix is ​​recorded as ; Scaled matrix A Dop and A Cov They are and ; Step b, polarization Doppler difference ΔF; The Doppler differences between different polarizations are obtained from the matrix A Dop Calculated from the off-diagonal elements of （16）； Among them, Δ F ij It is a polarization channel i and polarization channels j The Doppler frequency shift difference between them, Arg() represents the phase of the complex number in the brackets; （17）； Step c, the ratio Z of the high-order and zero-order terms of the backscatter coefficient; Ratio of the high-order to zero-order terms of the backscatter coefficient Z Using the Matrix A Dop and A Cov Calculation of diagonal elements of ; （18）； Where the denominator of Z represents the average backscattering of the non-sloped surface, while the numerator of Z corresponds to the change in NRCS caused by the surface slope; Step d, polarization correlation coefficient of radar backscattering high-order terms ; [ A Cov ], ( i , j = 1 (HH), 2 (HV / VH), 3 (VV), and i ≠ j ) is the polarization correlation coefficient Standard deviation of radar backscatter coefficient The product of The calculation is as follows: （19）； in, represents the polarization correlation coefficient of the high-order terms of the radar echo, is less than 1, which measures the degree of depolarization caused by high-order scattering. The phase is equivalent to The phase of the zero-order term is A Dop Derived from the matrix, the corresponding correlation coefficient of the zero-order term is 1; represents the polarization correlation coefficient Standard deviation of radar backscatter coefficient The product of ; represents conjugate transpose; is the ensemble average of the backscatter coefficients of the three polarizations, is the standard deviation of the backscatter coefficients for the three polarizations, are the Doppler centroid (DC) of HH, VV and HV polarizations, Defined as and The correlation coefficient between and The definition of same.

4. A SAR polarization measurement system based on a polarization autocorrelation matrix, the system being implemented by the SAR polarization measurement method based on a polarization autocorrelation matrix according to any one of claims 1 to 3, characterized in that: The system includes: A polarization autocorrelation matrix acquisition module is used to acquire a polarization autocorrelation matrix from a polarization covariance matrix through an autocorrelation operation; Phase encoding module, used to capture hidden Doppler information in the frequency domain through the acquired polarization autocorrelation matrix and encode the Doppler information into the phase using high-order statistical methods; The polarization and Doppler feature extraction module is used to simultaneously extract polarization and Doppler features from PolSAR data using matrix analysis based on the polarization autocorrelation matrix.

Images