Polarized SAR target enhanced extraction method in complex ground scene

Abstract

The invention discloses a polarized SAR (synthetic aperture radar) target enhanced extraction method in a complex ground scene, which belongs to the technical field of radars and comprises the following steps: S1, polarization distortion full-scene self-correction; s2, generating an equivalent sub-view image; s3, constructing a polarization statistical enhancement test model; s4, performing enhanced likelihood testing based on sub-view data characteristics; and S5, CFAR enhancement detection based on composite chi-square distribution. According to the method, the difference between the target and the background is effectively and quantitatively described, the signal-to-clutter ratio of the target and the detection robustness are greatly enhanced, and the requirements of automatic target enhancement extraction and batch application of large-scale SAR images are met.

Description

Technical Field

[0001] The present invention relates to the field of radar technology, and in particular to a polarimetric SAR target enhancement extraction method in a complex ground scene. Background Art

[0002] As an active microwave imaging system, synthetic aperture radar (SAR) has been widely used in military and national economic fields such as military reconnaissance, environmental monitoring, and land resource management. It has the advantages of all-day, all-weather, long-distance and wide mapping band, and can greatly improve the rapid information acquisition capability of radar targets in land and sea environments over long distances and over a large range of complex backgrounds.

[0003] SAR image target detection is the basis of SAR image interpretation. The SAR image target detection module is a binary detection processing process, that is, judging whether the target exists or not. Traditional detection and recognition algorithms have great limitations in the universality of target feature extraction. Each feature extraction method is only designed for a certain scene or a certain target. For example, the CFAR detection method based on different background clutter statistical distribution models provides the best CFAR detection for a variety of scene images such as uniform, generally non-uniform, and extremely non-uniform. In addition, polarization information is introduced, and through polarization target decomposition technology, an association model is established with the target polarization scattering characteristics and typical targets. On the other hand, deep learning methods clearly highlight the importance of feature learning and can solve the difficulties of feature extraction and recognition decision design in traditional recognition algorithms. However, the main idea of ​​SAR target detection is to use the feature differences between the target and the background clutter to extract the target in complex background clutter.

[0004] The complex background under complex background conditions refers to the target existing in the bright background, the background edge, the coherent speckle noise and other areas where the target is easily submerged by the background clutter. The existence of the target in these background clutter areas is often a difficulty and bottleneck in detection, resulting in the existing algorithms often being difficult to achieve satisfactory results in terms of detection rate, false alarm rate, signal-to-clutter ratio gain, background suppression rate and other indicators. The difficulties are mainly manifested in:

[0005] a. SAR targets usually do not have obvious shapes and textures, which is obviously not conducive to improving the algorithm detection rate;

[0006] b. The brightness of complex background areas such as highlighted background areas, background edges, and coherent speckle noise in SAR images is likely to drown out real targets, which may cause false alarms and reduce the detection rate of the algorithm;

[0007] c. Polarimetric SAR images contain rich polarization information of targets, which is beneficial to target detection and recognition. However, polarimetric SAR images have polarization crosstalk and amplitude imbalance between polarization channels, and polarization data correction is required to correctly interpret ground objects.

[0008] In general, when the target is in a complex background environment, the target may even be submerged by the complex background, and the signal-to-clutter ratio is lower, which greatly increases the difficulty of SAR target detection. Therefore, in-depth research on SAR target enhancement technology under complex background conditions has extremely important theoretical significance and practical engineering application value for improving SAR target detection performance. To this end, a polarimetric SAR target enhancement extraction method under complex ground scenes is proposed. Summary of the invention

[0009] The technical problem to be solved by the present invention is: how to solve the problem that the target is in a complex background environment, the target may even be submerged by the complex background, the signal-to-clutter ratio is lower, and the difficulty of SAR target detection is greatly increased. A polarization SAR target enhancement extraction method in a complex ground scene is provided.

[0010] The present invention solves the above technical problems through the following technical solutions, and the present invention comprises the following steps:

[0011] S1: Full-scenario self-correction of polarization distortion

[0012] The polarization distortion parameters of the acquired full-polarization SAR image are estimated and corrected using the natural scene distribution targets and artificial targets in the full scene to obtain the polarization-corrected SAR image.

[0013] S2: Equivalent subview image generation

[0014] The polarization-corrected SAR image is equivalently generated into two low-resolution sub-view images of the front view and the rear view through SAR image spectrum segmentation;

[0015] S3: Constructing a polarization statistical enhancement test model

[0016] Based on the distribution of polarization covariance matrix, a polarization statistical enhancement detection model for target and background clutter is constructed.

[0017] S4: Enhanced likelihood test based on sub-view data characteristics

[0018] Considering the incoherence of background clutter in the subview image, the subview image after target enhancement is obtained by using subview image operation, and the enhanced likelihood detection value of the image target is calculated by combining the background and target local window size setting and sliding window processing;

[0019] S5: CFAR enhanced detection based on composite chi-square distribution

[0020] Based on the approximate distribution representation of enhanced likelihood detection values, constant false alarm rate detection is used to extract targets from complex background clutter.

[0021] Furthermore, in step S1, the matrix form of the full polarization SAR image obtained by the polarization SAR system is as follows:

[0022]

[0023] in, is the undistorted polarization scattering matrix, S HH is the undistorted HH polarization channel data, S HV is the undistorted HV polarization channel data, S VH is the undistorted VH polarization channel data, S VV is the distortion-free VV polarization channel data, R e and T e are the distortion matrices introduced by reception and transmission respectively, N is the system noise matrix, f1, f2, f3, f4 are the amplitude and phase imbalance terms of the transmitting and receiving channels respectively, δ1, δ2, δ3, δ4 are the crosstalk factors between polarization channels.

[0024] Furthermore, in step S1, the specific processing process is as follows:

[0025] S11: Ignoring the influence of absolute radiation calibration and noise term, let the unbalanced parameter between polarization channels be Polarization channel imbalance parameters The matrix form of the fully polarized SAR image is rewritten into a vector form as follows:

[0026]

[0027] Among them, O HH is the observed HH polarization channel data, O HV is the observed HV polarization channel data, O VH is the observed VH polarization channel data, O VV is the observed VV polarization channel data, β is the imbalance parameter between polarization channels, and γ is the imbalance parameter within the polarization channel;

[0028] S12: Selecting a ground object whose total power SPAN value is lower than 70% of the maximum total power SPAN value of the entire scene image as a natural scene distribution target, and calculating the imbalance parameter β between polarization channels according to the formula;

[0029] S13: Select the target with single scattering and dihedral scattering characteristics in the image as the artificial target, the HH and VV polarization observation values ​​of the artificial target are equal, and calculate the imbalance parameter γ in the polarization channel according to the formula;

[0030] S14: Substitute the estimated imbalance parameter β between polarization channels and the imbalance parameter γ within polarization channels into the vector expression of the full polarization SAR image in step S11 to obtain a polarization-corrected SAR image.

[0031] Furthermore, in step S12, the calculation formula of the imbalance parameter β between polarization channels is:

[0032]

[0033] Among them, O VH,nat and O HV,nat are the VH and HV polarization statistics of naturally distributed objects, respectively.

[0034] Furthermore, in step S13, the calculation formula of the imbalance parameter γ in the polarization channel is:

[0035]

[0036] Among them, O HH,tri and O VV,tri are the HH and VV polarization observation values ​​of artificial targets, respectively.

[0037] Furthermore, in step S2, the specific processing process is as follows:

[0038] S21: The entire image spectrum is divided into two sub-video spectra of equal width and interval, namely:

[0039]

[0040] Among them, f ci represents the center frequency of subview image i;

[0041] S22: Obtain the equivalent spatial domain signal of the sub-view image as:

[0042]

[0043] S23: All sub-view spectra are converted back to the time domain through IFFT to obtain two low-resolution sub-view images of the front view and the rear view.

[0044] Furthermore, in step S3, the specific processing process is as follows:

[0045] S31: Incoherently average the n single-view covariance matrices to obtain the n-view covariance matrix W:

[0046]

[0047] Where n is the number of multi-views, S(k) is the kth single-view sampling vector;

[0048] S32: Let A = n W, then the covariance matrix A has the form of complex Wishart distribution:

[0049]

[0050] Where Tr(·) is the trace of the matrix, Σ=E(A), K(l,p)=π p(p-1) / 2 Γ(l)…Γ(l-p+1), Γ(n) is the Gamma function, and the parameter p is the dimension of the vector S(k);

[0051] S33: Based on statistical hypothesis testing, the polarization statistical enhancement test model R is defined as follows:

[0052]

[0053] Furthermore, in step S33, the statistical hypothesis testing process is as follows:

[0054] S331: Obtain the distribution function of the n-view covariance matrix W:

[0055]

[0056] If the polarization matrices X and Y are independent and both obey the complex Wishart distribution, that is, X∈W(p, k, Σ x ), Y∈W(p,l,Σ y ), then their sum X+Y also obeys the complex Wishart distribution, that is, (X+Y)∈W(p,k+l,Σ);

[0057] S332: When the data of the target window and the background window are different, Σ x ≠Σ y When:

[0058] S333: When the data of the target window and the background window are the same, Σ x =Σ y =Σ, then:

[0059]

[0060] Furthermore, in step S4, the specific processing process is as follows:

[0061] S41: The target data is the average of the sum of the two subview images in step S2, that is:

[0062]

[0063] The background data still uses full-polarization full-aperture image data;

[0064] S42: respectively set the target statistical area window size X wz and background statistics area window size Y wz , where X wz <Y wz ;

[0065] S43: By performing sliding window processing in the image, the logarithmic polarization statistical enhancement test value of the image target, that is, the enhanced likelihood detection value, is calculated. The constant term is removed and the logarithmic polarization statistical enhancement test value is simplified as follows:

[0066] -lnR=(k+l)ln|X+Y|-kln|X|-lln|Y|.

[0067] Furthermore, in step S5, the specific processing process is as follows:

[0068] S51: The probability density distribution function of the enhanced likelihood detection value is derived through the third-order Bernoulli polynomial and is approximately:

[0069] P{-lnR}≈P{χ 2 (p 2 )}+w·[P{χ 2 (p 2 +4)}-P{χ 2 (p 2 )}]

[0070] Among them, χ 2 is the chi-square distribution function, w is the weight factor;

[0071] S52: Under the condition of known enhanced likelihood detection value probability distribution function, using constant false alarm rate method and given constant false alarm probability P fa , to obtain the threshold for target enhanced detection.

[0072] Furthermore, in step S52, the threshold of target enhanced detection is obtained by solving an equation according to the CFAR detection principle, and the equation is as follows:

[0073]

[0074] Among them, T is the threshold of target enhanced detection.

[0075] Compared with the prior art, the present invention has the following advantages: the polarimetric SAR target enhancement extraction method in complex ground scenes effectively and quantitatively describes the difference between the target and the background, greatly enhances the target signal-to-clutter ratio and detection robustness, and meets the needs of automatic target enhancement extraction and batch application of large-scale SAR images. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1It is a flow chart of the polarimetric SAR target enhancement extraction method in a complex ground scene in the first embodiment of the present invention;

[0077] Figure 2 is the original SAR image (HH polarization) in the second embodiment of the present invention;

[0078] Figure 3 is the original SAR image (HV polarization) in the second embodiment of the present invention;

[0079] Figure 4 is the original SAR image (VV polarization) in the second embodiment of the present invention;

[0080] Figure 5 is the polarization-corrected SAR image in the second embodiment of the present invention;

[0081] Figure 6 This is a schematic diagram of the statistical window and range in the second embodiment of the present invention;

[0082] Figure 7 This is a diagram of enhanced likelihood detection results in the second embodiment of the present invention. DETAILED DESCRIPTION

[0083] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiment.

[0084] Embodiment 1

[0085] The present embodiment provides a technical solution: a polarimetric SAR target enhancement extraction method in a complex ground scene, firstly, the polarimetric distortion parameters of the acquired SAR image are estimated and corrected by using the naturally distributed ground objects and a small number of artificial targets in the whole scene; the sub-view images before and after polarization are equivalently generated by spectrum segmentation of the SAR image; then, based on the distribution of the polarimetric covariance matrix, a polarimetric statistical enhancement detection model of the target and background clutter is constructed; then, considering the incoherence of the background clutter in the sub-view image, the sub-view image after the target is enhanced is obtained by sub-view image operation, and the enhanced likelihood detection value of the image target is calculated in combination with the background and target local window size setting and sliding window processing; finally, based on the approximate distribution representation of the enhanced likelihood detection value, the constant false alarm rate detection is used to realize the extraction of the target from the complex background clutter.

[0086] like Figure 1 As shown, the polarimetric SAR target enhancement extraction method in complex ground scenes of the present invention specifically includes the following steps:

[0087] Step 1: Full-scene self-correction of polarization distortion

[0088] Based on the obtained fully polarized SAR image data to be processed, the polarization distortion parameters are quickly estimated through the natural scene distribution targets and artificial targets of the imaging scene itself, including the following:

[0089] The original 2*2 polarization scattering matrix observed by the polarimetric SAR system for:

[0090]

[0091] in, is the undistorted polarization scattering matrix, S HH is the distortion-free HH polarization channel data, S HV is the undistorted HV polarization channel data, S VH is the undistorted VH polarization channel data, S VV is the distortion-free VV polarization channel data, R e and T e are the distortion matrices introduced by reception and transmission respectively, N is the system noise matrix, f1, f2, f3, f4 are the amplitude and phase imbalance terms of the transmit and receive channels respectively, and the expected value is 1. The crosstalk terms δ1, δ2, δ3, δ4 are the crosstalk factors between polarization channels, which should be small enough compared with 1, δ1 = δ2 = δ3 = δ4 = 0, ignoring the influence of absolute radiation calibration and noise terms, let the imbalance parameters between polarization channels be Polarization channel imbalance parameters Formula (1) is written in vector form as:

[0092]

[0093] Among them, O HH is the observed HH polarization channel data, O HV is the observed HV polarization channel data, O VH is the observed VH polarization channel data, O VV is the observed VV polarization channel data;

[0094] Then, the statistical ratio of the image cross-polarization HV and VH is directly obtained, and the imbalance parameter β between polarization channels is obtained by formula (2):

[0095]

[0096] Among them, O VH,nat and O HV,nat are the VH and HV polarization statistics of naturally distributed objects, respectively;

[0097] Let S of the ground artificial target HH =S VV , then the imbalance parameter γ in the polarization channel can be obtained:

[0098]

[0099] Among them, O HH,tri and O VV,tri are the HH and VV polarization observation values ​​of ground artificial targets respectively.

[0100] Step 2: Equivalent subview image generation

[0101] In the SAR imaging process, the full-resolution SAR image of the target is obtained by synthesizing low-resolution echoes from many different oblique angles. Therefore, a SAR image target does not just correspond to a single radar line of sight direction, but to a series of oblique directions in azimuth. The movement of the radar platform relative to the ground target will cause the phase of the echo signal to change over time, thereby causing the change in the instantaneous frequency of the echo, forming a Doppler shift effect. The Doppler frequency f of the echo signal d The relationship between (t) and the oblique angle θ (i.e. the angle between the radar line of sight and the positive side view direction) is:

[0102]

[0103] Where v is the speed of the radar platform, R a is the distance from the radar to the ground target, and λ is the radar wavelength. However, what is often obtained is the SAR data after imaging, which is first converted to the azimuth\range spectrum domain through one-dimensional FFT. For a point target with a distance r0 from the SAR, its spatial domain response is expressed as:

[0104]

[0105] Where B is the signal bandwidth, t0 = 2r0 / c, is a constant. The Fourier transform X(f) of x(t) is:

[0106]

[0107] in,

[0108] Then the entire image spectrum is divided into two sub-video spectra of equal width and equal interval, namely:

[0109]

[0110] Among them, f ci represents the center frequency of the sub-view image i. Therefore, the equivalent spatial domain signal of the sub-view image can be written as:

[0111]

[0112] Finally, all sub-view spectra are converted back to the time domain through IFFT to obtain two low-resolution sub-view images of the front view and the rear view.

[0113] Step 3: Constructing a polarization statistical enhancement test model

[0114] By incoherently averaging n single-view covariance matrices, we can get the n-view covariance matrix W:

[0115]

[0116] Where n is the number of multi-views, and S(k) is the kth single-view sampling vector.

[0117] Let A = n W, then the matrix A has the form of complex Wishart distribution:

[0118]

[0119] Where Tr(·) is the trace of the matrix, Σ=E(A), K(l,p)=π p(p-1) / 2 Γ(l)…Γ(l-p+1), Γ(n) is the Gamma function, parameter p is the dimension of the vector S(k), for the fully polarized SAR reciprocal medium, p=3; for the dual-polarized SAR reciprocal medium, p=2.

[0120] In this way, it is easy to derive the distribution function of the n-view covariance matrix W:

[0121]

[0122] If the polarization matrices X and Y are independent and both obey the complex Wishart distribution, that is, X∈W(p,k,Σ x ), Y∈W(p,l,Σ y ), then their sum X+Y also obeys the complex Wishart distribution, that is, (X+Y)∈W(p,k+l,∑).

[0123] When the data of the target window and the background window are different, Σ x ≠Σ y , then:

[0124]

[0125] When the data of the target window and the background window are the same, Σ x =Σ y =Σ, then:

[0126]

[0127] Therefore, based on the above two statistical hypothesis tests, the polarization statistical enhancement test model can be defined as:

[0128]

[0129] Step 4: Enhanced likelihood test based on sub-view data characteristics

[0130] The value range of R is between [1, +∞). When the target and background data are completely different, the two polarization matrices X and Y can be considered independent of each other. According to the principle of probability, p(X,Y)=p(X)p(Y), then R is equal to 1. On the contrary, when the target and background data are closer, p(X,Y)<p(X)p(Y), then the value of R is larger.

[0131] Since the correlation of the background clutter distribution targets in different subview images is poor, the target data is averaged by the sum of the two left and right subview images with equal spectral width generated in step 2, that is:

[0132]

[0133] The background data still uses full-polarization full-aperture image data. Then, the target statistical area window size X is given respectively. wz and background statistics area window size Y wz ,in

[0134] X wz <Y wz (17)

[0135] In this way, the logarithmic polarization statistical enhancement test value of the image target is calculated by sliding window processing in the image. Removing the constant term, its logarithmic polarization statistical enhancement test value can be simplified to:

[0136] -lnR=(k+l)ln|X+Y|-kln|X|-lln|Y| (18)

[0137] Step 5: CFAR enhanced detection based on compound chi-square distribution

[0138] Since a certain order moment of the asymptotic expansion of the distribution of a random variable can generally be expressed as a specific function form, the probability density of -lnR can be derived through the third-order Bernoulli polynomial to be approximated as:

[0139] P{-lnR}≈P{χ 2 (p 2 )}+w·[P{χ 2 (p 2 +4)}-P{χ 2 (p 2 )}](19)

[0140] Among them, χ 2 is the chi-square distribution function, and w is the weight factor.

[0141] The -lnR value represents the degree of difference of the corresponding features.

[0142] Finally, under the condition that the probability density distribution function form of -lnR is known, the constant false alarm rate method and the given constant false alarm probability P are used. fa , to obtain the threshold for target enhanced detection.

[0143] Embodiment 2

[0144] The present invention uses the method of obtaining a SAR image by using the naturally distributed objects and a small number of artificial targets in the whole scene to estimate and correct the polarization crosstalk and polarization channel imbalance parameters, and generates polarization left and right sub-view images by equivalently dividing the SAR image spectrum, and then constructs a polarization statistical enhancement detection model of the target and background clutter on the basis of the polarization covariance matrix distribution, which effectively and quantitatively describes the difference between the target and the background. Next, considering the incoherence of the background clutter in the sub-view image, the sub-view image after the target is enhanced is obtained by sub-view image calculation, and the enhanced likelihood detection value of the image target is calculated by combining the background and target local window size setting and sliding window processing, which greatly enhances the target signal-to-clutter ratio and detection robustness. Finally, based on the approximate distribution representation of the enhanced likelihood detection value, the detection of the constant false alarm rate CFAR and the like can realize the extraction of the target from the complex background clutter, meet the needs of large-scale SAR image automatic target enhancement extraction and batch application, and finally develop programming software on an ordinary computer based on the technology of the present invention.

[0145] Combine the following Figure 2-Figure 7 The present invention is further described.

[0146] First, obtain the fully polarized SAR image according to step 1. Figure 2-Figure 4 They are the SAR images of the original HH polarization, HV polarization and VV polarization channels, respectively. The image size is 4000 pixels * 8192 pixels, the radar wavelength λ is 0.03125m, and the image resolution is 0.5m.

[0147] More specifically, the objects below 70% of the maximum value of the total power SPAN image are taken as natural scene distribution targets to calculate the VH and HV polarization statistics of the natural distribution objects, and the imbalance parameter β between polarization channels is calculated according to equation (3) in step 1.

[0148] As a more specific example, ground artificial targets are selected from power towers and electric poles in the image, which generally present single scattering and dihedral scattering characteristics. In this way, the HH and VV polarization observation values ​​of the ground artificial targets satisfy S HH =S VV, then according to formula (4), the imbalance parameter γ in the polarization channel can be obtained. Finally, according to the estimated imbalance parameters between polarization channels and within the channel, substitute them into formula (2) to obtain the polarization-corrected SAR image, as follows: Figure 5 shown.

[0149] According to step 2, the HH, HV, VH, and VV image spectra are divided into two sub-video spectra with equal width and interval, namely:

[0150]

[0151] Among them, f ci represents the center frequency of the sub-view image i. Therefore, the equivalent spatial domain signal of the sub-view image can be written as:

[0152]

[0153] All sub-view spectra of HH, HV, VH, and VV are reconverted to the time domain through IFFT to obtain two sets of low-resolution full-polarization sub-view images for the forward view and the rear view.

[0154] According to step 3, the covariance matrix A can be obtained by incoherently averaging the single-view covariance matrix. Then the matrix A has the form of complex Wishart distribution:

[0155]

[0156] Where Tr(·) is the trace of the matrix, Σ=E(A), K(l,p)=π p(p-1) / 2 Γ(l)…Γ(l-p+1), Γ(n) is the Gamma function, parameter p is the dimension of vector X, for fully polarized SAR reciprocal media, p=3;

[0157] Based on the statistical hypothesis test in step 3, the polarization statistical enhancement test model R can be calculated as:

[0158]

[0159] According to step 4, since the correlation of the background clutter distribution target in different subview images is poor, the target data adopts the sum average of the two left and right subviews with equal spectral width generated in step 2, that is:

[0160]

[0161] The background data still uses full-aperture image data, namely:

[0162] Y=x(t) (24)

[0163] Then, according to the target of interest, this embodiment takes the tower target as an example, and gives the target statistical area window size X wzis 31, and the background statistics area window size Y wz is 43, the statistical window and range diagram are as follows Figure 6 Then, according to formula (18) (where k=l=1), the logarithmic polarization statistical enhancement test values ​​of the image targets are calculated in sequence through image sliding window processing.

[0164] Finally, according to step 5, let p = 3, calculate the probability density distribution function of the log-polarization statistical enhancement test value according to formula (19), and set the constant false alarm probability P fa =1.0e-5, according to the CFAR detection principle, solve the following equation to calculate the threshold T of enhanced detection and obtain the enhanced likelihood detection result, see Figure 7 .

[0165]

[0166] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A polarimetric SAR target enhancement extraction method in complex ground scenes, characterized in that: The following steps are involved: S1: Full-scenario self-correction of polarization distortion The polarization distortion parameters of the acquired full-polarization SAR image are estimated and corrected using the natural scene distribution targets and artificial targets in the full scene to obtain the polarization-corrected SAR image. S2: Equivalent subview image generation The polarization-corrected SAR image is equivalently generated into two low-resolution sub-view images of the front view and the rear view through SAR image spectrum segmentation; S3: Constructing a polarization statistical enhancement test model Based on the distribution of polarization covariance matrix, a polarization statistical enhancement detection model for target and background clutter is constructed. S4: Enhanced likelihood test based on sub-view data characteristics Considering the incoherence of background clutter in the subview image, the subview image after target enhancement is obtained by using subview image operation, and the enhanced likelihood detection value of the image target is calculated by combining the background and target local window size setting and sliding window processing; S5: CFAR enhanced detection based on composite chi-square distribution Based on the approximate distribution representation of enhanced likelihood detection values, constant false alarm rate detection is used to extract targets from complex background clutter.

2. The method for polarimetric SAR target enhancement extraction in a complex ground scene according to claim 1, characterized in that: In step S1, the matrix form of the full polarization SAR image obtained by the polarization SAR system is as follows: in, is the undistorted polarization scattering matrix, S HH is the undistorted HH polarization channel data, S HV is the undistorted HV polarization channel data, S VH is the undistorted VH polarization channel data, S VV is the distortion-free VV polarization channel data, R e and T e are the distortion matrices introduced by reception and transmission respectively, N is the system noise matrix, f1, f2, f3, f4 are the amplitude and phase imbalance terms of the transmitting and receiving channels respectively, δ1, δ2, δ3, δ4 are the crosstalk factors between polarization channels.

3. The method for polarimetric SAR target enhancement extraction in complex ground scenes according to claim 2, characterized in that: In step S1, the specific processing process is as follows: S11: Ignoring the influence of absolute radiation calibration and noise term, let the unbalanced parameter between polarization channels be Polarization channel imbalance parameters The matrix form of the fully polarized SAR image is rewritten into a vector form as follows: Among them, O HH is the observed HH polarization channel data, O HV is the observed HV polarization channel data, O VH is the observed VH polarization channel data, O VV is the observed VV polarization channel data, β is the imbalance parameter between polarization channels, and γ is the imbalance parameter within the polarization channel; S12: Selecting a ground object whose total power SPAN value is lower than 70% of the maximum total power SPAN value of the entire scene image as a natural scene distribution target, and calculating the imbalance parameter β between polarization channels according to the formula; S13: Select the target with single scattering and dihedral scattering characteristics in the image as the artificial target, the HH and VV polarization observation values ​​of the artificial target are equal, and calculate the imbalance parameter γ in the polarization channel according to the formula; S14: Substitute the estimated imbalance parameter β between polarization channels and the imbalance parameter γ within polarization channels into the vector expression of the full polarization SAR image in step S11 to obtain a polarization-corrected SAR image.

4. The method for polarimetric SAR target enhancement extraction in complex ground scenes according to claim 3, characterized in that: In step S12, the calculation formula of the imbalance parameter β between polarization channels is: Among them, O VH,nat and O HV,nat are the VH and HV polarization statistics of naturally distributed objects, respectively.

5. The method for polarimetric SAR target enhancement extraction in complex ground scenes according to claim 3, characterized in that: In step S13, the calculation formula of the imbalance parameter γ in the polarization channel is: Among them, O HH,tri and O VV,tri are the HH and VV polarization observation values ​​of artificial targets, respectively.

6. The method for polarimetric SAR target enhancement extraction in complex ground scenes according to claim 3, characterized in that: In step S2, the specific processing process is as follows: S21: The entire image spectrum is divided into two sub-video spectra of equal width and interval, namely: Among them, f ci represents the center frequency of subview image i; S22: Obtain the equivalent spatial domain signal of the sub-view image as: S23: All sub-view spectra are converted back to the time domain through IFFT to obtain two low-resolution sub-view images of the front view and the rear view.

7. The method for polarimetric SAR target enhancement extraction in complex ground scenes according to claim 6, characterized in that: In step S3, the specific processing process is as follows: S31: Incoherently average the n single-view covariance matrices to obtain the n-view covariance matrix W: Where n is the number of multi-views, S(k) is the kth single-view sampling vector; S32: Let A = n W, then the covariance matrix A has the form of complex Wishart distribution: Where Tr(·) is the trace of the matrix, Σ=E(A), K(l,p)=π p(p-1) / 2 Γ(l)…Γ(l-p+1), Γ(n) is the Gamma function, and the parameter p is the dimension of the vector S(k); S33: Based on statistical hypothesis testing, the polarization statistical enhancement test model R is defined as follows:

8. The method for polarimetric SAR target enhancement extraction in complex ground scenes according to claim 7, characterized in that: In step S33, the statistical hypothesis testing process is as follows: S331: Obtain the distribution function of the n-view covariance matrix W: If the polarization matrices X and Y are independent and both obey the complex Wishart distribution, that is, X∈W(p, k, ∑ x ), Y∈W(p,l,Σ y ), then their sum X+Y also obeys the complex Wishart distribution, that is, (X+Y)∈W(p,k+l,∑); S332: When the data of the target window and the background window are different, Σ x ≠Σ y When: S333: When the data of the target window and the background window are the same, Σ x =Σ y =Σ, then:

9. The method for polarimetric SAR target enhancement extraction in complex ground scenes according to claim 8, characterized in that: In step S4, the specific processing process is as follows: S41: The target data is the average of the sum of the two subview images in step S2, that is: The background data still uses full-polarization full-aperture image data; S42: respectively set the target statistical area window size X wz and background statistics area window size Y wz , where X wz <Y wz ; S43: By performing sliding window processing in the image, the logarithmic polarization statistical enhancement test value of the image target, that is, the enhanced likelihood detection value, is calculated. The constant term is removed and the logarithmic polarization statistical enhancement test value is simplified as follows: -lnR=(k+l)ln|X+Y|-kln|X|-lln|Y|.

10. The method for polarimetric SAR target enhancement extraction in complex ground scenes according to claim 9, characterized in that: In step S5, the specific processing process is as follows: S51: The probability density distribution function of the enhanced likelihood detection value is derived through the third-order Bernoulli polynomial and is approximately: P{-lnR}≈P{x 2 (p 2 )}+w·[P{χ 2 (p 2 +4)}-P{x 2 (p 2 )}] Among them, χ 2 is the chi-square distribution function, w is the weight factor; S52: Under the condition of known enhanced likelihood detection value probability distribution function, using constant false alarm rate method and given constant false alarm probability P fa , to obtain the threshold for target enhanced detection.

Images