PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration

Abstract

The invention relates to a PolSAR-GeoSIFT (PolSAR-GeoSIFT) heterogeneous data matching method based on an SAR (Synthetic Aperture Radar) geometric configuration. The method comprises the following steps: obtaining a to-be-matched feature point in a to-be-matched image and an initial reference feature point in a reference image according to a PolSAR-Harris function; processing the initial reference feature points of the to-be-matched feature points through geographic information to obtain target feature points of the to-be-matched image and reference feature points in the reference image; wherein the target feature point and the reference feature point are mutually matched points. According to the technical scheme, the feature points are extracted according to the PolSAR-Harris function, then the feature points are processed according to the geographic information, the target feature points and the reference feature points which are matched with each other are obtained, and therefore the purpose of heterogeneous data matching is achieved, the method can be applied to feature point matching of heterogeneous SAR images of different wavebands, different time phases, different lower view angles and the like, and the matching accuracy of the heterogeneous SAR images is improved. And the method has relatively high accuracy and calculation efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of radar data processing, and in particular to a PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration. Background Art

[0002] Synthetic Aperture Radar (SAR) is a microwave imaging radar system. Because it is not restricted by weather and lighting conditions, existing ground-based, airborne, and satellite-based systems have widely used SAR in environmental monitoring, terrain mapping, resource exploration and other fields.

[0003] Polarization is an important property of electromagnetic waves, which can describe additional information about the surface characteristics of objects, such as vegetation structure, soil moisture, surface cover type, etc. Polarimetric Synthetic Aperture Radar (PolSAR) has the ability to measure the full polarization scattering information of the target and is often used in the fields of object decomposition, object classification and interpretation, and scattering mechanism analysis.

[0004] SAR image registration is a prerequisite for realizing joint analysis and processing of multi-band or multi-source SAR data. In Interferometric Synthetic Aperture Radar (InSAR), SAR image registration is mainly based on image grayscale information or image features. Registration based on image grayscale information generally uses sliding window search to achieve offset estimation of homonymous points by optimizing measurement functions. Measurement functions mainly include: correlation coefficient, spectral function, wave function, etc. Registration methods based on image grayscale information are sensitive to image noise, rotational deformation, etc., and are mainly applicable to SAR data of the same sensor without nonlinear distortion. For heterogeneous data, factors such as different bands or incident angles will lead to inconsistent image grayscale features. Different observation geometry and working modes will also lead to spatial variation in the image. At this time, the corresponding transformation relationship of the image can be established by extracting the invariant features in the image. In 1999, some scholars first proposed the Scale Invariant Feature Transform (SIFT) algorithm; this algorithm has rotation and scale invariance and has been widely used in the fields of computer vision and image processing. In 2014, some scholars proposed a SIFT-like algorithm for SAR data - SAR-SIFT. This method uses the exponentially weighted mean ratio to calculate the image gradient, which can effectively reduce the false alarm rate of SIFT feature selection; in 2016, some scholars introduced an enhanced feature matching method based on the SIFT algorithm - PSO-SIFT algorithm, which has good adaptability in multi-spectral and multi-source sensor image matching.

[0005] SIFT features are invariant to scale, rotation, and affine transformations, and have great advantages for SAR data registration. However, its high algorithm complexity, computational efficiency, and memory usage limit its application to a certain extent. In addition, for heterogeneous SAR data, the drastic rotation and scale changes between images also bring greater challenges to the accuracy and computational efficiency of local feature point detection. In view of the needs of cross-processing and joint application of SAR data, it is an indispensable task to achieve accurate and efficient heterogeneous image registration and feature point extraction. Summary of the invention

[0006] In order to solve the above problems existing in the prior art, the present invention provides a PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration.

[0007] According to a first aspect of an embodiment of the present invention, a PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration is provided, the method comprising:

[0008] According to the PolSAR-Harris function, feature points to be matched in the image to be matched and initial reference feature points in the reference image are obtained;

[0009] The initial reference feature points are processed by the feature points to be matched through geographic information to obtain the target feature points of the image to be matched and the reference feature points in the reference image; wherein the target feature points and the reference feature points are mutually matched points.

[0010] Optionally, after the feature points to be matched and the initial reference feature points are processed by using geographic information to obtain target feature points and reference feature points that match each other, the method further includes:

[0011] Respectively acquiring a first sub-feature point and a second sub-feature point in the target feature point and the reference feature point;

[0012] Obtain a transformation matrix according to the first sub-feature point and the second sub-feature point;

[0013] Verify whether the target feature point matches other target feature points and other reference feature points in the reference feature points according to the transformation matrix.

[0014] Optionally, obtaining the feature points to be matched in the image to be matched and the initial reference feature points in the reference image according to the PolSAR-Harris function includes:

[0015] Acquire a first image value of a first pixel in the image to be matched and a second image value of a second pixel in the reference image;

[0016] Obtaining a first exponentially weighted mean ratio of rectangular areas on both sides of the first pixel according to the first image value, and obtaining a first exponentially weighted mean ratio of rectangular areas on both sides of the second pixel according to the second image value;

[0017] Obtaining a first matrix and a second matrix according to the first mean ratio and the second mean ratio;

[0018] The feature points to be matched and the initial reference feature points are obtained according to the PolSAR-Harris function, the first matrix and the second matrix.

[0019] Optionally, acquiring a first image value of a first pixel in the image to be matched and a second image value of a second pixel in the reference image includes:

[0020] Obtaining a first surface scattering component corresponding to the first pixel and a second surface scattering component corresponding to the second pixel according to a surface scattering model;

[0021] Obtaining a first secondary scattering component corresponding to the first pixel and a second secondary scattering component corresponding to the second pixel according to a secondary scattering model;

[0022] Obtain a first image value according to the first surface scattering component and the first secondary scattering component;

[0023] A second image value is obtained according to the second surface scattered component and the second secondary scattered component.

[0024] Optionally, the processing of the initial reference feature points with respect to the feature points to be matched by using geographic information to obtain the target feature points of the image to be matched and the reference feature points in the reference image includes:

[0025] Acquire the first geographic information of the feature point to be matched and the first coordinates of the feature point to be matched in the image to be matched;

[0026] Obtaining a second coordinate of the feature point to be matched in the reference image according to the first geographic information and the first coordinate;

[0027] Obtaining a third coordinate of an initial reference feature point corresponding to the feature point to be matched in the reference image;

[0028] The feature point to be matched and the initial reference feature point whose difference between the second coordinate and the third coordinate is less than a preset threshold are respectively used as the target feature point and the reference feature point.

[0029] Optionally, the PolSAR-Harris function is represented by reference as follows:

[0030] R SH (x,y,σ)=Det(C PH (x,y,σ))-d·Trace(C PH (x,y,σ) 2 ;

[0031] Among them, R SH (x, y, σ) represents the coordinate information of the pixel point (x, y), Det() represents the determinant of the matrix, C PH (x, y, σ) is the matrix about the pixel point (x, y), σ is the scale factor, d is the preset extreme point detection threshold, and Trace() means finding the trace of the matrix.

[0032] Optionally, the matrix C about the pixel point (x, y) PH (x, y, σ) is represented as follows:

[0033]

[0034] Where G is the convolution function, represents the convolution operation, Γ x,σ' Represents the horizontal gradient of the pixel point (x, y), Γ y,σ' Represents the vertical gradient of the pixel point (x, y), and σ' is the exponential function scale factor.

[0035] Optionally, the first exponential weighted mean ratio and the second exponential weighted mean ratio are expressed as follows:

[0036]

[0037] Among them, R f,σ' Represents the exponential weighted mean ratio of the rectangular area on both sides of the pixel point (x, y), M 1,σ' Represents the exponential weighted mean of the rectangular area on the first side of the pixel point (x, y), M 2,σ' Represents the exponentially weighted mean of the rectangular area on the second side of the pixel point (x, y), and f represents the rotation angle of the mean calculation.

[0038] Optionally, the exponential weighted mean of the rectangular areas on both sides of the pixel point (x, y) is expressed as follows:

[0039]

[0040] in, is a weighted exponential function, x * is the difference between the integrated pixel and the pixel (x, y) in x, y* is the difference between the integral pixel and the pixel (x, y), R y' Represents the integration range, I1 represents the rectangular area on the first side of the pixel point (x, y), and I2 represents the rectangular area on the second side of the pixel point (x, y).

[0041] Optionally, the second coordinate of the feature point to be matched in the reference image is obtained by referring to the following formula:

[0042]

[0043] Y'=(|p t -p s |-R near ) / R cell ;

[0044] The first coordinates of the matching feature point in the image to be matched are (X, Y), and the first geographic information is p t =(X T ,Y T ,Z T ), p s is the reference image at time n in azimuth a The reference image platform position vector at the time, v is the velocity vector of the reference image platform relative to the matching feature point, and f' dc is the Doppler center frequency at a certain azimuth, λ is the wavelength, f dc is the Doppler center frequency, X' is the azimuth coordinate in the second coordinate, R near is the nearest slant distance of the reference image, R cell is the range sampling interval of the reference image, and Y' is the range coordinate in the second coordinate.

[0045] The technical solution provided by the present invention may include the following beneficial effects:

[0046] Through the above technical scheme, feature points are extracted according to the PolSAR-Harris function, and then the feature points are processed according to geographic information to obtain mutually matching target feature points and reference feature points, thereby achieving the purpose of heterogeneous data matching. It can be applied to feature point matching of heterogeneous SAR images of different bands, different phases, and different viewing angles, with high accuracy and computational efficiency.

[0047] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present invention but do not constitute a limitation of the present invention. In the accompanying drawings:

[0049] Figure 1 It is a flow chart of a PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to an exemplary embodiment.

[0050] Figure 2 It is a schematic diagram of an experimental scenario according to an exemplary embodiment.

[0051] Figure 3 The figure is a polarization pseudo-color composite image of GF-3 in an experimental scene according to an exemplary embodiment.

[0052] Figure 4 The figure is a polarization pseudo-color composite image of TanDEM-X in an experimental scene according to an exemplary embodiment.

[0053] Figure 5 is a schematic diagram of SIFT feature point matching and comparison results corresponding to situation 1 according to an exemplary embodiment.

[0054] Figure 6 It is a schematic diagram of SAR-SIFT feature point matching and comparison results corresponding to situation 1 according to an exemplary embodiment.

[0055] Figure 7 is a schematic diagram of PSO-SIFT feature point matching and comparison results corresponding to situation 1 according to an exemplary embodiment.

[0056] Figure 8 It is a schematic diagram of SAR-GeoSIFT feature point matching and comparison results corresponding to situation 1 according to an exemplary embodiment.

[0057] Fig. 9 It is a schematic diagram of PolSAR-GeoSIFT feature point matching and comparison results corresponding to situation 1 according to an exemplary embodiment.

[0058] Fig.10 FIG. 1 is a schematic diagram of SIFT feature point matching and comparison results corresponding to situation 2 according to an exemplary embodiment.

[0059] Fig.11 It is a schematic diagram of SAR-SIFT feature point matching and comparison results corresponding to situation 2 according to an exemplary embodiment.

[0060] Fig.12is a schematic diagram of PSO-SIFT feature point matching comparison results corresponding to situation 2 according to an exemplary embodiment.

[0061] Fig.13 It is a schematic diagram of SAR-GeoSIFT feature point matching and comparison results corresponding to situation 2 according to an exemplary embodiment.

[0062] Fig.14 It is a schematic diagram of PolSAR-GeoSIFT feature point matching and comparison results corresponding to situation 2 according to an exemplary embodiment.

[0063] Fig.15 is a schematic diagram of SIFT feature point matching and comparison results corresponding to situation three according to an exemplary embodiment.

[0064] Fig.16 is a schematic diagram of SAR-SIFT feature point matching and comparison results corresponding to situation three according to an exemplary embodiment.

[0065] Fig.17 is a schematic diagram of PSO-SIFT feature point matching comparison results corresponding to situation three according to an exemplary embodiment.

[0066] Fig.18 is a schematic diagram of SAR-GeoSIFT feature point matching and comparison results corresponding to situation three according to an exemplary embodiment.

[0067] Fig.19 is a schematic diagram of PolSAR-GeoSIFT feature point matching and comparison results corresponding to situation three according to an exemplary embodiment. DETAILED DESCRIPTION

[0068] Figure 1 is a flow chart of a PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to an exemplary embodiment. Figure 1 As shown, the method includes the following steps.

[0069] S101 . Obtaining feature points to be matched in the image to be matched and initial reference feature points in the reference image according to the PolSAR-Harris function.

[0070] It can be understood that the PolSAR-Harris function, i.e., the polarimetric synthetic aperture radar Harris function, is obtained based on the polarimetric synthetic aperture radar Harris function. The PolSAR target decomposition is a polarimetric data analysis method based on the actual physical constraints to interpret the target scattering mechanism. By decomposing the complex polarimetric scattering matrix, coherence matrix or covariance matrix into several physically meaningful scattering mechanisms, different ground objects are identified and classified. It is worth mentioning that the present invention is used in any two images of heterogeneous data, and any one of them is used as the image to be matched and the other as the reference image.

[0071] Optionally, S101 may include:

[0072] Acquire a first image value of a first pixel in the image to be matched and a second image value of a second pixel in the reference image;

[0073] Obtaining a first exponentially weighted mean ratio of a rectangular area on both sides of a first pixel according to the first image value, and obtaining a first exponentially weighted mean ratio of a rectangular area on both sides of a second pixel according to the second image value;

[0074] Obtain a first matrix and a second matrix according to the first mean ratio and the second mean ratio;

[0075] The feature points to be matched and the initial reference feature points are obtained according to the PolSAR-Harris function, the first matrix and the second matrix.

[0076] It can be understood that the first image value of the first pixel point in the image to be matched and the second image value of the second pixel point in the reference image respectively obtain the weighted mean ratio of the polarization indexes corresponding to the first pixel point and the second pixel point, and further obtain the matching feature points and the initial reference feature points.

[0077] In one embodiment, taking any pixel point (x, y) in the image to be matched or the reference image as an example, the image value I(x, y) of the pixel point (x, y) includes the image value of the polarization data surface scattering and the secondary scattering mechanism, and the image value I(x, y) can be expressed as I(x, y) = f S (x,y)+f D (x,y),f S (x,y) and f D (x, y) are the surface scattering component and secondary scattering component at (x, y) respectively; then the exponential weighted mean of the rectangular area on both sides of the pixel point (x, y) is expressed as follows:

[0078]

[0079] in, is a weighted exponential function, x *is the difference between the integrated pixel and the pixel (x, y) in x, y * is the difference between the integral pixel and the pixel (x, y), R y' Indicates the integration range, I1 indicates the rectangular area on the first side of the pixel point (x, y), and I2 indicates the rectangular area on the second side of the pixel point (x, y);

[0080] Optionally, both the first exponential weighted mean ratio and the second exponential weighted mean ratio may be expressed with reference to the following mean ratio:

[0081]

[0082] Among them, R f,σ' Represents the exponential weighted mean ratio of the rectangular area on both sides of the pixel point (x, y), M 1,σ' Represents the exponential weighted mean of the rectangular area on the first side of the pixel point (x, y), M 2,σ' Represents the exponentially weighted mean of the rectangular area on the second side of the pixel point (x, y). f represents the rotation angle of the mean calculation. When f=1, it represents the vertical weighting ratio; f=3 represents the horizontal weighting ratio.

[0083] Furthermore, the exponentially weighted mean ratio is normalized to obtain the horizontal gradient Γ of the pixel point (x, y) x,σ' and the vertical gradient Γ y,σ' , refer to the following representation:

[0084] Γ x,σ' =log(R 1,σ' )

[0085] Γ y,σ' =log(R 3,σ' );

[0086] According to the horizontal gradient Γ of the pixel point (x, y) x,σ' and the vertical gradient Γ y,σ' You can get the matrix C about the pixel point (x, y) PH (x,y,σ), the matrix C PH (x, y, σ) is represented as follows:

[0087]

[0088] Among them, G is the convolution function, represents the convolution operation, Γ x,σ' Represents the horizontal gradient of the pixel point (x, y), Γ y,σ' Represents the vertical gradient of the pixel point (x, y), σ' is the exponential function scale factor;

[0089] According to the PolSAR-Harris function and matrix CPH (x, y, σ) obtains the coordinate information of the pixel point (x, y), that is, obtains the coordinate information of the feature point to be matched in the image to be matched and the initial reference feature point in the reference image;

[0090] The PolSAR-Harris function reference is expressed as follows:

[0091] R SH (x,y,σ)=Det(C PH (x,y,σ))-d·Trace(C PH (x,y,σ) 2 ;

[0092] Among them, R SH (x, y, σ) represents the coordinate information of the pixel point (x, y), Det() represents the determinant of the matrix, C PH (x, y, σ) is the matrix about the pixel point (x, y), σ is the scale factor, d is the preset extreme point detection threshold, which can usually be set to 0.04-0.06 based on experience, and Trace() means finding the trace of the matrix.

[0093] Optionally, obtaining a first image value of a first pixel in the image to be matched and a second image value of a second pixel in the reference image includes:

[0094] Obtaining a first surface scattering component corresponding to a first pixel and a second surface scattering component corresponding to a second pixel according to a surface scattering model;

[0095] Obtaining a first secondary scattering component corresponding to the first pixel and a second secondary scattering component corresponding to the second pixel according to the secondary scattering model;

[0096] Obtaining a first image value according to a first surface scattering component and a first secondary scattering component;

[0097] A second image value is obtained according to the second surface scattered component and the second secondary scattered component.

[0098] It can be understood that the scattering matrix S1 of the surface scattering model is:

[0099]

[0100] Among them, R H and R V are the complex scattering coefficients for horizontal and vertical polarizations, respectively.

[0101] Furthermore, the covariance matrix C of the surface model can be obtained based on the scattering matrix S1 of the surface scattering model: S :

[0102]

[0103] Among them, f S represents the contribution of the surface scattering component, f S =|R V | 2 , β * represents the conjugate matrix of β,

[0104] The secondary scattering model can simulate the scattering of dihedral reflectors composed of different dielectric materials. The scattering matrix S2 of the secondary scattering model can be expressed as follows:

[0105]

[0106] Among them, R TH and R TV are the scattering coefficients of the vertical trunk for horizontally polarized waves and vertically polarized waves, R GH and R GV are the scattering coefficients of the horizontal surface, and are the horizontal and vertical polarization propagation factors, respectively.

[0107] Furthermore, according to the scattering matrix S2 of the secondary scattering model, the covariance matrix C corresponding to the secondary scattering can be obtained: D :

[0108]

[0109] Among them, f D represents the contribution of the secondary scattering component, f D =|R TV R GV | 2 ,

[0110] S102, processing the initial reference feature points with the feature points to be matched by using geographic information to obtain target feature points of the image to be matched and reference feature points in the reference image; wherein the target feature points and the reference feature points are mutually matched points.

[0111] It is understandable that in order to achieve the rotation invariance of the SIFT algorithm, the traditional SIFT method is based on the texture structure of the image, and the main direction and auxiliary direction of the feature point distribution are determined by the gradient value of the pixel neighborhood. Due to the SAR side-view imaging mechanism, there are not only rotational scale changes between SAR data, but also image geometric distortion caused by terrain undulations. Since the actual coordinates of the feature points representing the same geographical location in the two images should be the same, the feature points in the two images can be projected into the same coordinate space through geographic information to eliminate the rotation and scale transformation between the feature points, and obtain target feature points and reference feature points with rotation and scale stability.

[0112] Optionally, S102 may include:

[0113] Acquire first geographic information of the feature point to be matched and first coordinates of the feature point to be matched in the image to be matched;

[0114] Obtaining a second coordinate of the feature point to be matched in the reference image according to the first geographic information and the first coordinate;

[0115] Obtain the third coordinate of the initial reference feature point corresponding to the feature point to be matched in the reference image;

[0116] The feature point to be matched and the initial reference feature point whose difference between the second coordinate and the third coordinate is less than a preset threshold are respectively used as the target feature point and the reference feature point.

[0117] It can be understood that, according to the PolSAR-Harris function, the feature points to be matched in the image to be matched and the initial reference feature points in the reference image respectively include the coordinate information of the feature points to be matched and the coordinate information of the initial reference feature points. Taking a single feature point to be matched in the first image to be matched as an example, assuming that the coordinates of the feature point to be matched in the image to be matched are (X, Y), positioning is performed according to the platform position vector of the image to be matched and the load parameters, and the first geographic information p is obtained. t =(X T ,Y T ,Z T ), in one implementation, the second coordinate of the feature point to be matched in the reference image is obtained by referring to the following formula:

[0118]

[0119] Y'=(|p t -p s |-R near ) / R cell ;

[0120] Among them, the first coordinate of the matching feature point in the image to be matched is (X, Y), and the first geographic information is p t=(X T ,Y T ,Z T ), p s is the reference image at time n in azimuth a The reference image platform position vector at the time, v is the velocity vector of the reference image platform relative to the matching feature point, and f' dc is the Doppler center frequency at a certain azimuth, λ is the wavelength, f dc is the Doppler center frequency, X' is the azimuth coordinate in the second coordinate, R near is the closest slant distance to the reference image, R cell is the distance sampling interval of the reference image, Y' is the distance coordinate in the second coordinate, and the feature description vector F of the matching feature point in the reference image can be obtained. R = [X'Y'] and the feature description vector F of the matching feature point in the image to be matched M =[XY], further, the feature point to be matched and the initial reference feature point whose difference between the second coordinate and the third coordinate is less than a preset threshold are respectively used as the target feature point and the reference feature point.

[0121] Optionally, after S102, the method may further include:

[0122] Obtaining the first sub-feature point and the second sub-feature point in the target feature point and the reference feature point respectively;

[0123] Obtaining a transformation matrix according to the first sub-feature point and the second sub-feature point;

[0124] Verify whether the target feature point matches other target feature points and other reference feature points in the reference feature point according to the transformation matrix.

[0125] It can be understood that feature point matching is to find corresponding feature points in two or more images to obtain the transformation relationship between images to achieve image matching. Image transformation usually includes affine transformations such as rotation, translation, scaling, and flipping. For SAR images, side-view imaging makes the transformation relationship between images more complicated. Therefore, the present invention uses a perspective transformation relationship to describe the transformation relationship between the reference image and the image to be matched, which can be expressed as:

[0126]

[0127] Where H is the perspective transformation matrix, represents the linear transformation of the image, [h 31 h 32 ] represents the perspective transformation of the image, h 13 、h 23 , 1 represents the translation transformation of the image.

[0128] Expanding the above formula, we can get:

[0129]

[0130] The perspective transformation matrix H can be solved by the mutually matching feature points of the two images, so that the transformation relationship between the two images can be described by the perspective transformation matrix H, and then it can be verified whether the extracted target feature points and reference feature points are reasonable.

[0131] In one implementation, the present invention is further described below in conjunction with the accompanying drawings and examples, but the implementation of the present invention is not limited thereto.

[0132] In this embodiment, Figure 2 is a schematic diagram of an experimental scenario according to an exemplary embodiment. Figure 3 is a polarization pseudo-color composite image of GF-3 in an experimental scene according to an exemplary embodiment. Figure 4 It is a polarization pseudo-color composite image of TanDEM-X in an experimental scene shown according to an exemplary embodiment. The C-band full polarization data of GF-3 in Xi'an and the X-band full polarization data of TanDEM-X are used to verify the performance of the heterogeneous SAR data matching algorithm. At the same time, the traditional SIFT method is used to compare and analyze the SAR-SIFT method commonly used for SAR data, the enhanced feature matching method PSO-SIFT and the method proposed in the present invention. In order to further verify the necessity of using surface scattering and secondary scattering components to improve the accuracy and efficiency of the algorithm, the present invention is divided into the SAR-GeoSIFT method and the PolSAR-GeoSIFT method, wherein the SAR-GeoSIFT method uses the SAR-Harris space consistent with SAR-SIFT to extract feature points, and the PolSAR-GeoSIFT method uses the PolSAR-Harris function to extract feature points related to the polarization scattering mechanism. In this embodiment, the algorithm matching performance and operation performance are compared in three cases, and all methods do not perform manual initialization when matching feature points. The reference images are all GF-3 data, and the data to be matched are TanDEM-X data at different translation and scaling scales, which have different degrees of geometric deformation compared with the reference images. The specific parameters are shown in Table 1.

[0133] Table 1

[0134]

[0135] In the verification of the results, in order to ensure the accurate matching of PolSAR-GeoSIFT feature points, in each case, for the initially extracted feature points, the feature points are first screened based on the nearest neighbor distance ratio (NNDR), and the correct matching feature points after the initial screening are purified based on the random sample consensus (RANSAC) purification algorithm. Finally, the projection transformation matrix is ​​fitted according to the correct feature points.

[0136] Case 1:

[0137] Figure 5 is a schematic diagram of SIFT feature point matching and comparison results corresponding to a situation 1 according to an exemplary embodiment, Figure 6 is a schematic diagram of SAR-SIFT feature point matching and comparison results corresponding to a situation 1 according to an exemplary embodiment, Figure 7 is a schematic diagram of PSO-SIFT feature point matching comparison results corresponding to a situation 1 according to an exemplary embodiment, Figure 8 is a schematic diagram of SAR-GeoSIFT feature point matching and comparison results corresponding to a situation 1 according to an exemplary embodiment, Fig. 9 It is a schematic diagram of PolSAR-GeoSIFT feature point matching and comparison results corresponding to Case 1 shown in an exemplary embodiment. Multi-view processing and common area interception are performed on the initial TanDEM-X data so that the scales of the two sets of data are consistent but there is a certain translation. The data sizes are 1143×2172 and 1100×2000 respectively.

[0138] The comparison parameters are shown in Table 2. The results show that when there is no obvious scale change, the five SIFT methods can all achieve accurate matching of feature points. The SIFT method has the smallest number of matching feature points and their distribution is concentrated in the urban buildings in the center of the scene. The number of feature points of the other methods is comparable. Among them, the matching feature points of SAR-SIFT, PSO-SIFT, and SAR-GeoSIFT algorithms are similarly distributed. Compared with the SIFT algorithm, their feature points are more evenly distributed, but still concentrated in urban and suburban buildings. The matching feature points of the PolSAR-GeoSIFT method cover the entire scene. In addition to the building area, they are distributed along the river banks on the left and upper right sides of the scene, proving that the perspective transformation matrix estimated based on matching feature points can be effective for the entire scene.

[0139] The comparison results were further quantitatively analyzed to count the number of feature points initially extracted, the number of matching feature points initially screened by NNDR (Nearest Neighbor Distance Ratio), the number of matching feature points purified by RANSAC (Random Sample Consensus), and the matching running time, as shown in the following table. From the perspective of the number of image feature points, the number of image feature points extracted by SIFT and PSO-SIFT methods is similar, and PSO-SIFT significantly improves the number of final matching feature points by sacrificing computational efficiency. The number of initial feature point detections and the number of feature point matches of SAR-SIFT, SAR-GeoSIFT and PolSAR-GeoSIFT methods are better than those of the other two methods, but the number of feature points of the two GeoSIFT methods after NNDR matching is significantly increased by nearly two times compared with the SAR-SIFT method, and the number of matching points after RANSAC is increased by about 20.43% compared with the SAR-SIFT method, and the algorithm running time is increased by an order of magnitude. The reason is that the Geo-SIFT description vector is a 2-dimensional vector and describes the rotation, scale, and displacement invariant features of the reference image and the image to be matched in three-dimensional geographic space, while the SAR-SIFT description vector is 128-dimensional, and since the description vector corresponding to the feature point is not unique, confusion is easily caused when making similarity judgments.

[0140] Table 2

[0141]

[0142] Case 2:

[0143] Fig.10 is a schematic diagram of SIFT feature point matching and comparison results corresponding to situation 2 according to an exemplary embodiment. Fig.11 is a schematic diagram of SAR-SIFT feature point matching and comparison results corresponding to situation 2 according to an exemplary embodiment, Fig.12 is a schematic diagram of PSO-SIFT feature point matching comparison results corresponding to situation 2 according to an exemplary embodiment, Fig.13 is a schematic diagram of SAR-GeoSIFT feature point matching and comparison results corresponding to situation 2 according to an exemplary embodiment, Fig.14 It is a schematic diagram of the PolSAR-GeoSIFT feature point matching comparison results corresponding to the second situation shown according to an exemplary embodiment. Only the common area of ​​the initial TanDEM-X data is intercepted, so that there are obvious scale transformation and translation phenomena in the two sets of data. The data sizes are 1143×2172 and 2200×2000 respectively.

[0144] The comparison parameters are shown in Table 3. The results show that the introduction of scale change leads to a sharp deterioration in the performance of SIFT, PSO-SIFT and SAR-SIFT methods. SIFT and PSO-SIFT methods only obtain one correct matching feature point, which leads to the failure of perspective matrix estimation and the inability to achieve normal data matching. SAR-SIFT method only obtains 7 matching points, all of which are concentrated near the urban buildings in the lower part of the scene. On the contrary, the increase in the number of feature points caused by scale transformation also further increases the number of matching feature points extracted by GeoSIFT method. Compared with case 1, the distribution range of feature points matched by SAR-GeoSIFT method has expanded to the river bank on the left and upper right side of the scene; the feature points matched by PolSAR-SIFT method also cover the entire scene compared with case 1, and the distribution density of matching feature points has been significantly improved.

[0145] The comparison results are also quantitatively analyzed, as shown in the following table. It can be seen that the number of feature points detected in the image to be matched by each method increases proportionally, but the NNDR initial feature point matching performance of SIFT, SAR-SIFT, and PSO-SIFT methods has deteriorated to varying degrees compared with case 1. SIFT and PSO-SIFT are even unable to purify to the minimum number of correct matching feature points that meet the perspective matrix estimation. In contrast, the GeoSIFT method can still achieve accurate matching of feature points, and the PolSAR-GeoSIFT method shows better scale robustness than the SAR-GeoSIFT method in this case.

[0146] Table 3

[0147]

[0148] Case 3:

[0149] Fig.15 is a schematic diagram of SIFT feature point matching and comparison results corresponding to situation three according to an exemplary embodiment, Fig.16 is a schematic diagram of SAR-SIFT feature point matching and comparison results corresponding to situation three according to an exemplary embodiment, Fig.17 is a schematic diagram of PSO-SIFT feature point matching comparison results corresponding to situation three according to an exemplary embodiment, Fig.18 is a schematic diagram of SAR-GeoSIFT feature point matching and comparison results corresponding to situation three according to an exemplary embodiment, Fig.19It is a schematic diagram of the PolSAR-GeoSIFT feature point matching comparison results corresponding to case three according to an exemplary embodiment. No operation is performed on the initial TanDEM-X data. At this time, the two sets of data not only have obvious scale transformation, but also have a large translation phenomenon. The data sizes are 1143×2172 and 2550×2766 respectively.

[0150] The comparison parameters are shown in Table 4. The results show that SIFT and SAR-SIFT methods obtain 1 and 2 pairs of correctly matched feature points respectively, and PSO-SIFT method can no longer obtain correctly matched feature points. GeoSIFT method can still robustly obtain correct matching feature points. The matching feature points obtained by PolSAR-GeoSIFT method are more evenly distributed and have a higher distribution density around the scene than those by SAR-GeoSIFT method.

[0151] The quantitative analysis of Case 3 is shown in the following table. The results show that the number of feature points detected by each method in the matching image still increases proportionally, but except for SIFT and PSO-SIFT methods, SAR-SIFT method cannot extract the minimum number of correct matching feature points that meet the perspective matrix estimation. The GeoSIFT method can still achieve accurate matching of feature points, and the number of matching points is similar to that of Case 2. This is because the size of the reference image has not changed, and the GeoSIFT method has completely extracted accurate feature points in the reference image in Cases 2 and 3.

[0152] Table 4

[0153]

[0154] By comparing and analyzing various SIFT methods under three conditions, it can be concluded that: under the condition of consistent scale, SIFT, SAR-SIFT and PSO-SIFT methods can all achieve correct matching of feature points. SIFT method has the highest computational efficiency but the least number of matched feature points; SAR-SIFT method is more suitable for SAR data, with the largest number of matched feature points, but the matching takes the longest time; PSO-SIFT method achieves both efficiency and performance, and can match more feature points in a shorter time. Under the condition of inconsistent scale, SIFT, SAR-SIFT and PSO-SIFT methods deteriorate significantly, and it is almost impossible to achieve accurate image matching. The PolSAR-Harris feature point detection proposed in the present invention effectively extracts stable feature points in heterogeneous SAR data, and the Geo-SIFT description vector achieves true scale rotation invariance and reduces the vector dimension from 128 dimensions to 2 dimensions, overcoming the two core problems of accuracy and efficiency faced by heterogeneous SAR data matching.

[0155] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, a variety of simple modifications can be made to the technical solution of the present invention, and these simple modifications all belong to the protection scope of the present invention.

[0156] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations.

[0157] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the present invention, they should also be regarded as the contents disclosed by the present invention.

Claims

1. A PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration, characterized in that: The method comprises: According to the PolSAR-Harris function, feature points to be matched in the image to be matched and initial reference feature points in the reference image are obtained; The initial reference feature points are processed by the feature points to be matched through geographic information to obtain the target feature points of the image to be matched and the reference feature points in the reference image; wherein the target feature points and the reference feature points are mutually matched points.

2. The PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to claim 1, characterized in that: After the initial reference feature points are processed by the geographic information to obtain the target feature points and reference feature points that match each other, the method further includes: Respectively acquiring a first sub-feature point and a second sub-feature point in the target feature point and the reference feature point; Obtain a transformation matrix according to the first sub-feature point and the second sub-feature point; Verify whether the target feature point matches other target feature points and other reference feature points in the reference feature points according to the transformation matrix.

3. The PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to claim 1, characterized in that: The step of obtaining the feature points to be matched in the image to be matched and the initial reference feature points in the reference image according to the PolSAR-Harris function includes: Acquire a first image value of a first pixel in the image to be matched and a second image value of a second pixel in the reference image; Obtaining a first exponentially weighted mean ratio of rectangular areas on both sides of the first pixel according to the first image value, and obtaining a first exponentially weighted mean ratio of rectangular areas on both sides of the second pixel according to the second image value; Obtaining a first matrix and a second matrix according to the first mean ratio and the second mean ratio; The feature points to be matched and the initial reference feature points are obtained according to the PolSAR-Harris function, the first matrix and the second matrix.

4. The PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to claim 3 is characterized in that: The obtaining a first image value of a first pixel in the image to be matched and a second image value of a second pixel in the reference image includes: Obtaining a first surface scattering component corresponding to the first pixel and a second surface scattering component corresponding to the second pixel according to a surface scattering model; Obtaining a first secondary scattering component corresponding to the first pixel and a second secondary scattering component corresponding to the second pixel according to a secondary scattering model; Obtain a first image value according to the first surface scattering component and the first secondary scattering component; A second image value is obtained according to the second surface scattered component and the second secondary scattered component.

5. The PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to claim 1, characterized in that: The step of processing the initial reference feature points with respect to the feature points to be matched by using geographic information to obtain the target feature points of the image to be matched and the reference feature points in the reference image includes: Acquire the first geographic information of the feature point to be matched and the first coordinates of the feature point to be matched in the image to be matched; Obtaining a second coordinate of the feature point to be matched in the reference image according to the first geographic information and the first coordinate; Obtaining a third coordinate of an initial reference feature point corresponding to the feature point to be matched in the reference image; The feature point to be matched and the initial reference feature point whose difference between the second coordinate and the third coordinate is less than a preset threshold are respectively used as the target feature point and the reference feature point.

6. The PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to claim 3, characterized in that: The PolSAR-Harris function reference is expressed as follows: R SH (x,y,σ)=Det(C PH (x,y,σ))-d·Trace(C PH (x,y,σ)) 2 ; Among them, R SH (x, y, σ) represents the coordinate information of the pixel point (x, y), Det(·) represents the determinant of the matrix, C PH (x, y, σ) is the matrix about the pixel point (x, y), σ is the scale factor, d is the preset extreme point detection threshold, and Trace(·) means finding the trace of the matrix.

7. The PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to claim 6, characterized in that: The matrix C about the pixel point (x, y) PH (x, y, σ) is represented as follows: Where G is the convolution function, represents the convolution operation, Γ x，σ' Represents the horizontal gradient of the pixel point (x, y), Γ y，σ' Represents the vertical gradient of the pixel point (x, y), and σ' is the exponential function scale factor.

8. The PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to claim 7, characterized in that: The first exponential weighted mean ratio and the second exponential weighted mean ratio are expressed as follows: Among them, R f,σ' Represents the exponential weighted mean ratio of the rectangular area on both sides of the pixel point (x, y), M 1,σ' Represents the exponential weighted mean of the rectangular area on the first side of the pixel point (x, y), M 2,σ' Represents the exponentially weighted mean of the rectangular area on the second side of the pixel point (x, y), and f represents the rotation angle of the mean calculation.

9. The PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to claim 8, characterized in that: The exponential weighted mean of the rectangular area on both sides of the pixel point (x, y) is expressed as follows: in, is a weighted exponential function, x * is the difference between the integrated pixel and the pixel (x, y) in x, y * is the difference between the integral pixel and the pixel (x, y), R y' Represents the integration range, I1 represents the rectangular area on the first side of the pixel point (x, y), and I2 represents the rectangular area on the second side of the pixel point (x, y).

10. The PolSAR-GeoSIFT heterogeneous data matching method based on SAR geometric configuration according to claim 5, characterized in that: The second coordinate of the feature point to be matched in the reference image is obtained by referring to the following formula: Y'=(|p t -p s |-R near ) / R cell ; The first coordinates of the matching feature point in the image to be matched are (X, Y), and the first geographic information is p t =(X T ,Y T ,Z T ), p s is the reference image at time n in azimuth a The reference image platform position vector at the time, v is the velocity vector of the reference image platform relative to the matching feature point, and f' dc is the Doppler center frequency at a certain azimuth, λ is the wavelength, f dc is the Doppler center frequency, X' is the azimuth coordinate in the second coordinate, R near is the nearest slant distance of the reference image, R cell is the range sampling interval of the reference image, and Y' is the range coordinate in the second coordinate.

Images