Optical and SAR satellite image matching point location refining method and system

Abstract

The invention discloses an optical and SAR satellite image matching point refining method and system, and the method comprises the steps: extracting the edge structure feature maps of an optical satellite image and an SAR satellite image through a multi-scale Log-Gabor filter, and extracting the feature points of the edge structure feature maps of the two satellite images through a Harris algorithm; constructing radiation independent feature descriptors containing local abstract structures for the feature points; and based on the radiation-independent feature descriptors of the feature points, refining of the matching point positions of the optical satellite image and the SAR satellite image in the research area is realized by using a coarse-to-fine enhanced matching strategy. According to the method, the radiation independent feature descriptor containing the local abstract structure is constructed to describe and represent the features, the complexity of a matching algorithm can be reduced, mismatching existing in initial matching is eliminated by using a coarse-to-fine enhanced matching strategy, the refining of matching point locations is realized, and the matching precision is greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of image feature extraction, and in particular relates to a method and system for refining matching points of optical and SAR satellite images. Background Art

[0002] The rapid development of high-resolution remote sensing technology provides crucial technical support for detailed monitoring and analysis of the Earth's surface. In particular, with the successful launch of domestically produced high-resolution optical satellites such as Tianhui-1 and Ziyuan-3, as well as synthetic aperture radar (SAR) satellites such as Gaofen-3 and Tianhui-2, the acquisition and processing of large-scale, multi-source remote sensing data worldwide has become increasingly feasible and accurate. The combination of domestically produced optical and SAR satellite imagery enables high-precision terrain mapping, target identification and positioning, three-dimensional reconstruction, change detection, and other tasks. The success of these applications relies heavily on the accuracy of matching optical and SAR imagery.

[0003] Optical and SAR satellite image matching is an important component of multimodal remote sensing image matching. Its goal is to identify the same or similar content in optical and SAR satellite images and complete high-precision alignment and matching of the images. This process is particularly challenging for satellites that image at different times, under different conditions, and with different sensors. Traditional image matching methods usually rely on the gradient or grayscale information of the image to detect and describe features. However, due to the significant nonlinear radiometric and geometric differences between multimodal images, traditional image matching methods often perform poorly in such tasks. Therefore, how to eliminate the impact of linear radiometric and geometric differences on optical and SAR image matching and improve the accuracy and robustness of the matching algorithm are current problems that need to be solved in optical and SAR image matching technology. Summary of the Invention

[0004] In view of the shortcomings of the prior art, the present invention provides a method for refining the matching points of optical and SAR satellite images, comprising the following steps: Step 1: Obtain optical satellite images and SAR satellite images of the study area; Step 2: Use multi-scale Log-Gabor filter to extract edge structure feature maps of optical satellite images and SAR satellite images, and use Harris algorithm to extract feature points of edge structure feature maps of the two satellite images; Step 3: Construct a radiation-independent feature descriptor containing a local abstract structure for the feature points extracted in step 2; Step 4: Based on the radiation-independent feature descriptor of the feature points, a coarse-to-fine enhanced matching strategy is used to refine the matching points between the optical satellite image and the SAR satellite image of the study area.

[0005] Furthermore, the Log-Gabor filter in step 2 is defined as follows: (1) Where, L represents Log-Gabor filtering, represents the polar diameter of logarithmic polar coordinates, represents the polar angle in log polar coordinates, represents the center frequency, The scale is s , direction is o The center direction of the Log-Gabor filter, express The bandwidth parameter, express bandwidth parameter.

[0006] Convolution operations are performed on optical satellite images and SAR satellite images with Log-Gabor filters to generate frequency domain response components at different scales and directions. The formula is as follows: (2) Where I represents the input image, which is an optical satellite image or a SAR satellite image; represents an even-symmetric Log-Gabor filter on scale s and direction o; represents an odd-symmetric Log-Gabor filter at scale s and direction o; express Frequency domain response components at scale s and direction o; express Frequency domain response components at scale s and direction o; Represents a convolution operation. The amplitude response component and phase response component of the input image are calculated based on the frequency domain response components at different scales and directions. The calculation formula is as follows: (3) (4) Where, Indicates that the input image is at scale s and direction o The magnitude response component on Indicates that the input image is at scale s and direction o The phase response component on .

[0007] The phase consistency of the input image is calculated based on the amplitude response component and the phase response component. The calculation formula is as follows: (5) Where PC represents the phase consistency of the input image; is the weight function; Represents the amplitude response component of the input image at scale s and direction o; is the phase deviation function; T is the noise threshold; is a minimum value, preventing the denominator from being zero; the symbol Indicates that when its value is positive, the enclosed amount is equal to itself, otherwise it is zero.

[0008] The second-order moment of the input image is calculated by phase consistency to obtain the edge structure feature map of the input image. The calculation formula of the second-order moment is as follows: (6) Where, The second-order moment of the input image, express o The angle of direction.

[0009] The Harris operator is used to extract the feature points of the edge structure feature map of the input image. The calculation formulas of the covariance matrix and response function in the Harris operator are as follows: (7) (8) Where, M is the covariance matrix; w is the Gaussian window function; Represents the convolution operation; express projection in the vertical direction; express Projection in the horizontal direction; R is the response function; Denotes the matrix determinant value; k is an empirical constant with a value range of 0.04-0.06; represents the trace of the matrix determinant.

[0010] Points whose response function values are greater than the set threshold are selected as feature points.

[0011] Furthermore, in step 3, the spatial domain convolution kernel is constructed by using Log-Gabor filtering on the result of filtering the image through the spatial domain even symmetric filter and the result of filtering the image through the spatial domain odd symmetric filter, and an inverse Fourier transform is performed in the spatial domain. The formula is as follows: (9) Where, represents the inverse Fourier transform, L represents Log-Gabor filtering,i represents a unit imaginary number, o represents the direction of the Log-Gabor filter, Represents the real part of the spatial domain convolution kernel, corresponding to the result of filtering by an even symmetric filter, Represents the imaginary part of the spatial domain convolution kernel, corresponding to the result of odd-symmetric filter filtering, the formula is as follows: (10) Where, i represents a unit imaginary number, o represents the direction of the Log-Gabor filter, L (o) indicates o Directional Log-Gabor filtering.

[0012] Hypothesized feature points The gray value is , then the feature point In the Log-Gabor filter direction o Response value on The calculation formula is: (11) Where, represents the convolution operation, Represents the imaginary part of the convolution kernel in the spatial domain.

[0013] For each feature point, calculate its response value in each direction, and then combine it into a vector as the feature descriptor of the feature point. The formula is as follows: (12) Where, is the feature descriptor, is the Log-Gabor filter direction The response value on .

[0014] right Perform Gaussian convolution to obtain the radiation-independent feature descriptor of each feature point , the calculation formula is as follows: (13) Where, is a radiation-independent feature descriptor, is the feature descriptor, is a two-dimensional Gaussian kernel, is the standard deviation of the two-dimensional Gaussian kernel, Represents a convolution operation.

[0015] Furthermore, in step 4, when obtaining the radiation-independent feature descriptor After that, the feature points are roughly matched using the minimum Euclidean distance as the measure. After the rough matching is completed, the error is eliminated by the FSC algorithm, and the matching error is less than The matching points of pixels are taken as correct matching points, and the initial affine transformation model between the optical satellite image and the SAR satellite image is obtained from all correct matching points. .

[0016] Based on the initial affine transformation model The error between the predicted matching position and the true value is assumed to follow a Gaussian distribution with mean 0, and the Euclidean distance with position and scale constraints is calculated. , and The minimum measure is used for precise matching. The specific calculation formula is as follows: (14) Where, Represents feature points in optical satellite images location, Represents the feature points in the SAR satellite image to be matched location, Are constraint coefficients, and the calculation formula is as follows: (15) Where, Represents feature points in optical satellite images and the feature points in the SAR satellite image to be matched The position error calculation symbol, Express Coordinates after affine transformation; Indicates the absolute value calculation symbol, and Respectively and scale.

[0017] After the precise matching is completed, the FSC algorithm is used again to eliminate the mismatched points and reduce the matching error to less than The matching points of pixels are the correct matching points, and the matching points of optical and SAR satellite images are finally refined.

[0018] The present invention also provides a system for refining the matching points of optical and SAR satellite images, which is used to implement the above-mentioned method for refining the matching points of optical and SAR satellite images.

[0019] Furthermore, the invention comprises a processor and a memory, wherein the memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the above-mentioned method for refining the position of optical and SAR satellite image matching points.

[0020] Alternatively, the method comprises a readable storage medium having a computer program stored thereon, and when the computer program is executed, the method for refining the position of optical and SAR satellite image matching points as described above is implemented.

[0021] Compared with the prior art, the present invention has the following advantages: 1) The present invention uses Log-Gabor filtering to extract edge structure feature maps of optical and SAR images, which can eliminate the influence of nonlinear radiation differences on the matching of optical and SAR images; 2) The present invention constructs a radiation-independent feature descriptor containing local abstract structures to describe and represent features, which can reduce the complexity of the matching algorithm; 3) The present invention proposes a coarse-to-fine enhanced matching strategy to eliminate mismatches in the initial matching, achieve refinement of matching points, and greatly improve matching accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 This is a flowchart of refining matching points between optical satellite images and SAR satellite images according to an embodiment of the present invention.

[0024] FIG2( a ) and FIG2( b ) are a pair of optical satellite images and SAR satellite images according to an embodiment of the present invention, wherein FIG2( a ) is an optical satellite image and FIG2( b ) is a SAR satellite image.

[0025] FIG3( a ) is a diagram showing edge structure characteristics of an optical satellite image according to an embodiment of the present invention, and FIG3( b ) is a diagram showing edge structure characteristics of a SAR satellite image according to an embodiment of the present invention.

[0026] FIG4( a ) shows the feature points of the optical satellite image edge structure feature map extracted by an embodiment of the present invention, and FIG4( b ) shows the feature points of the SAR satellite image edge structure feature map extracted by an embodiment of the present invention.

[0027] Figure 5 These are radiation-independent feature maps of optical satellite images and SAR satellite images according to an embodiment of the present invention. The upper figure is a radiation-independent feature map of optical satellite images, and the lower figure is a radiation-independent feature map of SAR satellite images.

[0028] FIG6( a ) is an initial matching result of an optical satellite image and a SAR satellite image according to an embodiment of the present invention, and FIG6( b ) is a final matching result of an optical satellite image and a SAR satellite image according to an embodiment of the present invention.

[0029] 7(a)-7(f) are the local point matching results of the optical satellite image and the SAR satellite image according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention are further described below with reference to the accompanying drawings and embodiments. It is obvious that the embodiments described are only some of the embodiments of the present invention, not all of them. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0031] Example 1 like Figure 1 As shown, an embodiment of the present invention provides a method for refining the matching points of optical and SAR satellite images, comprising the following steps: Step 1: Obtain optical satellite images and SAR satellite images of the study area.

[0032] Step 2: Use multi-scale Log-Gabor filter to extract edge structure feature maps of optical satellite images and SAR satellite images, and use Harris algorithm to extract feature points of edge structure feature maps of the two satellite images.

[0033] The Log-Gabor filter is defined as follows: (1) Where, L represents Log-Gabor filtering, represents the polar diameter of logarithmic polar coordinates, represents the polar angle in log polar coordinates, represents the center frequency, The scale is s , direction is o The center direction of the Log-Gabor filter, express The bandwidth parameter, express bandwidth parameter.

[0034] Convolution operations are performed on optical satellite images and SAR satellite images with Log-Gabor filters to generate frequency domain response components at different scales and directions. The formula is as follows: (2) Where I represents the input image, which is an optical satellite image or a SAR satellite image; represents an even-symmetric Log-Gabor filter on scale s and direction o; represents an odd-symmetric Log-Gabor filter at scale s and direction o; express Frequency domain response components at scale s and direction o; express Frequency domain response components at scale s and direction o; Represents a convolution operation.

[0035] The amplitude response component and phase response component of the input image are calculated based on the frequency domain response components at different scales and directions. The calculation formula is as follows: (3) (4) Where, Indicates that the input image is at scale s and direction o The magnitude response component on Indicates that the input image is at scale s and direction o The phase response component on .

[0036] The phase consistency of the input image is calculated based on the amplitude response component and the phase response component. The calculation formula is as follows: (5) Where PC represents the phase consistency of the input image; is the weight function; Represents the amplitude response component of the input image at scale s and direction o; is the phase deviation function; T is the noise threshold; is a minimum value, preventing the denominator from being zero; the symbol Indicates that when its value is positive, the enclosed amount is equal to itself, otherwise it is zero.

[0037] The second-order moment of the input image is calculated by phase consistency to obtain the edge structure feature map of the input image. The calculation formula of the second-order moment is as follows: (6) Where, The second-order moment of the input image, express o The angle of direction.

[0038] Considering that image moments have a strong response at edges and corners, the Harris operator is used to extract the feature points of the input image edge structure feature map. The calculation formulas for the covariance matrix and response function in the Harris operator are as follows: (7) (8) Where, M is the covariance matrix; w is the Gaussian window function; Represents the convolution operation; express projection in the vertical direction; express Projection in the horizontal direction; R is the response function; Denotes the matrix determinant value; k is an empirical constant with a value range of 0.04-0.06; represents the trace of the matrix determinant.

[0039] For a pair of optical and SAR satellite images, as shown in Figures 2(a) and 2(b), Figure 2(a) is the optical satellite image, and Figure 2(b) is the SAR satellite image. The edge structure feature maps of the image pair are extracted, as shown in Figures 3(a) and 3(b). Figure 3(a) shows the edge structure feature map of the optical satellite image, while Figure 3(b) shows the edge structure feature map of the SAR satellite image. It can be seen that the edge structure feature maps of the optical and SAR satellite images eliminate nonlinear radiometric differences between the images and effectively represent the image structure information. Feature detection is performed on the edge structure feature maps using an operator, resulting in the results shown in Figures 4(a) and 4(b). Figure 4(a) shows the feature points extracted from the edge structure feature map of the optical satellite image, while Figure 4(b) shows the feature points extracted from the edge structure feature map of the SAR satellite image. It can be seen that the proposed multi-scale Log-Gabor filter + Harris algorithm method can extract a large number of evenly distributed feature points from both the optical and SAR satellite images.

[0040] Step 3: For the feature points extracted in step 2, a radiation-independent feature descriptor containing a local abstract structure is constructed.

[0041] Considering the advantages of the Log-Gabor filter in terms of frequency response characteristics and directional selectivity, it can accurately extract the edge and texture information of the image at multiple scales and directions. Therefore, the Log-Gabor filter is selected to filter the remote sensing image. The Log-Gabor filter presents a log-normal distribution in the spatial frequency domain. This characteristic enables it to effectively suppress noise and prevent the attenuation problem of high-frequency bands while extracting edge information. Therefore, the filtered result can be regarded as the equivalent gradient of the image in all directions. However, since the Log-Gabor filter consumes a lot of memory when processing large-scale remote sensing data in the frequency domain. The spatial domain convolution kernel is constructed by using the Log-Gabor filter to filter the image after the spatial domain even symmetric filter and the image after the spatial domain odd symmetric filter, and the inverse Fourier transform is performed in the spatial domain. The formula is as follows: (9) Where, represents the inverse Fourier transform, L represents Log-Gabor filtering, i represents a unit imaginary number, o represents the direction of the Log-Gabor filter, Represents the real part of the spatial domain convolution kernel, corresponding to the result of filtering by an even symmetric filter, Represents the imaginary part of the spatial domain convolution kernel, corresponding to the result of odd-symmetric filter filtering, the formula is as follows: (10) Where, i represents a unit imaginary number, o represents the direction of the Log-Gabor filter, L (o) indicates o Directional Log-Gabor filtering.

[0042] Taking into account High sensitivity to energy changes at the edge of the image, select The filtering result is the equivalent of the first-order gradient of the image. Assume that the feature point The gray value is , then the feature point In the Log-Gabor filter direction o Response value on The calculation formula is: (11) Where, represents the convolution operation, Represents the imaginary part of the convolution kernel in the spatial domain.

[0043] For each feature point, calculate its response value in each direction, and then combine it into a vector as the feature descriptor of the feature point. The formula is as follows: (12) Where, is the feature descriptor, is the Log-Gabor filter direction The response value on ; In order to improve the robustness of the feature descriptor, Perform Gaussian convolution to obtain the radiation-independent feature descriptor of each feature point , the calculation formula is as follows: (13) Where, is a radiation-independent feature descriptor, is the feature descriptor, is a two-dimensional Gaussian kernel, is the standard deviation of the two-dimensional Gaussian kernel, Represents a convolution operation.

[0044] according to At the spatial position in the image, the radiation-independent feature descriptors are organized into a radiation-independent feature map. This feature map contains the local abstract feature structure of the image, so it can be used as a dense feature representation of the image. In this embodiment, the radiation-independent feature map containing the local abstract structure is obtained as follows: Figure 5 As shown, the upper figure is the radiation-independent feature map of optical satellite images, and the lower figure is the radiation-independent feature map of SAR satellite images.

[0045] Step 4: Based on the radiation-independent feature descriptor of the feature points, a coarse-to-fine enhanced matching strategy is used to refine the matching points between the optical satellite image and the SAR satellite image of the study area.

[0046] In obtaining the radiation-independent feature descriptor Finally, the feature points are roughly matched using the minimum Euclidean distance as the measure. After the rough matching is completed, the errors are eliminated by the FSC algorithm, and the matching points with matching errors less than 10 pixels are regarded as correct matching points. The initial affine transformation model between the optical satellite image and the SAR satellite image is obtained from all the correct matching points. .

[0047] Based on the initial affine transformation model The error between the predicted matching position and the true value is assumed to follow a Gaussian distribution with mean 0, and the Euclidean distance with position and scale constraints is calculated. , and The minimum measure is used for precise matching. The specific calculation formula is as follows: (14) Where, Represents feature points in optical satellite images location, Represents the feature points in the SAR satellite image to be matched location, Are constraint coefficients, and the calculation formula is as follows: (15) Where, Represents feature points in optical satellite images and the feature points in the SAR satellite image to be matched The position error calculation symbol, Express Coordinates after affine transformation; Indicates the absolute value calculation symbol, and Respectively and scale.

[0048] After the precise matching is completed, the FSC algorithm is used again to eliminate the incorrect matching points, and the matching points with a matching error of less than 3 pixels are regarded as correct matching points, ultimately achieving the refinement of the matching points between optical and SAR satellite images.

[0049] This embodiment uses the minimum Euclidean distance as a measure to perform rough matching of feature points, eliminates errors through the FSC algorithm, and regards matching points with matching errors less than 10 pixels as correct matching points, obtaining the initial matching result of Figure 6 (a). It can be found that there are a large number of mismatches in Figure 6 (a). Calculate the Euclidean distance with position and scale constraints , and The minimum measurement is used for precise matching. Errors are eliminated using the FSC algorithm, and matching points with matching errors less than 3 pixels are considered correct matching points. The final matching result in Figure 6(b) is obtained. It can be seen that Figure 6(b) retains high-quality feature correspondences. Figures 7(a), 7(b), 7(c), 7(d), 7(e), and 7(f) show the local point positioning results. It can be seen that the proposed method achieves high-precision point positioning.

[0050] Example 2 Based on the same inventive concept, the present invention also provides a system for refining optical and SAR satellite image matching points, including a processor and a memory, the memory being used to store program instructions, and the processor being used to call the program instructions in the memory to execute the above-mentioned method for refining optical and SAR satellite image matching points.

[0051] Example 3 Based on the same inventive concept, the present invention also provides a system for refining optical and SAR satellite image matching points, including a readable storage medium, on which a computer program is stored. When the computer program is executed, it implements the above-mentioned method for refining optical and SAR satellite image matching points.

[0052] In specific implementation, the method proposed in the technical solution of the present invention can be automatically run by those skilled in the art using computer software technology. System devices that implement the method, such as computer-readable storage media that store the corresponding computer program of the technical solution of the present invention and computer equipment that runs the corresponding computer program, should also be within the scope of protection of the present invention.

[0053] The specific embodiments described herein are merely illustrative of the spirit of the present invention. Persons skilled in the art may make various modifications, additions, or substitutions to the described specific embodiments without departing from the spirit of the present invention or exceeding the scope of the appended claims.

Claims

1. A method for refining the matching points of optical and SAR satellite images, characterized in that: The following steps are involved: Step 1: Obtain optical satellite images and SAR satellite images of the study area; Step 2: Use multi-scale Log-Gabor filter to extract edge structure feature maps of optical satellite images and SAR satellite images, and use Harris algorithm to extract feature points of edge structure feature maps of the two satellite images; Step 3: Construct a radiation-independent feature descriptor containing a local abstract structure for the feature points extracted in step 2; Step 4: Based on the radiation-independent feature descriptor of the feature points, a coarse-to-fine enhanced matching strategy is used to refine the matching points between the optical satellite image and the SAR satellite image of the study area.

2. The method for refining optical and SAR satellite image matching points according to claim 1, characterized in that: The Log-Gabor filter in step 2 is defined as follows: （1） Where, L represents Log-Gabor filtering, represents the polar diameter of logarithmic polar coordinates, represents the polar angle in log polar coordinates, represents the center frequency, The scale is s , direction is o The center direction of the Log-Gabor filter, express The bandwidth parameter, express Bandwidth parameters; Convolution operations are performed on optical satellite images and SAR satellite images with Log-Gabor filters to generate frequency domain response components at different scales and directions. The formula is as follows: (2) Where I represents the input image, which is an optical satellite image or a SAR satellite image; represents an even-symmetric Log-Gabor filter on scale s and direction o; represents an odd-symmetric Log-Gabor filter at scale s and direction o; express Frequency domain response components at scale s and direction o; express Frequency domain response components at scale s and direction o; Represents a convolution operation.

3. The method for refining the matching points of optical and SAR satellite images according to claim 2, characterized in that: In step 2, the amplitude response component and phase response component of the input image are calculated based on the frequency domain response components at different scales and directions. The calculation formula is as follows: （3） （4） Where, Indicates that the input image is at scale s and direction o The magnitude response component on Indicates that the input image is at scale s and direction o The phase response component on ; The phase consistency of the input image is calculated based on the amplitude response component and the phase response component. The calculation formula is as follows: （5） Where PC represents the phase consistency of the input image; is the weight function; Represents the amplitude response component of the input image at scale s and direction o; is the phase deviation function; T is the noise threshold; is a minimum value, preventing the denominator from being zero; the symbol Indicates that when its value is positive, the closed quantity is equal to itself, otherwise it is zero; The second-order moment of the input image is calculated by phase consistency to obtain the edge structure feature map of the input image. The calculation formula of the second-order moment is as follows: （6） Where, The second-order moment of the input image, express o The angle of direction.

4. The method for refining the matching points of optical and SAR satellite images according to claim 3, characterized in that: In step 2, the Harris operator is used to extract the feature points of the edge structure feature map of the input image. The calculation formulas of the covariance matrix and response function in the Harris operator are as follows: （7） （8） Where, M is the covariance matrix; w is the Gaussian window function; Represents the convolution operation; express projection in the vertical direction; express Projection in the horizontal direction; R is the response function; Denotes the matrix determinant value; k is an empirical constant with a value range of 0.04-0.06; represents the trace of the matrix determinant; Points whose response function values are greater than the set threshold are selected as feature points.

5. The method for refining optical and SAR satellite image matching points according to claim 1, wherein: In step 3, the Log-Gabor filter is used to construct the spatial domain convolution kernel by filtering the image with the spatial domain even symmetric filter and the image with the spatial domain odd symmetric filter, and then perform the inverse Fourier transform in the spatial domain. The formula is as follows: （9） Where, represents the inverse Fourier transform, L represents Log-Gabor filtering, i represents a unit imaginary number, o represents the direction of the Log-Gabor filter, Represents the real part of the spatial domain convolution kernel, corresponding to the result of filtering by an even symmetric filter, Represents the imaginary part of the spatial domain convolution kernel, corresponding to the result of odd-symmetric filter filtering, the formula is as follows: （10） Where, i represents a unit imaginary number, o represents the direction of the Log-Gabor filter, L (o) indicates o Directional Log-Gabor filtering; Hypothesized feature points The gray value is , then the feature point In the Log-Gabor filter direction o Response value on The calculation formula is: （11） Where, represents the convolution operation, Represents the imaginary part of the convolution kernel in the spatial domain.

6. The method for refining optical and SAR satellite image matching points according to claim 5, characterized in that: In step 3, for each feature point, calculate its response value in each direction, and then combine it into a vector as the feature descriptor of the feature point. The formula is as follows: （12） Where, is the feature descriptor, is the Log-Gabor filter direction The response value on ; right Perform Gaussian convolution to obtain the radiation-independent feature descriptor of each feature point , the calculation formula is as follows: （13） Where, is a radiation-independent feature descriptor, is the feature descriptor, is a two-dimensional Gaussian kernel, is the standard deviation of the two-dimensional Gaussian kernel, Represents a convolution operation.

7. The method for refining optical and SAR satellite image matching points according to claim 1, wherein: In step 4, we obtain the radiation-independent feature descriptor After that, the feature points are roughly matched using the minimum Euclidean distance as the measure. After the rough matching is completed, the error is eliminated by the FSC algorithm, and the matching error is less than The matching points of pixels are taken as correct matching points, and the initial affine transformation model between the optical satellite image and the SAR satellite image is obtained from all correct matching points. .

8. The method for refining optical and SAR satellite image matching points according to claim 7, characterized in that: In step 4, based on the initial affine transformation model The error between the predicted matching position and the true value is assumed to follow a Gaussian distribution with mean 0, and the Euclidean distance with position and scale constraints is calculated. , and The minimum measure is used for precise matching. The specific calculation formula is as follows: （14） Where, Represents feature points in optical satellite images location, Represents the feature points in the SAR satellite image to be matched location, Are constraint coefficients, and the calculation formula is as follows: （15） Where, Represents feature points in optical satellite images and the feature points in the SAR satellite image to be matched The position error calculation symbol, Express Coordinates after affine transformation; Indicates the absolute value calculation symbol, and Respectively and Scale; After the precise matching is completed, the FSC algorithm is used again to eliminate the mismatched points and reduce the matching error to less than The matching points of pixels are the correct matching points, and the matching points of optical and SAR satellite images are finally refined.

9. A system for refining the matching points of optical and SAR satellite images, characterized in that: The method comprises a processor and a memory, wherein the memory is used to store program instructions, and the processor is used to call the program instructions in the memory to execute the method for refining the position of optical and SAR satellite image matching points as described in any one of claims 1 to 8.

10. A system for refining optical and SAR satellite image matching points, characterized in that: The method comprises a readable storage medium having a computer program stored thereon, and when the computer program is executed, the method for refining the position of optical and SAR satellite image matching points according to any one of claims 1 to 8 is implemented.

Images