High-resolution SAR satellite image line segment extraction method

Abstract

This application relates to a method for extracting line segments from high-resolution SAR satellite images, comprising: processing the image into an amplitude image; obtaining a seed point set based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding; obtaining a candidate growth point set based on constant false alarm rate detection, non-maximum suppression, and adaptive amplitude thresholding of the amplitude image; constructing a circular neighborhood window centered on each point in the seed point and the candidate growth point set; obtaining the principal direction of each point based on the pixel amplitude value of the pixels within the circular neighborhood window of each point through weighted principal component analysis; for each seed point, determining a first growth point based on the candidate growth point set using spatial distance and direction consistency constraints; and determining a current second growth point using spatial distance, direction consistency, and fitting direction vector change constraints. This method can achieve high-precision line segment extraction from linear features composed of discrete strong scattering points in the image.

Description

Technical Field

[0001] This application relates to the field of high-resolution SAR satellite image information extraction technology, and in particular to a method for extracting line segments from high-resolution SAR satellite images. Background Technology

[0002] With the development of high-resolution synthetic aperture radar (SAR) satellite technology, the image data provided by high-resolution SAR satellites possesses high spatial resolution. High-resolution SAR satellite imagery contains numerous linear features composed of discrete strong scattering points, primarily originating from strong reflections from structures such as building facades and bridges. Extracting line segments from these linear features yields valuable applications in areas such as building structure analysis and 3D modeling.

[0003] Existing image line segment extraction methods mainly include gradient-based edge detection and Hough transform based on the transform domain. These existing methods mostly originate from the field of computer vision and can only perform high-precision line segment extraction on linear features of planar target boundaries in SAR satellite images. They cannot perform high-precision line segment extraction on linear features composed of discrete strong scattering points in high-resolution SAR satellite images. Summary of the Invention

[0004] This application provides a method for extracting line segments from high-resolution SAR satellite images, which can achieve high-precision extraction of line segments from linear features composed of discrete strong scattering points in high-resolution SAR satellite images.

[0005] In a first aspect, embodiments of this application provide a method for extracting line segments from high-resolution SAR satellite images, including:

[0006] Acquire high-resolution SAR satellite imagery and process it into an amplitude image; obtain a seed point set based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding; obtain a candidate growth point set based on non-maximum suppression and adaptive amplitude thresholding of the amplitude image.

[0007] With each point in the seed point set and the candidate growth point set as the center, a circular neighborhood window is constructed. Based on the pixel amplitude value of the pixels in the circular neighborhood window of each point, the principal direction of each seed point in the seed point set and the principal direction of each candidate growth point in the candidate growth point set are obtained through weighted principal component analysis.

[0008] For each seed point, based on the main direction of the seed point and the main direction of the candidate growth points, and using spatial distance and direction consistency constraints, a first growth point is determined from the set of candidate growth points to generate an initial fitted line segment; Step-size growth step: Step-size growth is performed along the initial fitted line segment or the two opposite directions of the current fitted line segment to obtain the current step-size growth line segment; Second growth point determination step: Based on the endpoints of the current step-size growth line segment, and using spatial distance, direction consistency, and fitting direction vector change constraints, a current second growth point is determined from the set of candidate growth points to generate the current fitted line segment; The step-size growth step and the current second growth point determination step are repeated until no new second growth point can be found.

[0009] Secondly, embodiments of this application provide another method for extracting line segments from high-resolution SAR satellite images, including:

[0010] Acquire high-resolution SAR satellite imagery and process it into an amplitude image; obtain a seed point set based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding; obtain a candidate growth point set based on non-maximum suppression and adaptive amplitude thresholding of the amplitude image.

[0011] With each point in the seed point set and the candidate growth point set as the center, a circular neighborhood window is constructed. Based on the pixel amplitude value of the pixels in the circular neighborhood window of each point, the principal direction of each seed point in the seed point set and the principal direction of each candidate growth point in the candidate growth point set are obtained through weighted principal component analysis.

[0012] For each seed point, based on the main direction of the seed point and the main direction of the candidate growth points, a first growth point is determined from the set of candidate growth points using spatial distance and direction consistency constraints. Based on the first growth point, a second growth point is determined from the set of candidate growth points using spatial distance, direction consistency, and fitting direction vector change constraints, generating the current fitted line segment. The new second growth point determination step is as follows: based on the current second growth point and the current fitted line segment, a new second growth point is determined from the set of candidate growth points using spatial distance, direction consistency, and fitting direction vector change constraints, generating a new fitted line segment. The new second growth point determination step is repeated until no new second growth point can be found.

[0013] Thirdly, embodiments of this application provide a high-resolution SAR satellite image line segment extraction system, including:

[0014] The growth point determination module is used to: acquire high-resolution SAR satellite images and process the images into amplitude images; obtain a seed point set based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding; and obtain a candidate growth point set based on non-maximum suppression and adaptive amplitude thresholding of the amplitude image.

[0015] The main direction determination module for growth points is used to: construct an annular neighborhood window centered on each point in the seed point set and the candidate growth point set; and obtain the main direction of each seed point in the seed point set and the main direction of each candidate growth point in the candidate growth point set through weighted principal component analysis based on the pixel amplitude value of the pixels in the annular neighborhood window of each point.

[0016] The line segment growth module is used for: for each seed point, based on the main direction of the seed point and the main direction of the candidate growth points, using spatial distance and direction consistency constraints, determining a first growth point from the set of candidate growth points to generate an initial fitted line segment; a step-size growth step: performing step-size growth along the initial fitted line segment or the opposite direction of the current fitted line segment to obtain the current step-size grown line segment; a second growth point determination step: based on the endpoints of the current step-size grown line segment, using spatial distance, direction consistency, and fitting direction vector change constraints, determining a current second growth point from the set of candidate growth points to generate the current fitted line segment; repeating the step-size growth step and the current second growth point determination step until no new second growth point can be found.

[0017] Fourthly, embodiments of this application provide a high-resolution SAR satellite image line segment extraction system, comprising:

[0018] The growth point determination module is used to: acquire high-resolution SAR satellite images and process the images into amplitude images; obtain a seed point set based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding; and obtain a candidate growth point set based on non-maximum suppression and adaptive amplitude thresholding of the amplitude image.

[0019] The main direction determination module for growth points is used to: construct an annular neighborhood window centered on each point in the seed point set and the candidate growth point set; and obtain the main direction of each seed point in the seed point set and the main direction of each candidate growth point in the candidate growth point set through weighted principal component analysis based on the pixel amplitude value of the pixels in the annular neighborhood window of each point.

[0020] The line segment growth module is used for: for each seed point, determining a first growth point from the set of candidate growth points based on the main direction of the seed point and the main direction of the candidate growth points, using spatial distance and direction consistency constraints; determining a second growth point from the set of candidate growth points based on the first growth point, using spatial distance, direction consistency, and fitting direction vector change constraints, and generating the current fitted line segment; the new second growth point determination step: determining a new second growth point from the set of candidate growth points based on the current second growth point and the current fitted line segment, using spatial distance, direction consistency, and fitting direction vector change constraints, and generating a new fitted line segment; repeating the new second growth point determination step until no new second growth point can be found.

[0021] Fifthly, embodiments of this application provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the high-resolution SAR satellite image line segment extraction method provided in the first or second aspect.

[0022] In a sixth aspect, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the high-resolution SAR satellite image line segment extraction method provided in the first or second aspect.

[0023] In a seventh aspect, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the high-resolution SAR satellite image line segment extraction method provided in the first or second aspect.

[0024] The method provided in this application, in a first aspect, utilizes constant false alarm rate (CFAR) detection to identify candidate strong scattering points in an image; then, it filters candidate growth points from these candidate strong scattering points using non-maximum suppression (NMS); further, it thresholds corner response values ​​to filter out noise points from the candidate strong scattering points, retaining the candidate strong scattering points as target strong scattering points. These target strong scattering points are used as seed points for line segment extraction of linear features composed of discrete strong scattering points in the image; and for all pixels in the image, it filters out points with maximum amplitude and points with amplitude thresholds greater than a set threshold to obtain a set of candidate growth points.

[0025] Secondly, for each seed point and candidate growth point, based on the pixel amplitude value of the pixel points within the constructed annular neighborhood window, the principal direction of the seed point is obtained through weighted principal component analysis (PCA). The principal direction of the seed point is used to provide directional constraints for the extraction of linear features composed of discrete strong scattering points in the image.

[0026] Thirdly, each seed point is used as the starting point for growth, and candidate growth points are used as the growth objects. Line segments are grown along the main direction and the opposite direction of the seed point. Based on spatial distance, directional consistency and fitting direction vector change constraints, new growth points are searched from the candidate growth points to grow line segments. The direction of the line segments is updated by fitting the current set of growth points with a straight line until the growth of each line segment terminates, thus obtaining a set of candidate line segments. This enables the complete extraction of line segments from the linear features composed of discrete strong scattering points in the image.

[0027] Fourthly, the direction consistency of each candidate line segment in the candidate line segment set is verified, and the candidate line segments with direction consistency are selected to obtain the target line segment set, thereby realizing high-precision line segment extraction of linear features composed of discrete strong scattering points in the image.

[0028] In summary, the method provided in this application can effectively suppress the influence of noise in the image and improve the accuracy of line segment extraction in the process of extracting line segments of linear features composed of discrete strong scattering points in the image, thereby achieving high-precision extraction of linear structures of buildings. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] To gain a more complete understanding of this application and its beneficial effects, the following description will be provided in conjunction with the accompanying drawings, wherein the same reference numerals in the following description denote the same parts.

[0031] Figure 1 This is a flowchart illustrating a high-resolution SAR satellite image line segment extraction method provided in this application embodiment.

[0032] Figure 2 This is an example of a high-resolution SAR satellite image of a certain urban area provided in an embodiment of this application.

[0033] Figure 3 The embodiments provided in this application are based on Figure 2 A schematic diagram of the seed points obtained from the image shown.

[0034] Figure 4 This is a schematic diagram of the ring-shaped neighborhood window provided in the embodiments of this application.

[0035] Figure 5 The embodiments provided in this application are based on Figure 2The image shown is an estimate of the principal direction of the seed point obtained from the image.

[0036] Figure 6 The embodiments provided in this application are based on Figure 2 The image shown is a diagram of the grouped growth results obtained from the image.

[0037] Figure 7 The embodiments provided in this application are based on Figure 2 The image shown is a diagram of the effective line segment extraction results.

[0038] Figure 8 Different methods are used in the embodiments of this application to Figure 2 The image shown is a comparison of line segment extraction.

[0039] Figure 9 This is a schematic diagram of a high-resolution SAR satellite image line segment extraction system provided in the embodiments of this application.

[0040] Figure 10 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0041] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the protection scope of this application.

[0042] Figure 1 This is a flowchart illustrating a high-resolution SAR satellite image line segment extraction method provided in this application embodiment. Please refer to... Figure 1 This application provides a method for extracting line segments from high-resolution SAR satellite images, including steps 100 to 300, which will be described in detail below.

[0043] Step 100: Acquire high-resolution SAR satellite imagery and process it into an amplitude image; obtain a seed point set based on constant false alarm rate detection, non-maximum suppression, and corner response value thresholding; obtain a candidate growth point set based on non-maximum suppression and adaptive amplitude thresholding of the amplitude image.

[0044] Step 200: Construct a circular neighborhood window centered on each point in the seed point set and the candidate growth point set. Based on the pixel amplitude value of each pixel in the circular neighborhood window, obtain the principal direction of each seed point in the seed point set and the principal direction of each candidate growth point in the candidate growth point set through weighted principal component analysis.

[0045] Step 300: For each seed point, based on the main direction of the seed point and the main direction of the candidate growth points, and using spatial distance and direction consistency constraints, determine the first growth point from the candidate growth point set to generate an initial fitted line segment; Step-size growth step: Perform step-size growth along the initial fitted line segment or the opposite direction of the current fitted line segment to obtain the current step-size growth line segment; Second growth point determination step: Based on the endpoints of the current step-size growth line segment, and using spatial distance, direction consistency, and fitting direction vector change constraints, determine the current second growth point from the candidate growth point set to generate the current fitted line segment; Repeat the step-size growth step and the current second growth point determination step until no new second growth point can be found.

[0046] The method provided in this application, in a first aspect, utilizes constant false alarm rate (CFAR) detection to identify candidate strong scattering points in an image; selects candidate growth points from the candidate strong scattering points using non-maximum suppression (NMS); further, it filters out noise points from the candidate strong scattering points by thresholding corner response values, retaining the candidate strong scattering points as target strong scattering points. These target strong scattering points are used as seed points for line segment extraction of linear features composed of discrete strong scattering points in the image; and for all pixels in the image, it selects points with maximum amplitude and points with amplitude thresholds greater than a set threshold to obtain a set of candidate growth points.

[0047] Secondly, for each seed point and candidate growth point, based on the pixel amplitude value of the pixel points within the constructed annular neighborhood window, the principal direction of the seed point is obtained through weighted principal component analysis (PCA). The principal direction of the seed point is used to provide directional constraints for the extraction of linear features composed of discrete strong scattering points in the image.

[0048] Thirdly, each seed point is used as the starting point for growth, and candidate growth points are used as the growth objects. Line segments are grown along the main direction and the opposite direction of the seed point. Based on spatial distance, directional consistency and fitting direction vector change constraints, new growth points are searched from the candidate growth points to grow line segments. The direction of the line segments is updated by fitting the current set of growth points with a straight line until the growth of each line segment terminates, thus obtaining a set of candidate line segments. This enables the complete extraction of line segments from the linear features composed of discrete strong scattering points in the image.

[0049] Fourthly, the direction consistency of each candidate line segment in the candidate line segment set is verified, and the candidate line segments with direction consistency are selected to obtain the target line segment set, thereby realizing high-precision line segment extraction of linear features composed of discrete strong scattering points in the image.

[0050] In summary, the method provided in this application can effectively suppress the influence of noise in the image and improve the accuracy of line segment extraction during the extraction of line segments of linear features composed of discrete strong scattering points in the image, thereby achieving high-precision extraction of linear structures of buildings.

[0051] It should be noted that, in the embodiments of this application, the line segment accuracy reflects the degree to which the line segment extraction result approximates the real structure.

[0052] The following description Figure 1 Further optional specific embodiments for each step performed during the high-resolution SAR satellite image line segment extraction processing in the example.

[0053] Step 100: Acquire high-resolution SAR satellite imagery and process it into an amplitude image; obtain a seed point set based on constant false alarm rate detection, non-maximum suppression, and corner response value thresholding; obtain a candidate growth point set based on non-maximum suppression and adaptive amplitude thresholding of the amplitude image.

[0054] In some embodiments, the step of selecting a set of seed points from an amplitude image based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding includes:

[0055] The pixels in the amplitude image are detected one by one through the detection window. For each pixel to be detected, all pixels in the non-protected area of ​​the detection window are sorted according to their pixel amplitude values. The local pixel amplitude threshold is obtained based on the sorted sequence and the empirical quantile ratio. If the pixel amplitude value of the pixel to be detected is greater than the local pixel amplitude threshold, the pixel to be detected is determined as a candidate target pixel, and a set of candidate target pixels is obtained.

[0056] In the third defined neighborhood of each candidate target pixel, the pixel amplitude values ​​of all the candidate target pixels are compared, and only the candidate target pixel with the largest pixel amplitude value is retained as the candidate feature point to obtain the candidate feature point set.

[0057] Calculate the corner response value of each candidate feature point; compare the corner response value of each candidate feature point with the corner response threshold, and retain the candidate feature points whose corner response values ​​are greater than the set threshold as seed points to obtain the seed point set.

[0058] It is understood that in this embodiment, candidate target pixels can be obtained through constant false alarm rate detection. These candidate target pixels are strong scattering points in the image. Through non-maximum suppression, only the pixel with the largest pixel amplitude value in the third set neighborhood of the candidate target pixels is retained to obtain a set of candidate feature points, which is used to highlight isolated scattering structures. Through non-maximum suppression, noise points are filtered out for the initial location of irregular structures in the image.

[0059] It is understandable that the detection window is a local neighborhood closely surrounding the pixel to be detected. This local neighborhood includes a protected area and an unprotected area. The protected area is set closely around the pixel to be detected, and the unprotected area is set closely around the protected area.

[0060] For example, please refer to Figure 2 , Figure 2 This is a legend for a high-resolution SAR satellite image of a certain urban area. In constant false alarm rate (CFAR) detection, each pixel in the image is detected, and a detection window is established centered on the pixel to be detected. Within this detection window, pixels within the protected area are removed, while pixels outside the protected area are retained. The pixels outside the protected area are then sorted in ascending order of pixel amplitude value. Combined with empirical quantile ratios, the local pixel amplitude threshold for the pixel to be detected is obtained. The formula for calculating the local pixel amplitude threshold is:

[0061] ，

[0062] in: For the pixel to be detected Local pixel amplitude threshold; A set representing the pixel amplitude values ​​of pixels located outside the protected area; For the pixel to be detected The set of pixels within the detection window For the pixel to be detected The collection of pixels within the non-protected area of ​​the detection window; From The middle screening satisfies A set; This is a sorting function used to obtain a sorted sequence; This is an indexing operator used to retrieve the element at a specific position from a sorted sequence. Indicates rounding down to the nearest integer. The empirical quantile ratio, This represents the number of pixels outside the protected area.

[0063] If the pixel amplitude value of the pixel to be detected satisfy If so, the pixel to be detected is determined as a constant false alarm rate detection response point, which is also a candidate target pixel, or a strong scattering point.

[0064] For example, the empirical quantile ratio can be 90%. The third set neighborhood of the candidate target pixel can be a 3×3 neighborhood.

[0065] In some embodiments, the step of thresholding the corner response values ​​includes:

[0066] Calculate the corner response value of each candidate feature point; compare the corner response value of each candidate feature point with the corner response threshold, and retain the candidate feature points whose corner response values ​​are greater than the set threshold as seed points.

[0067] It is understood that in this embodiment, the corner response value is calculated and compared with a set threshold to evaluate the quality of candidate feature points, eliminate candidate feature points with low corner response caused by noise and clutter, and retain stable and reliable strong scattering candidate feature points as seed points for line segment growth.

[0068] For example, corner response values ​​can be obtained using any one of the following methods: Harris response function, Shi-Tomasi corner detection, and FAST corner detection. This application Figures 2 to 7 In a specific embodiment, the corner response value is obtained through the Harris response function.

[0069] For example, when calculating the corner response values ​​of candidate feature points using the Harris response function, the expression for the Harris response function is as follows:

[0070] ,

[0071] ,

[0072] Where R is the corner response value of the candidate feature point to be calculated, and M is the structure tensor calculated from the local image gradient of the candidate feature point. This is an empirical coefficient. . and These represent the gradients in the x and y directions, respectively. Please refer to [link / reference]. Figure 3 , Figure 3 The embodiments provided in this application are based on Figure 2 A schematic diagram of the seed points obtained from the image shown.

[0073] In some embodiments, the step of obtaining a candidate growth point set based on non-maximum suppression and an adaptive amplitude threshold of the amplitude image includes:

[0074] In the fourth defined neighborhood of each pixel in the amplitude image, the pixel amplitude values ​​are compared, and only the pixel with the largest pixel amplitude value is retained to obtain the set of candidate pixels.

[0075] By using the adaptive amplitude threshold of the overall amplitude image, the candidate growth point set is further filtered to obtain an optimized candidate growth point set.

[0076] The formula for calculating the adaptive amplitude threshold is:

[0077] ,

[0078] in, For adaptive amplitude threshold, As a control factor, This represents the average background value of the entire image. This represents the background standard deviation for the entire image. For example, The value range is greater than or equal to 2.0 and less than or equal to 3.0.

[0079] For example, for all response points obtained through nonmaximum suppression Preserve pixel amplitude values satisfy response point As a candidate growth point.

[0080] For example, the fourth defined neighborhood of each pixel can be the region composed of four directly adjacent pixels above, below, left, and right of each pixel.

[0081] Understandably, in this embodiment, seed points for line segment growth are further selected from the candidate feature point set obtained through non-maximum suppression. This step aims to eliminate false response points with low signal-to-noise ratio by calculating an adaptive amplitude threshold for the background noise of the entire image, thereby improving the accuracy and reliability of subsequent growth.

[0082] Understandably, the average background value of the entire image is... and the background standard deviation of the entire image The mean background value can be obtained by: identifying the region with the smallest local standard deviation within the entire image; and then using this region as the mean background value for the entire image. and the background standard deviation of the entire image .

[0083] Step 200: Construct a circular neighborhood window centered on each point in the seed point set and the candidate growth point set. Based on the pixel amplitude value of each pixel in the circular neighborhood window, obtain the principal direction of each seed point in the seed point set and the principal direction of each candidate growth point in the candidate growth point set through weighted principal component analysis.

[0084] It is understandable that each seed point has a dominant direction. The dominant direction of a seed point indicates the dominant direction of the gray-scale structure of its neighborhood, providing a structurally stable and locally dominant directional angle value for the seed point. This provides directional guidance for the subsequent line segment growth of the seed point and directional constraints for its subsequent growth.

[0085] It is understandable that each candidate growth point has a principal direction, which is used to indicate the principal direction of the gray-scale structure in the neighborhood of the candidate growth point. This provides a constraint for determining the consistency of the direction of candidate points during the line segment growth process and serves as the basis for subsequent estimation and verification of the direction of the fitted line segment.

[0086] In some embodiments, the step of obtaining the principal direction of each seed point in the seed point set and the principal direction of each candidate growth point in the candidate growth point set through weighted principal component analysis based on the pixel amplitude value of the pixels within the annular neighborhood window of each point includes:

[0087] The steps for calculating the weighted average coordinates are as follows: For the current seed point or the current candidate growth point, obtain the relative coordinates and corresponding pixel amplitude values ​​of multiple sample pixels within the annular neighborhood window. The relative coordinates are the coordinates of multiple sample pixels relative to the current seed point. Using the relative coordinates and corresponding pixel amplitude values, calculate the weighted average coordinates of the multiple sample pixels.

[0088] Steps for constructing the weighted covariance matrix: Decenter the relative coordinates of all sample pixels relative to the weighted mean coordinates; use the decentralized coordinates and the corresponding pixel amplitude values ​​to calculate the weighted covariance matrix;

[0089] Matrix eigenvalue decomposition steps: Perform eigenvalue decomposition on the weighted covariance matrix to obtain an eigenvector matrix composed of eigenvectors and an eigenvalue diagonal matrix composed of eigenvalues;

[0090] The main direction determination steps are as follows: Select the eigenvalue with the largest value from the eigenvalue diagonal matrix, and define the eigenvector corresponding to this eigenvalue in the eigenvector matrix as the target eigenvector; use the target eigenvector as the main direction of the current seed point.

[0091] For example, with the seed point as the center, construct an inner diameter of... =7, outer diameter is A circular neighborhood window. See also Figure 4 , Figure 4 This is a schematic diagram of the ring-shaped neighborhood window provided in the embodiments of this application.

[0092] Satisfy the requirements in the annular region window All pixels are used as the sample pixel sum, where... The coordinates of the target seed point. These are the coordinates of the pixel.

[0093] For example, the set of sample pixels can be defined as:

[0094] ,

[0095] in, For the set of sample pixels, The coordinates of the target seed point. The coordinates of the pixel. The inner diameter of the annular neighborhood window, is the outer diameter of the annular neighborhood window.

[0096] For example, the weighted mean coordinates of the sample pixel set are calculated using the following formula:

[0097] ,

[0098] in, For weighted mean coordinates, For the first The relative coordinates of each sample pixel , The number of sample pixels. For the first The pixel amplitude value of each sample pixel;

[0099] The weighted covariance matrix is ​​calculated using the following formula:

[0100] ,

[0101] in, For the weighted covariance matrix, For weighted mean coordinates, Represents the transpose of a matrix. For the first The relative coordinates of each sample pixel , The number of sample pixels. For the first The pixel amplitude value of each sample pixel;

[0102] The steps for performing eigenvalue decomposition on the weighted covariance matrix include:

[0103] ,

[0104] in, For the weighted covariance matrix, V The eigenvector matrix, D It is an eigenvalue diagonal matrix. Represents the transpose of a matrix;

[0105] The principal direction of the target seed point is expressed as:

[0106] ,

[0107] in, For the eigenvalue diagonal matrix D The eigenvalue with the largest value selected is in the eigenvector matrix. V The corresponding feature vector; for exist Components on the axis, for exist Components on the axis.

[0108] It is understandable that each seed point has a dominant direction. The dominant direction of a seed point indicates the dominant direction of its neighborhood grayscale structure, providing a structurally stable and locally dominant directional angle value. This provides directional guidance for subsequent line segment growth and directional constraints for the seed point's subsequent growth. Please refer to [link to relevant documentation]. Figure 5 , Figure 5 The embodiments provided in this application are based on Figure 2 The image shown is an estimate of the principal orientation of the seed point obtained from the image. Figure 5 The red arrow shown indicates the main direction of the seed point.

[0109] Step 300: For each seed point, based on the main direction of the seed point and the main direction of the candidate growth points, and using spatial distance and direction consistency constraints, determine the first growth point from the candidate growth point set to generate an initial fitted line segment; Step-size growth step: Perform step-size growth along the initial fitted line segment or the opposite direction of the current fitted line segment to obtain the current step-size growth line segment; Second growth point determination step: Based on the endpoints of the current step-size growth line segment, and using spatial distance, direction consistency, and fitting direction vector change constraints, determine the current second growth point from the candidate growth point set to generate the current fitted line segment; Repeat the step-size growth step and the current second growth point determination step until no new second growth point can be found.

[0110] It is understood that step 300 is a line segment growth step. For example, the specific steps of line segment growth may include: current seed point determination step 310, initial candidate growth point determination step 320, step size growth step 330, second growth point determination step 340, current fitting growth step 350, and iterative growth step 360.

[0111] Current seed point determination step 310: Sort the seed points in the seed point set according to the size of the corner response values, and determine the current seed point.

[0112] Initial candidate growth point determination step 320: In the first set neighborhood of the current seed point, search for candidate growth points that meet the first condition and use them as initial candidate growth points. The first condition is: the angle between the main direction of the candidate growth point and the main direction of the seed point is less than the first angle threshold, and the Euclidean distance between the candidate growth point and the target seed point is the smallest.

[0113] Step-size growth step 330: Based on the fitted line segment, which can be either the initial fitted line segment or the current fitted line segment, grow along both the positive and negative directions of the fitted line segment according to the step size. To grow, , This is the step size coefficient. , The length of the initially fitted line segment is used to obtain the line segment with the current step size.

[0114] Step 340 for determining the second growth point: Within the second defined neighborhood of the endpoint of the current step-size growth segment, find candidate growth points that have not yet been added to the current point set and meet the second condition, and use them as the second growth points. The second condition is: the vertical distance from the candidate growth point to the current fitted line segment is less than the vertical distance threshold, and the value of the vertical distance threshold is greater than 1 and less than 3; the angle difference between the main direction of the candidate growth point and the direction of the current fitted line segment is less than the second angle threshold, and the value of the second angle threshold is greater than 20° and less than 30°; and after adding the candidate growth point to the point set, the updated point set is fitted with a straight line to obtain a new fitted line segment; the angle difference between the direction of the new fitted line segment and the direction of the previously fitted line segment is less than the third angle threshold, and the value of the third angle threshold is 20°.

[0115] Current fitting growth step 350: If a second growth point is found, add the second growth point to the current point set to obtain the updated point set; perform line fitting on the updated point set using the least squares method to obtain the current fitted line segment; if no second growth point is found, terminate the growth.

[0116] Iterative growth step 360: After obtaining the fitted line segment, return to the step growth step; if no candidate growth point that meets the second condition is found after returning to the step growth step, return to the step growth step again. When the total number of times the step growth step is returned reaches the preset number of returns, and no candidate growth point that meets the second condition is found, stop returning and terminate the growth.

[0117] Understandably, the set of points formed during the growth of all line segments, along with the orientation information of each point within that set, will serve as a candidate set for subsequent orientation consistency verification of the fitted line segments. Please refer to [link / reference]. Figure 6 , Figure 6 The embodiments provided in this application are based on Figure 2 The image shown is a diagram of the grouped growth results obtained from the image.

[0118] For example, the coordinate calculation formula for the endpoints of the current step-size growth segment is: ,in As the endpoint, For the current set of grown points The first in One point, For the current set of growth points The coordinates of the points; Step size, The unit direction vector of the fitted line segment obtained by fitting a straight line to the current set of points. The calculation formula is:

[0119] ,

[0120] in, The slope of the fitted line segment obtained by fitting a straight line to the current point set; by fitting a straight line to the current point set, a linear model of the fitted line segment is obtained. ,in The x-axis is... The vertical axis is , The slope The intercept is used to obtain the unit direction vector of the fitted line segment. .

[0121] For example, line fitting methods can include least squares method, RANSAC line fitting, or orthogonal fitting method. This application Figures 2 to 7 In a specific embodiment, the least squares method is used for line fitting.

[0122] In some embodiments, for each seed point, the step of determining a first growth point from the set of candidate growth points based on the principal direction of the seed point and the principal direction of the candidate growth points, using spatial distance and orientation consistency constraints, includes:

[0123] Determine the current seed point from the set of seed points;

[0124] Based on the candidate growth point set, within the first defined neighborhood of the current seed point, a candidate growth point that simultaneously satisfies the spatial distance and orientation consistency constraints is identified as the first growth point. The spatial distance and orientation consistency constraints include:

[0125] The angle between the principal direction of the candidate growth point and the principal direction of the current seed point is less than the first angle threshold; and

[0126] The candidate growth point has the smallest Euclidean distance to the current seed point.

[0127] Understandably, the aforementioned spatial distance constraint and directional consistency constraint are used to ensure the consistency of the first growth point and the seed point in spatial position and direction, thereby guaranteeing the continuity and accuracy of line segment growth.

[0128] For example, the first angle threshold The range of values ​​is greater than or equal to and less than or equal to .

[0129] The first defined neighborhood of the seed point can be a region centered at the seed point with a radius of [missing information]. The neighborhood,

[0130] Furthermore, in some embodiments, the step of determining the current seed point from the seed point set includes:

[0131] For each seed point in the seed point set, sort them in descending order of corner response value, and determine the current seed point based on the sorting sequence.

[0132] In some embodiments, the step of determining the current second growth point from the set of candidate growth points based on the endpoints of the current step-size growth segment, using spatial distance constraints, directional consistency constraints, and fitting direction vector change constraints, includes:

[0133] Based on the candidate growth point set, within the second defined neighborhood of the current step-size growth segment, candidate growth points that simultaneously satisfy the constraints of spatial distance, directional consistency, and fitting direction vector change are identified as the current second growth point; wherein, the constraints of spatial distance, directional consistency, and fitting direction vector change include:

[0134] The vertical distance from the candidate growth point to the current fitted line segment is less than the vertical distance threshold, and the value of the vertical distance threshold is greater than 1 and less than 3.

[0135] The angle difference between the principal direction of the candidate growth point and the direction of the currently fitted line segment is less than the second angle threshold; and

[0136] After adding candidate growth points to the point set, the updated point set is fitted with a straight line to obtain a new fitted line segment; the angle difference between the direction of the new fitted line segment and the direction of the previously fitted line segment is less than the third angle threshold.

[0137] For example, the vertical distance threshold can be greater than 1 and less than 3. The second angle threshold and the third angle threshold can be greater than 20° and less than 30°.

[0138] It is understandable that the above three constraints—spatial distance constraint, directional consistency constraint, and fitting direction vector change constraint—are used to ensure the stable growth of line segments with consistent direction, avoid structural breakage, and ensure the stability and accuracy of line segment extraction.

[0139] Furthermore, in some embodiments, after the step of finding candidate growth points that simultaneously satisfy the constraints of spatial distance, directional consistency, and fitting direction vector change, the method further includes:

[0140] If no candidate growth point that simultaneously satisfies the constraints of spatial distance, directional consistency, and fitting direction vector change is found in the second set neighborhood, the process jumps to the step size growth step. When the number of jumps reaches the first threshold, the growth process terminates.

[0141] If a candidate growth point that simultaneously satisfies the constraints of spatial distance, directional consistency, and change of fitting direction vector is found within the second defined neighborhood, the current seed point, the first growth point, and all found second growth points are constructed into the current point set. Based on the current point set, a straight line is fitted to generate the current fitted line segment.

[0142] For example, the preset number of returns can be set to 3.

[0143] Furthermore, in some embodiments, after the step of finding candidate growth points that simultaneously satisfy the constraints of spatial distance, directional consistency, and fitting direction vector change, the method further includes:

[0144] If a candidate growth point is found within the second defined neighborhood that simultaneously satisfies the constraints of spatial distance, directional consistency, and the change in the fitted direction vector, then...

[0145] Candidate growth points that satisfy the constraints of spatial distance, directional consistency, and fitting direction vector change are identified as potential growth points. These potential growth points are then checked, including:

[0146] Obtain the current point set corresponding to the currently fitted line segment; determine whether the candidate point belongs to the current point set; if it does, do not determine the candidate point as the current second growth point, and continue to search for new candidate growth points; if it does not belong, determine the candidate point as the current second growth point.

[0147] In some embodiments, the step of growing the current step size line segment by performing step size growth in both directions along the initial fitted line segment or the current fitted line segment includes:

[0148] When the fitted line segment is the initial fitted line segment, proceed along both the positive and negative directions of the initial fitted line segment, according to the step size. To grow, , This is the step size coefficient. The length of the initially fitted line segment;

[0149] When the fitted line segment is the current fitted line segment, proceed along both the positive and negative directions of the current fitted line segment, according to the step size. To grow, , This is the step size coefficient. The length of the initially fitted line segment.

[0150] For example, The range of values ​​can be .

[0151] For example, step size This is a multiple of the length of the initially fitted line segment. Step size. The calculation formula is: in, This is the step size coefficient. , The initial fitted line segment length, , Indicates the initial candidate growth point. Indicates the target seed point. Indicates from point to The length of the displacement vector also represents the length of the initially fitted line segment.

[0152] In some embodiments, after repeating the step-size growth step and the current second growth point determination step until no new second growth point can be found, the method further includes: performing direction consistency verification on the current fitted line segment obtained based on each seed point, including:

[0153] Based on the current seed point and the corresponding currently fitted line segment, construct the current point set by the current seed point, the first growth point, and all found second growth points; perform line fitting on the current point set to obtain the current fitting direction;

[0154] Calculate the angle difference between the principal direction and the current fitted direction for each point in the current point set, count the number of points whose angle difference is less than the fourth angle threshold, count the total number of points in the current point set, and divide the number of points by the total number to obtain the first ratio.

[0155] If the first ratio is greater than or equal to the ratio threshold, the current fitted line segment is determined to be a valid line segment and retained; if the first ratio is less than the ratio threshold, the current fitted line segment is determined to be an invalid line segment and discarded.

[0156] For example, the set of all points in the currently fitted line segment is called the current point set;

[0157] For the current point set First, the least squares method is used to determine the spatial position of each point in the current point set. Perform linear fitting to obtain the current fitted line. Calculate the current fitting direction of the currently fitted line. For nearly perpendicular lines, define .

[0158] For each point in the current point set, calculate the principal direction of that point. Current fitting direction of the fitted line angular difference :

[0159] ,

[0160] in, |·| represents taking the minimum value, and |·| represents taking the absolute value. If the angle difference at that point... Less than the fourth angle threshold , in The range of values ​​is greater than and less than If a point is found to be aligned with the overall structural direction of the currently fitted line, then that point is considered a point with consistent direction. It can be understood that the current fitting direction is used to characterize the overall structural direction of the currently fitted line.

[0161] Count the number of points in the current point set that have the same direction across all directions. And calculate the first ratio, which is the directional consistency ratio of the current point set:

[0162] ,

[0163] in, The directional consistency ratio of the current point set. This represents the number of points in the current point set that share the same direction. This represents the number of points in the current point set.

[0164] If the current point set has a consistent direction ratio ,in For example, a ratio threshold. If the value is 0.8, the currently fitted line segment is determined to be a valid line segment and is retained; if... If the current fitted line segment is invalid, it will be discarded.

[0165] Ultimately, all currently fitted line segments that pass the orientation consistency verification constitute the effective linear structure feature set extracted from the image.

[0166] Figure 7 The embodiments provided in this application are based on Figure 2 The image shown is a diagram of the effective line segment extraction results.

[0167] Figure 8 This section compares the performance of LSDSAR, POE, Hough, and EDLine algorithms on the same image. LSDSAR (Line Segment Detector for SAR images) is a line segment detection algorithm for SAR images; POE (Pixel Orientation Estimation) is a pixel orientation estimation algorithm; Hough is a Hough transform algorithm; and EDLines (Edge Drawing Lines) is a line segment detection algorithm based on edge drawing.

[0168] By Figure 8 and Figure 7 as well as Figure 1 A comparison of the original images shows that the line segments extracted using the method provided in this application have higher accuracy and precision, resulting in a more complete and clearer building structure. This verifies the reliability of the method provided in this application for extracting linear structures from high-resolution SAR satellite imagery. The line segment accuracy reflects the proportion of valid line segments in the extracted results; the line segment precision reflects the degree to which the extracted results approximate the actual structure.

[0169] Please see Figure 9 This application also provides a high-resolution SAR satellite image line segment extraction system, including: a growth point determination module 901, a growth point main direction determination module 902, and a line segment growth module 903.

[0170] The growth point determination module 901 is used to: acquire high-resolution SAR satellite images and process the images into amplitude images; filter candidate growth point sets from the amplitude images based on constant false alarm rate detection and non-maximum suppression; and filter seed point sets from the candidate growth point sets by thresholding the response values ​​of the corner points.

[0171] The main direction determination module 902 for growth points is used to: construct an annular neighborhood window centered on each point in the seed point set and the candidate growth point set; and obtain the main direction of each seed point in the seed point set and the main direction of each candidate growth point in the candidate growth point set through weighted principal component analysis based on the pixel amplitude value of the pixels in the annular neighborhood window of each point.

[0172] The line segment growth module 903 is used for: for each seed point, based on the main direction of the seed point and the main direction of the candidate growth points, using spatial distance and direction consistency constraints, determining the first growth point from the candidate growth point set, and generating an initial fitted line segment; step-size growth step: performing step-size growth along the initial fitted line segment or the opposite direction of the current fitted line segment to obtain the current step-size growth line segment; second growth point determination step: based on the endpoint of the current step-size growth line segment, using spatial distance, direction consistency, and fitting direction vector change constraints, determining the current second growth point from the candidate growth point set, and generating the current fitted line segment; repeating the step-size growth step and the current second growth point determination step until no new second growth point can be found.

[0173] This application also provides another method for extracting line segments from high-resolution SAR satellite images, including:

[0174] Acquire high-resolution SAR satellite imagery and process it into an amplitude image; obtain a seed point set based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding; obtain a candidate growth point set based on non-maximum suppression and adaptive amplitude thresholding of the amplitude image.

[0175] With each point in the seed point set and the candidate growth point set as the center, a circular neighborhood window is constructed. Based on the pixel amplitude value of the pixels in the circular neighborhood window of each point, the principal direction of each seed point in the seed point set and the principal direction of each candidate growth point in the candidate growth point set are obtained through weighted principal component analysis.

[0176] For each seed point, based on the main direction of the seed point and the main direction of the candidate growth points, a first growth point is determined from the set of candidate growth points using spatial distance and direction consistency constraints. Based on the first growth point, a second growth point is determined from the set of candidate growth points using spatial distance, direction consistency, and fitting direction vector change constraints, generating the current fitted line segment. The new second growth point determination step is as follows: based on the current second growth point and the current fitted line segment, a new second growth point is determined from the set of candidate growth points using spatial distance, direction consistency, and fitting direction vector change constraints, generating a new fitted line segment. The new second growth point determination step is repeated until no new second growth point can be found.

[0177] This application also provides another high-resolution SAR satellite image line segment extraction system, including:

[0178] The growth point determination module is used to: acquire high-resolution SAR satellite images and process the images into amplitude images; obtain a seed point set based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding; and obtain a candidate growth point set based on non-maximum suppression and adaptive amplitude thresholding of the amplitude image.

[0179] The main direction determination module for growth points is used to: construct an annular neighborhood window centered on each point in the seed point set and the candidate growth point set; and obtain the main direction of each seed point in the seed point set and the main direction of each candidate growth point in the candidate growth point set through weighted principal component analysis based on the pixel amplitude value of the pixels in the annular neighborhood window of each point.

[0180] The line segment growth module is used for: for each seed point, determining a first growth point from the set of candidate growth points based on the main direction of the seed point and the main direction of the candidate growth points, using spatial distance and direction consistency constraints; determining a second growth point from the set of candidate growth points based on the first growth point, using spatial distance, direction consistency, and fitting direction vector change constraints, and generating the current fitted line segment; the new second growth point determination step: determining a new second growth point from the set of candidate growth points based on the current second growth point and the current fitted line segment, using spatial distance, direction consistency, and fitting direction vector change constraints, and generating a new fitted line segment; repeating the new second growth point determination step until no new second growth point can be found.

[0181] In practical applications, the aforementioned system can be a terminal device or a chip applied to a terminal device. In this application, the system can implement the functions of multiple units through software, hardware, or a combination of both, enabling the system to execute the high-resolution SAR satellite image line segment extraction method steps provided in any of the above embodiments. Furthermore, the technical effects of each technical solution in this system can be referenced to the technical effects of the corresponding technical solutions in the high-resolution SAR satellite image line segment extraction method, and will not be elaborated upon further in this application.

[0182] Figure 10 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application.

[0183] Based on the hardware implementation of each unit in the above system, embodiments of this application also provide an electronic device, such as... Figure 8 As shown, the electronic device 10 includes a memory 110 and a processor 120. The memory 110 stores a computer program, and the processor 120 executes the computer program to implement the steps of the high-resolution SAR satellite image line segment extraction method provided in any of the above embodiments.

[0184] Of course, in practical applications, such as Figure 8 As shown, the various components in the electronic device 10 are coupled together via a bus system 130. It is understood that the bus system 130 is used to enable communication between these components. In addition to a data bus, the bus system 130 also includes a power bus, a control bus, and a status signal bus. However, for clarity, all buses are labeled as bus system 130 in the figure.

[0185] In practical applications, the aforementioned processor can be at least one of the following: Application-Specific Integrated Circuit (ASIC), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field-Programmable Gate Array (FPGA), controller, microcontroller, and microprocessor. It is understood that, for different devices, the electronic devices used to implement the functions of the aforementioned processor can also be other types, and the embodiments of this application do not specifically limit this.

[0186] The aforementioned memory can be volatile memory, such as random-access memory (RAM); or non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); or a combination of the above types of memory, and provides instructions and data to the processor.

[0187] The electronic devices described in the embodiments of this application can be terminal devices or chips applied to terminal devices. The terminal devices described in the embodiments of this application can include devices such as mobile phones, tablet computers, laptops, and PDAs.

[0188] In an exemplary embodiment, this application also provides a computer-readable storage medium, such as a memory including a computer program, which can be executed by a processor of an electronic device to perform the steps of the aforementioned method.

[0189] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any one of the embodiments of this application.

[0190] Optionally, the computer program product can be applied to the electronic device in the embodiments of this application, and the computer program instructions cause the computer to execute the corresponding processes implemented by the electronic device in the various methods of the embodiments of this application. For the sake of brevity, they will not be described in detail here.

[0191] This application also provides a computer program.

[0192] Optionally, the computer program can be applied to the electronic device in the embodiments of this application. When the computer program is run on a computer, it causes the computer to execute the corresponding processes implemented by the electronic device in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

[0193] It should be understood that in the embodiments of this application, data such as user information are involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0194] It should be understood that the terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items. The expressions “having,” “may have,” “comprising,” and “including,” or “may include” and “may contain” used herein may be used to indicate the presence of a corresponding feature (e.g., an element such as a number, function, operation, or component), but do not exclude the presence of additional features.

[0195] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another, and are not necessarily used to describe a specific order or sequence. For example, without departing from the scope of this invention, first information may also be referred to as second information, and similarly, second information may also be referred to as first information.

[0196] The technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0197] In the several embodiments provided in this application, it should be understood that the disclosed methods, apparatus, and devices can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical, or other forms.

[0198] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0199] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0200] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for extracting line segments from high-resolution SAR satellite images, characterized in that, include: Acquire high-resolution SAR satellite imagery and process the imagery into amplitude images; Based on constant false alarm rate detection, nonmaximum suppression, and thresholding of diagonal response values, a seed point set is obtained; A set of candidate growth points is obtained based on nonmaximum suppression and adaptive amplitude thresholding of the amplitude image; With each point in the seed point set and the candidate growth point set as the center, a circular neighborhood window is constructed. Based on the pixel amplitude value of the pixels in the circular neighborhood window of each point, the principal direction of each seed point in the seed point set and the principal direction of each candidate growth point in the candidate growth point set are obtained through weighted principal component analysis. For each seed point, based on the main direction of the seed point and the main direction of the candidate growth points, the first growth point is determined from the set of candidate growth points using spatial distance and direction consistency constraints, and an initial fitted line segment is generated. Step-size growth step: Perform step-size growth along the initial fitted line segment or the opposite direction of the current fitted line segment to obtain the current step-size growth line segment; The second growth point determination step is as follows: Based on the endpoints of the current step-size growth segment, the current second growth point is determined from the candidate growth point set using spatial distance, directional consistency, and fitting direction vector change constraints, and the current fitting segment is generated; the step-size growth step and the current second growth point determination step are repeated until no new second growth point can be found.

2. The high-resolution SAR satellite image line segment extraction method according to claim 1, characterized in that, The step of obtaining the seed point set based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding includes: The pixels in the amplitude image are detected one by one through the detection window. For each pixel to be detected, all pixels in the non-protected area of ​​the detection window are sorted according to their pixel amplitude values. The local pixel amplitude threshold is obtained based on the sorted sequence and the empirical quantile ratio. If the pixel amplitude value of the pixel to be detected is greater than the local pixel amplitude threshold, the pixel to be detected is determined as a candidate target pixel, and a set of candidate target pixels is obtained. Within the third defined neighborhood of each candidate target pixel, the pixel amplitude values ​​of all pixels that are determined to be candidate target pixels are compared, and only the candidate target pixel with the largest pixel amplitude value is retained as a candidate feature point to obtain the candidate feature point set. Calculate the corner response value of each candidate feature point; compare the corner response value of each candidate feature point with the corner response threshold, and retain the candidate feature points whose corner response values ​​are greater than the set threshold as seed points to obtain the seed point set.

3. The high-resolution SAR satellite image line segment extraction method according to claim 1, characterized in that, The step of obtaining the candidate growth point set based on nonmaximum suppression and adaptive amplitude thresholding of the amplitude image includes: In the fourth defined neighborhood of each pixel in the amplitude image, the pixel amplitude values ​​are compared, and only the pixel with the largest pixel amplitude value is retained to obtain the set of candidate pixels. The set of candidate pixels is further filtered using the adaptive amplitude threshold of the overall amplitude image to obtain an optimized set of candidate growth points. The formula for calculating the adaptive amplitude threshold is: ， in, For adaptive amplitude threshold, As a control factor, This represents the average background value of the entire image. This represents the background standard deviation of the entire image.

4. The high-resolution SAR satellite image line segment extraction method according to claim 1, characterized in that, The step of obtaining the principal direction of each seed point in the seed point set and the principal direction of each candidate growth point in the candidate growth point set by weighted principal component analysis based on the pixel amplitude value of each pixel within the annular neighborhood window of each point includes: The steps for calculating the weighted average coordinates are as follows: For the current seed point or the current candidate growth point, obtain the relative coordinates and corresponding pixel amplitude values ​​of multiple sample pixels within the annular neighborhood window. The relative coordinates are the coordinates of the multiple sample pixels relative to the current seed point. Using the relative coordinates and the corresponding pixel amplitude values, calculate the weighted average coordinates of the multiple sample pixels. The steps for constructing the weighted covariance matrix are as follows: The relative coordinates of all sample pixels are decentered relative to the weighted mean coordinates; the weighted covariance matrix is ​​calculated using the decentered coordinates and the corresponding pixel amplitude values. Matrix eigenvalue decomposition steps: Perform eigenvalue decomposition on the weighted covariance matrix to obtain an eigenvector matrix composed of eigenvectors and an eigenvalue diagonal matrix composed of eigenvalues; Main direction determination steps: Select the eigenvalue with the largest value from the eigenvalue diagonal matrix, define the eigenvector corresponding to the eigenvalue in the eigenvector matrix as the target eigenvector; use the target eigenvector as the main direction of the current seed point.

5. The high-resolution SAR satellite image line segment extraction method according to claim 1, characterized in that, The step of determining the first growth point from the set of candidate growth points for each seed point, based on the principal direction of the seed point and the principal direction of the candidate growth points, using spatial distance and direction consistency constraints, includes: Determine the current seed point from the set of seed points; Based on the candidate growth point set, within the first defined neighborhood of the current seed point, a candidate growth point that simultaneously satisfies the spatial distance and orientation consistency constraints is identified as the first growth point; the spatial distance and orientation consistency constraints include: The angle between the principal direction of the candidate growth point and the principal direction of the current seed point is less than the first angle threshold; and The candidate growth point has the smallest Euclidean distance to the current seed point.

6. The high-resolution SAR satellite image line segment extraction method according to claim 1, characterized in that, The step of determining the current second growth point from the candidate growth point set based on the endpoints of the current step-size growth segment, using spatial distance constraints, directional consistency constraints, and fitting direction vector change constraints, includes: Based on the candidate growth point set, within the second defined neighborhood of the current step-size growth segment, a candidate growth point that simultaneously satisfies the constraints of spatial distance, directional consistency, and fitting direction vector change is identified as the current second growth point; wherein, the constraints of spatial distance, directional consistency, and fitting direction vector change include: The vertical distance from the candidate growth point to the currently fitted line segment is less than the vertical distance threshold; The angle difference between the principal direction of the candidate growth point and the direction of the currently fitted line segment is less than the second angle threshold; and After adding the candidate growth points to the point set, the updated point set is obtained by performing line fitting to obtain a new fitted growth line segment; the angle difference between the direction of the new fitted growth line segment and the direction of the previously fitted growth line segment is less than the third angle threshold.

7. The high-resolution SAR satellite image line segment extraction method according to claim 6, characterized in that, After the step of finding candidate growth points that simultaneously satisfy the constraints of spatial distance, directional consistency, and fitting direction vector change, the method further includes: If no candidate growth point is found in the second set neighborhood that simultaneously satisfies the constraints of spatial distance, directional consistency, and fitting direction vector change, then jump to the step size growth step. When the number of jumps reaches the first threshold, the growth is terminated. If a candidate growth point that simultaneously satisfies the constraints of spatial distance, directional consistency, and fitting direction vector change is found within the second defined neighborhood, the current seed point, the first growth point, and all found second growth points are constructed as the current point set. Based on the current point set, a straight line is fitted to generate the current fitted line segment.

8. The high-resolution SAR satellite image line segment extraction method according to claim 1, characterized in that, The step of growing the line segment by step size along both the positive and negative directions of the initial fitted line segment or the current fitted line segment to obtain the current step size growth line segment includes: When the fitted line segment is the initial fitted line segment, along both the positive and negative directions of the initial fitted line segment, according to the step size... To grow, , This is the step size coefficient. The length of the initially fitted line segment; When the fitted line segment is the current fitted line segment, proceed along both the positive and negative directions of the current fitted line segment, according to the step size. To grow, , This is the step size coefficient. The length of the initially fitted line segment.

9. The high-resolution SAR satellite image line segment extraction method according to claim 1, characterized in that, After repeating the step-size growth step and the current second growth point determination step until no new second growth point can be found, the method further includes: performing direction consistency verification on the current fitted line segment obtained based on each seed point, including: Based on the current seed point and the corresponding currently fitted line segment, construct the current point set by the current seed point, the first growth point, and all found second growth points; perform line fitting on the current point set to obtain the current fitting direction; Calculate the angle difference between the main direction and the current fitted direction for each point in the current point set, count the number of points whose angle difference is less than the fourth angle threshold, count the total number of points in the current point set, and divide the number by the total number to obtain the first ratio; If the first ratio is greater than or equal to the ratio threshold, the currently fitted line segment is determined to be a valid line segment and retained; if the first ratio is less than the ratio threshold, the currently fitted line segment is determined to be an invalid line segment and discarded.

10. A method for extracting line segments from high-resolution SAR satellite images, characterized in that, include: Acquire high-resolution SAR satellite imagery and process it into amplitude images; obtain a seed point set based on constant false alarm rate detection, non-maximum suppression, and diagonal response value thresholding; A set of candidate growth points is obtained based on nonmaximum suppression and adaptive amplitude thresholding of the amplitude image; With each point in the seed point set and the candidate growth point set as the center, a circular neighborhood window is constructed. Based on the pixel amplitude value of the pixels in the circular neighborhood window of each point, the principal direction of each seed point in the seed point set and the principal direction of each candidate growth point in the candidate growth point set are obtained through weighted principal component analysis. For each seed point, based on the main direction of the seed point and the main direction of the candidate growth points, a first growth point is determined from the set of candidate growth points using spatial distance and direction consistency constraints. Based on the first growth point, a second growth point is determined from the set of candidate growth points using spatial distance, direction consistency, and fitting direction vector change constraints, generating the current fitted line segment. The new second growth point determination step is as follows: based on the current second growth point and the current fitted line segment, a new second growth point is determined from the set of candidate growth points using spatial distance, direction consistency, and fitting direction vector change constraints, generating a new fitted line segment. The new second growth point determination step is repeated until no new second growth point can be found.

11. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the high-resolution SAR satellite image line segment extraction method provided in any one of claims 1 to 10.

Images