SAR image ship target sidelobe removing method based on distance constraint
By using a distance-constrained method, the ship target is rotated and the distances from its two sides to the main axis are calculated. Combined with morphological preprocessing, the side lobes of the ship target in the SAR image are gradually removed, which solves the problems of estimation error and poor imaging effect in the existing technology and improves the accuracy of ship target identification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-10-30
- Publication Date
- 2026-03-03
AI Technical Summary
Existing technologies suffer from estimation errors and poor imaging results when removing side lobes and trailing phenomena of ship targets in SAR images. This leads to inaccurate estimation of the geometric parameters of ship targets, affecting subsequent identification results.
By using a distance constraint-based method, the ship target is rotated to the horizontal direction, the distances from both sides of the ship to the main axis are calculated, and the side lobes are removed using distance constraints. Combined with morphological preprocessing and connected component analysis, the side lobe pixels of the ship target are gradually removed to obtain the true width.
It effectively removes the side lobes of ship targets in SAR images, improving the accuracy and robustness of ship target identification, and is suitable for images with different imaging effects.
Smart Images
Figure CN115797644B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of image processing, specifically relating to a method for removing sidelobes of SAR images of ship targets based on distance constraints. This method is used to remove sidelobes caused by strong scattering points and trailing caused by geometric distortion during the imaging process from ship targets in SAR images. Background Technology
[0002] Synthetic Aperture Radar (SAR) is an active microwave imaging sensor with the advantage of being unaffected by lighting and weather conditions, enabling it to acquire data around the clock and in all weather conditions. It has wide applications in military and agriculture. Ship detection and identification plays a crucial role in maritime traffic control and marine resource protection. In civilian applications, ship target identification assists relevant departments in planning maritime and port traffic and searching for distressed vessels; in military applications, it provides information on enemy ships, helping the navy to deploy defenses in advance.
[0003] When a ship sails on the sea, its velocity can be decomposed along the range and azimuth dimensions. The azimuth component introduces geometric distortion into the ship's imaging, manifesting in actual SAR images as a strong scattering region near the azimuth direction, known as a trailing phenomenon. Simultaneously, due to the characteristics of the ship's strong scattering structure, when a particular structure within the ship exhibits strong backscattering, it suppresses surrounding weakly scattering targets, producing cross-shaped white bright spots in the SAR image—a phenomenon known as the sidelobe effect. When the sidelobe effect exists, the detected target area extends beyond the main ship area, causing the extracted ship outline (minimum bounding rectangle) to be wider / longer than its true value, severely impacting target recognition performance.
[0004] The following two methods are for removing ship sidelobes from SAR images based on grayscale images:
[0005] 1. Stepwise approximation method:
[0006] The successive approximation method first estimates the azimuth angle of the target ship in the SAR image, rotates the target to a horizontal position, and obtains an initial bounding rectangle with length L and width W. The successive approximation method then reduces the initial bounding rectangle, finally obtaining the final bounding rectangle. The specific method is as follows:
[0007] (1) Count the number of target pixels, denoted as N;
[0008] (2) Calculate the average number of pixels S in each row. W =N / W;
[0009] (3) Calculate the number of ship pixels on the upper and lower boundaries respectively. If it is less than aS WThen shrink the circumscribed rectangle boundary inwards until it is not less than aS. W until.
[0010] (4) The same treatment applies to the horizontal direction.
[0011] The successive approximation method is simple to implement and computationally inexpensive. Its implementation assumes that the initial bounding rectangle of the target is not significantly different in size from the overall size of the ship, allowing for the iterative removal of side lobes. However, practical experiments show that the successive approximation method does not account for the impact of large side lobes on the initial bounding rectangle. When the side lobes or trailing tails are too long, the estimated width or length of the initial bounding rectangle is too large, resulting in a threshold calculated using the formula that is too small, causing the loop to terminate before reaching the target area of the ship. Although this can be improved by adjusting the coefficient 'a', adaptive adjustment of 'a' is not possible when the side lobe size is unknown.
[0012] 2. Radon Transform Method
[0013] The Radon transform can transform an image to the ρ-θ space. The essence of the transformation is the line integral along a straight line over the image pixels. Each pixel on each straight line in the image domain corresponds to a point in the ρ-θ space after the line integral. Therefore, in a SAR image, lines with high pixel gray values correspond to bright spots in the ρ-θ space, while lines with low pixel gray values correspond to dark spots. Ships in SAR images are considered bright areas, corresponding to a set of bright areas in the ρ-θ domain.
[0014] The Radon transform method first performs a Radon transform on the binary image to determine the principal axis direction, i.e., the azimuth angle. Using the estimated azimuth angle, the binary image is rotated to a horizontal orientation. Then, a Radon transform is performed again on the rotated binary image. All line integral values along the 90° projection direction are used as the target width estimate. To remove sidelobes, k times the maximum line integral value is used as a threshold to remove points below the threshold along the 90° direction. The remaining values are used to estimate the target width.
[0015] The Radon transform method transforms the sidelobe removal problem into a maximum peak detection problem in the ρ-θ domain. However, in some cases, the angle corresponding to the maximum peak point in ρ-θ is not in the direction of the principal axis, which leads to incorrect target azimuth estimation and thus prevents sidelobe removal. Furthermore, the target width estimation uses k times the maximum line integral value as a threshold. When the ship imaging effect is poor, the brightness of the ship target in the ρ-θ domain is low, which may cause the main body of the ship to be misidentified as a sidelobe and removed. Summary of the Invention
[0016] Technical problems to be solved
[0017] The technical problem this invention aims to solve is that the presence of sidelobes leads to incorrect estimation of the geometric parameters of ship targets in SAR images, thus affecting subsequent identification. The presence of large sidelobes and trailing edges causes the initial minimum bounding rectangle size estimated by the successive approximation method to be too large, prematurely terminating the subsequent sidelobe removal process and failing to completely remove large sidelobes. Furthermore, the Radon transform method is prone to errors in estimating azimuth angles and performs poorly on ship images with low imaging quality. Based on these problems, this invention proposes a range-constrained SAR image ship target sidelobe removal method.
[0018] Technical solution
[0019] A method for removing sidelobes of ship targets in SAR images based on range constraints, characterized by the following steps:
[0020] Step 1: Rotate the binary image so that the long line containing the centroid of the target in the horizontal direction is taken as the target's principal axis. Using the principal axis as the dividing line, target pixels above the principal axis are considered to be the upper half of the ship, and target pixels below the principal axis are considered to be the lower half of the ship.
[0021] Step 2: First, perform sidelobe removal on the upper half of the target. Count the number of target pixels detected on the upper side of the i-th column of the main axis of the ship target, and define it as the distance G_on from the upper half of the ship to the main axis, where G_on = [G_on1, G_on2, ... G_on...]. w ], where w is the length of the target ship; the maximum value in G_on is denoted as max_on;
[0022] Step 3: Sort the obtained G_on, remove the first 25% of the close-range cells and the last 25% of the far-range cells, and use the average distance mean_on calculated from the remaining middle 50% of cells as an estimate of the ship's true top width.
[0023] Step 4: Compare the size of p*max_on with mean_on. If p*max_on > mean_on, the pixels outside p*max_on are considered sidelobe pixels and are set to 0; where p is a parameter.
[0024] Step 5: Update the binary image, repeat steps 1-4 until p*max_on ≤ mean_on; the loop ends.
[0025] Step 6: Process the pixels below the main axis using the same method.
[0026] A further technical solution of the present invention includes preprocessing the original slices to obtain a binary image: First, CFAR detection is performed on the original SAR image slices to obtain a binary image. In the binary image, the target region has a value of 1, and the rest of the region has a value of 0. The obtained binary image contains holes in the ship target part, which are removed using opening and closing operations in morphology. The binary image after morphological filtering contains small clutter regions. By calculating the area of the connected region and limiting the minimum size of the connected region, the smaller regions are removed.
[0027] A further technical solution of the present invention: in step 4, p is set to 0.9.
[0028] A computer system is characterized by comprising: one or more processors, and a computer-readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method described above.
[0029] A computer-readable storage medium is characterized by storing computer-executable instructions, which, when executed, are used to implement the above-described method.
[0030] Beneficial effects
[0031] This invention provides a distance-constrained SAR image ship target sidelobe removal method. Compared with existing sidelobe removal algorithms, it offers the following advantages: The successive approximation method relies on the parameters of the initial bounding rectangle. When the length or width of the initial bounding rectangle is excessive due to the influence of sidelobes, the successive approximation method cannot remove all sidelobes. Furthermore, the Radon transform method is prone to removing ship pixels as sidelobes when the imaging quality is poor. The method proposed in this invention, based on the distribution of sidelobes within the ship in the SAR image, starts with the farthest-distance pixels. It only needs to update the estimated ship width and the farthest distance from both sides of the ship to the principal axis during the iteration process to progressively remove sidelobe pixels and obtain the true target width. It does not rely on prior knowledge and is robust to SAR images with poor imaging quality. Attached Figure Description
[0032] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.
[0033] Figure 1 Schematic diagram of the successive approximation method;
[0034] Figure 2 Schematic diagram of Radon transform method;
[0035] Figure 3Schematic diagrams showing the failure of the successive approximation method and Radon transform; left (successive approximation method fails) right (Radon transform method fails);
[0036] Figure 4 A schematic diagram of the SAR image ship target sidelobe removal method based on distance constraints of the present invention. Detailed Implementation
[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.
[0038] The solution to the technical problem of this invention is as follows: First, the azimuth angle of the ship target is estimated based on the moment technique. The ship target is then rotated to a horizontal position using the estimated azimuth angle. The centroid of the ship target after rotation is then calculated. Since the ship target has good symmetry, the position of the centroid can be considered as the position of the ship's main axis. Then, the distances from each column on both sides of the ship target to the main axis are statistically analyzed in the binary image rotated to the horizontal position. The statistically obtained distances are sorted and truncated. The width of the ship is roughly estimated using the middle part of the distance. Finally, the sidelobe pixels are iteratively removed based on the relationship between the farthest distances from both sides to the main axis and the estimated width of the ship, and the true width estimate of the ship target is finally obtained.
[0039] To more clearly illustrate the specific implementation, this invention preprocesses the original slices to obtain a binary image. First, CFAR detection is performed on the original SAR image slices to obtain a binary image. In the binary image, the target region has a value of 1, and the remaining regions have a value of 0. Holes exist in the ship target portion of the obtained binary image, which are removed using opening and closing operations in morphology. Small clutter regions exist in the morphologically filtered binary image; these smaller regions are removed by calculating the area of connected regions and limiting the minimum connected region size. Finally, a range-constrained SAR image ship target sidelobe removal method is applied to the obtained binary image.
[0040] Figure 4 This is a schematic diagram of a range-constrained SAR image sidelobe removal method for ship targets. The sidelobe removal algorithm consists of the following steps:
[0041] (1) High-resolution ship targets are symmetrical, with their centroid roughly located near the principal axis. Therefore, the long straight line containing the centroid of the target rotated to the horizontal direction is considered the target's principal axis. Using the principal axis as the dividing line, target pixels above the principal axis are considered the upper half of the ship, and target pixels below the principal axis are considered the lower half of the ship. Since the estimated ship principal axis deviates somewhat from the actual ship principal axis, the upper and lower parts are processed separately when removing side lobes.
[0042] (2) First, perform sidelobe removal in the upper half. Count the number of target region pixels detected on the upper side of the i-th column of the main axis of the ship target, and define it as the distance G_on from the upper half of the ship to the main axis, where G_on = [G_on1, G_on2, ... G_on...]. w [], where w is the length of the target ship. The maximum value in G_on is denoted as max_on.
[0043] (3) The side lobes of the ship target are far from the main axis, while the distances of the ship pixels on both sides without side lobes from the main axis are basically the same. Additionally, in some cases, due to imaging issues, target occlusion, etc., the imaging effect in some areas of the ship is poor, resulting in the calculated upper edge of the ship being closer to the main axis. To remove the influence of these two situations on the estimated ship width, the obtained G_on is sorted, the first 25% of near-range cells and the last 25% of far-range cells are removed, and the average distance mean_on calculated from the remaining middle 50% of cells is used as the estimate of the true upper width of the ship.
[0044] (4) Remove pixels far from the main axis. Compare the size of p*max_on with mean_on. If p*max_on > mean_on, the pixels outside p*max_on are considered sidelobe pixels and are set to 0. Here, p is a parameter, which is set to 0.9 in the experiment.
[0045] (5) Update the binary graph and repeat steps 1-4 until p*max_on ≤ mean_on. The loop ends.
[0046] (6) Use the same method to process the pixels below the main axis.
[0047] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the scope of the technology disclosed in the present invention, and such modifications or substitutions should all be covered within the scope of protection of the present invention.
Claims
1. A method for sidelobe removal of ship targets in SAR images based on distance constraint, characterized in that The steps are as follows: Step 1: Rotate the binary image to the long straight line where the centroid of the target in the horizontal direction is the target principal axis, and use the principal axis as the demarcation line, and consider the target pixel points above the principal axis as the upper half of the ship and the target pixel points below the principal axis as the lower half of the ship; Step 2: First, the upper half of the sidelobe is removed, and the target of the warship is counted The number of pixels of the target area detected on the upper side of the main shaft is defined as the distance from the upper half of the warship to the main shaft , , The length of the target warship is defined as L The maximum value in the column is recorded as L ; Step 3: Sort the obtained Sort the obtained distances, remove the closest 25% and the farthest 25%, and calculate the average distance from the remaining 50% of the middle As an estimate of the true beam of the ship Step 4: Compare with in size, if , consider pixels other than the mainlobe pixel to be sidelobe pixels, set them to 0; where is a parameter; Step 5: Update the binary map, repeat steps 1-4 until ; loop ends; Step 6: Use the same method to process the pixel points below the principal axis.
2. The method of claim 1, wherein the method is a distance constraint based SAR image ship target sidelobe removal method. Also includes pre-processing the original slice to obtain a binary image: first, perform CFAR detection on the original SAR image slice to obtain a binary image, in which the target region is 1 and the remaining regions are 0; the obtained binary image has a hollow in the ship target part, and open operation and close operation in morphology are used for removal; The binary image after morphological filtering has small clutter regions, and the area of the connected region is calculated to limit the minimum connected region size to remove the regions with smaller areas.
3. The method of claim 1, wherein the method is a distance constraint based SAR image ship target sidelobe removal method. In step 4 was set to 0.
9.
4. A computer system, characterized by Comprise: One or more processors, a computer readable storage medium for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method of claim 1.
5. A computer readable storage medium characterized by Store computer executable instructions, which when executed, are used to implement the method of claim 1.