A ship detection method for SAR images based on arbitrary directional morphological analysis

Abstract

This invention discloses a method for ship detection in SAR images based on arbitrary directional morphological analysis, comprising the following steps: Step 10, constructing a convolution kernel and convolving the original SAR image; Step 20, uniformly sampling the preprocessed image to obtain a binary image; Step 30, labeling the binary image; Step 40, determining multiple preliminary detection regions; Step 50, determining the background window of each preliminary detection region; Step 60, determining the target subject region; Step 70, calculating the center of the target subject region; Step 80, determining the width and length of the target subject region; Step 90, determining the detection result based on the width and length of the target subject region. This invention reduces the complexity of image preprocessing while significantly improving image interpretation speed and ensuring the accuracy of target detection.

Description

Technical Field

[0001] The invention belongs to the technical field of image processing, and in particular relates to a SAR image ship detection method for arbitrary direction morphological analysis. Background Art

[0002] The terrain at the junction of land and sea is often complex, and ship targets need to be accurately detected in such a complex environment. In the complex environment of the sea-land junction, the echo scattering is complex, and non-ship targets cannot be eliminated simply by relying on the echo amplitude, which makes the detection of ship targets more difficult. Most traditional ships use CFAR (Constant False Alarm Rate Detector) detection. During image preprocessing, there is often some residue when the land environment at the sea-land junction is eliminated. During detection, the sliding window traverses the image, and the sliding window size is fixed. The image preprocessing process is relatively complicated, which makes the calculation efficiency low when there are many images, and the final detection performance is poor. Summary of the invention

[0003] In order to solve the above problems existing in the prior art, the present invention provides a SAR image ship detection method based on arbitrary direction morphological analysis. The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0004] A SAR image ship detection method based on arbitrary direction morphological analysis comprises the following steps:

[0005] Step 10, constructing a convolution kernel, and convolving the convolution kernel with the original SAR image of the ship target to be detected to obtain a preprocessed image; wherein the size of the convolution kernel is an average of a first preset size and a second preset size; the first preset size is the size of the smallest ship target on the SAR image, and the second preset size is the size of the largest ship target on the SAR image;

[0006] Step 20, uniformly sampling the preprocessed image to obtain amplitude data, and binarizing the preprocessed image according to the amplitude data and a K distribution model to obtain a binary image;

[0007] Step 30, marking the areas in the binary image where the pixel value is 1, to obtain a marked image having a plurality of marked areas;

[0008] Step 40, determining a plurality of first regions in a plurality of marked regions according to the second preset size and the size of the convolution kernel, and then determining a plurality of preliminary detection regions according to the average amplitude of each of the first regions and the average amplitude and amplitude standard deviation of the plurality of first regions;

[0009] Step 50, determining a background window of each preliminary detection area in each preliminary detection area according to the size of the convolution kernel and the pixel with the largest amplitude in each preliminary detection area;

[0010] Step 60, determining the target main body area according to the average amplitude of the background window and the amplitude of each pixel point in the preliminary detection area;

[0011] Step 70, calculating the center of the target body area according to the zero-order moment and the first-order moment of the target body area;

[0012] Step 80, sampling from the center of the target body area to equally spaced positions on both sides, and determining the sampling center of each sampling position, and determining the width and length of the target body area according to the sampling center and the center of the target body area;

[0013] Step 90: Determine a detection result according to the first preset size, the second preset size, and the width and length of the target body area.

[0014] Beneficial effects of the present invention:

[0015] 1. The present invention performs a convolution preprocessing operation before detecting the SAR image, which can not only reduce the number of holes in the SAR image, but also increase the amplitude of the ship target, making the ship target easier to detect.

[0016] 2. The present invention combines the brightness distribution and spatial distribution of SAR images, which can not only avoid the problem of inaccurate detection results caused by the inability to distinguish between the detection target and the background in the bright domain, but also accurately detect the spatial position of the ship and improve the accuracy of the detection results.

[0017] 3. The present invention adopts the method of taking the average value of multiple tests, which not only avoids the randomness and uncertainty of the ship detection results, but also improves the accuracy of the detection results, ensures the efficiency of the detection results, and can eliminate detection errors to a certain extent.

[0018] The present invention simplifies the image by constructing a convolution kernel and convolving it with the image, eliminating holes and separation parts to reduce the frequency of image amplitude changes, making it easier to distinguish ship targets and reducing the complexity of image preprocessing. At the same time, the target is subsequently screened based on energy and shape for preliminary detection, and the target body part of the part with the target is further extracted based on the peak sidelobe ratio, and the posture and size of the target body are extracted, which significantly improves the image interpretation speed and ensures the accuracy of target detection.

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The present invention provides a flowchart of a method for detecting ships in SAR images by morphological analysis in any direction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.

[0022] like Figure 1 As shown, a SAR image ship detection method based on arbitrary direction morphological analysis includes the following steps:

[0023] Step 10, construct a convolution kernel, and convolve the convolution kernel with the original SAR image of the ship target to be detected to obtain a preprocessed image; wherein, the size of the convolution kernel is the average of the first preset size and the second preset size; the first preset size is the size of the smallest ship target on the SAR image, and the second preset size is the size of the largest ship target on the SAR image; the first preset size and the second preset size are known prior information, and the sizes are length and width.

[0024] Step 20, uniformly sampling the preprocessed image to obtain amplitude data, and binarizing the preprocessed image according to the amplitude data and the K distribution model to obtain a binary image.

[0025] Specifically, the specific steps of step 20 include steps 21 to 24:

[0026] Step 21, uniformly sample the preprocessed image to obtain a sampled image.

[0027] Step 22, taking the first N amplitudes with the largest number of occurrences in the sampled image as the statistical data of the K distribution model; in this step, all the amplitudes are sorted from most to least according to the number of occurrences, and the first N amplitudes are the amplitudes with the largest number of occurrences, and N is an integer greater than or equal to 2, for example, taking the first 10 amplitudes with the largest number of occurrences.

[0028] Step 23, calculating a threshold value based on the K distribution model, statistical data and a preset false alarm probability.

[0029] In this step, according to the characteristics of sea clutter conforming to K distribution, the amplitude is used as the statistical data of K distribution model, and a suitable false alarm probability is set to calculate the threshold value. Specifically, for SAR ship target background sea clutter, the pixel distribution is relatively scattered, and K distribution has good statistical characteristics for this. In view of the distribution of ship target and its background, it is proposed to use K distribution to describe the sea background. The probability density function of K distribution is for:

[0030] .

[0031] in, is the probability density function of K distribution, , is the mean, is the shape parameter, is the image view number (which can be obtained by imaging processing); is the Gamma function, for To obtain accurate Distribution needs to The traditional parameter estimation method is the maximum likelihood method. In order to avoid complex estimation, tedious iteration and save computational efficiency, the statistical value of the pixel points in the background window is used to replace the estimated value for approximate fitting. Distribution model, The statistical value of can be expressed as:

[0032] ;

[0033] ;

[0034] in For the expected operation, This is the statistical data described in step 22.

[0035] Using the above model as the sea surface background distribution model in detection, the false alarm probability formula can be obtained:

[0036] .

[0037] in, is the false alarm probability. By setting a reasonable false alarm probability and solving the above formula, we can get the threshold value (threshold value): After that, threshold detection is performed on the target to be detected and the background, and a preliminary judgment can be made on the target to be detected. In this embodiment, after the threshold value is calculated, the amplitude of each pixel in the preprocessed image is compared with the threshold value.

[0038] Step 24, setting the values ​​of the pixels in the preprocessed image whose amplitude is less than the threshold value to 0, and setting the values ​​of the pixels whose amplitude is greater than or equal to the threshold value to 1, to obtain a binary image.

[0039] Step 30, marking the areas where the pixel values ​​in the binary image are 1, and obtaining a marked image with multiple marked areas; specifically, the pixels with a value of 1 in the binary image constitute a region, and the binary image has multiple regions, and each region needs to be marked.

[0040] The specific steps of step 30 include steps 31 to 36:

[0041] Step 31, constructing an initial connected region labeling matrix based on any pixel point with a value of 1 in the binary image as a starting point; the initial connected region labeling matrix is ​​recorded as the i-th result labeling matrix; wherein i is an integer greater than or equal to 1; in this step, the initial connected region labeling matrix is ​​the same as the matrix corresponding to the binary image. The starting point and the position in the initial connected region labeling matrix correspond to any pixel point with a value of 1 selected in the binary image.

[0042] Step 32, dilate the i-th result label matrix with the starting point as the center to obtain the i-th dilated label matrix; the dilation process is to add pixel values ​​to the adjacent pixels around the edge of a certain area. In this step, for example, dilate the eight adjacent pixels around the starting point as the center, and add pixel value 1 to each of the eight pixels to obtain the first dilated label matrix.

[0043] Step 33, multiply the i-th expanded labeling matrix by the binary image to obtain the i+1-th result labeling matrix.

[0044] Step 34, if the (i+1)th result marking matrix is ​​the same as the (i)th result marking matrix, then the marking of the area where the starting point is located is completed, and a marked area is obtained;

[0045] If the i+1th result marking matrix is ​​different from the i-th result marking matrix, the i+1th result marking matrix is ​​expanded to obtain the i+1th expanded marking matrix, and the i+1th expanded marking matrix is ​​multiplied by the binary image to obtain the i+2th result marking matrix. This cycle is repeated until the i+nth result marking matrix is ​​the same as the i+n-1th result marking matrix, and the marking area is obtained, where n is an integer greater than or equal to 1.

[0046] Step 35, setting the values ​​of the pixels in the area in the binary image corresponding to the marked area to 0.

[0047] The above steps 32 to 35 are the process of marking an area, and the marking process for each area is the same.

[0048] Step 36, according to a pixel point with a value of 1 in the binary image except the area corresponding to the marked area, the corresponding pixel point is used as the starting point in the result marking matrix with the previous marked area, and steps 32 to 35 are executed in the result marking matrix of the previous marked area until the values ​​of the pixel points in all areas of the binary image are 0, and the last result marking matrix is ​​used as the marked image; the marked image has multiple marked areas.

[0049] Step 40, determine multiple first areas in multiple marked areas according to the second preset size and the size of the convolution kernel, and then determine multiple preliminary detection areas according to the average amplitude of each first area and the average amplitude of multiple first areas and the amplitude standard deviation; in this step, perform preliminary detection to obtain a preliminary detection image with multiple preliminary detection areas.

[0050] Specifically, the specific steps of step 40 include steps 41 to 43:

[0051] Step 41, setting the amplitude of the pixel points of the marked areas whose size is greater than the sum of the second preset size and the size of the convolution kernel among the multiple marked areas to 0, and each of the remaining marked areas is regarded as a first area; in this step, the number of pixels in each marked area is counted, and if the sum of the number of pixels in the marked area is greater than the sum of the number of pixels in the second preset size and the number of pixels in the convolution kernel, the marked area that does not meet the conditions is removed, and after all the marked areas that do not meet the conditions are removed, each of the remaining marked areas is regarded as a first area.

[0052] Step 42, calculating the average amplitude of each first region, the average amplitude of all first regions, and the amplitude standard deviation of all first regions.

[0053] Step 43, according to 3 In principle, among multiple first regions, the amplitude of the pixel points in the first region whose difference between the average amplitude of each first region and the average amplitude of all first regions is equal to three times the amplitude standard deviation is set to 0, and each of the remaining first regions is used as a preliminary detection region. In this step, among multiple first regions, the regions that do not meet the conditions are further eliminated, and the remaining first regions are used as a preliminary detection region.

[0054] Step 50, determining the background window of each preliminary detection area in each preliminary detection area according to the size of the convolution kernel and the pixel with the largest amplitude in each preliminary detection area; this step can determine the background window of each preliminary detection area in the preliminary detection image.

[0055] The specific steps of this step are: in each preliminary detection area, expand the pixel with the largest amplitude to the area with the same size as the convolution kernel to obtain the target window, and use the area except the target window in the preliminary detection area as the background window.

[0056] Step 60, determine the target body area according to the average amplitude of the background window and the amplitude of each pixel point in the preliminary detection area where the background window is located; in this step, the preliminary detection image is further processed to determine the target body area in each preliminary detection area, thereby obtaining a secondary detection image.

[0057] Specifically, the specific steps of step 60 include step 61-step 62:

[0058] Step 61, judging whether each pixel point belongs to the pixel point of the ship target to be detected according to the average amplitude of the background window and the amplitude of each pixel point in the preliminary detection area corresponding to the background window.

[0059] In this step, specifically, the image amplitude statistical characteristics are used to determine whether the pixel belongs to the ship target to be detected, and the determination method is as follows:

[0060] .

[0061] in, is the amplitude of a single pixel in the preliminary detection area, is the mean amplitude of the background window, is the variance of the amplitude of the background window, It is a normalization factor, which is generally set according to actual needs.

[0062] Step 62, if the pixel does not belong to the ship target to be detected, the amplitude of the pixel is set to 0, and the pixel belonging to the ship target to be detected is the target main area. In this step, each preliminary detection area is further processed, and the pixel points that do not belong to the ship target to be detected in each preliminary detection area are further eliminated. After elimination, the remaining pixel points constitute the target main area. At this time, a secondary detection image with multiple target main areas can be obtained, and only the pixel points in the target main area in the image have amplitudes.

[0063] Step 70: Calculate the center of the target body area according to the zero-order moment and the first-order moment of the target body area.

[0064] Specifically, the specific steps of step 70 include step 71-step 72:

[0065] Step 71, calculate the zero-order moment and the first-order moment of the target main area; specifically, calculate the zero-order moment and the first-order moment according to the following formula:

[0066] .

[0067] in, is the zero-order moment of the target main area, , are the first-order moments of the rows and columns of the target body area, respectively. is the number of rows in the target body area, is the number of columns in the target body area, is the image matrix of the target subject area.

[0068] Step 72: Calculate the center of the target body area based on the zero-order moment and the first-order moment.

[0069] Specifically, the center of the target main area is calculated according to the following formula:

[0070] ;

[0071] Indicates the position coordinates of the center of the target body area.

[0072] Step 80, for each target body region, sampling is performed from the center of the target body region to equally spaced positions on both sides, and the sampling center of each sampling position is determined, and the width and length of the target body region are determined according to the sampling center and the center of the target body region.

[0073] Specifically, the specific steps of step 80 include step 81 to step 86:

[0074] Step 81, sampling is performed from the center of the target main body area to equally spaced positions on both sides; in this step, sampling is performed from the center of the target main body area to equally spaced rows or columns on both sides.

[0075] Step 82, taking the pixel point with the largest amplitude at each sampling position as the sampling center; the sampling position is also the row or column for sampling. In this step, a sampling center can be obtained at each sampling position, and finally multiple sampling centers are obtained.

[0076] Step 83, calculate the slope of the line between each sampling center and the center of the target body area; in this step, the slope of the line between the sampling center and the center of the target body area is calculated based on the coordinates of each sampling center and the coordinates of the center of the target body area.

[0077] Step 84, calculating the average value of the multiple slopes and the inclination angle corresponding to the average value of the slope.

[0078] Step 85, if the inclination angle is greater than 45 degrees, calculate the total number of pixel points with amplitude in the rows where all sampling centers are located, as well as the maximum number of rows and the minimum number of rows with amplitude; in this step, for example, if there are 10 sampling centers, the total number of pixel points with amplitude in the rows where the sampling centers are located, as well as the maximum number of rows (which row) and the minimum number of rows (which row) among all rows with amplitude are counted.

[0079] If the inclination angle is less than or equal to 45 degrees, the total number of pixel points with amplitude in the column where all sampling centers are located, as well as the maximum number of columns and the minimum number of columns with amplitude are calculated; in this step, for example, there are 10 sampling centers, and the total number of pixel points with amplitude in the column where the sampling centers are located, as well as the maximum number of columns and the minimum number of columns in all columns with amplitude are counted.

[0080] Step 86 corresponds to step 85. There are two cases in this step:

[0081] If the tilt angle is greater than 45 degrees, the width of the ship target to be detected is determined according to the average value of the number of pixels and the tilt angle, and the length of the ship target to be detected is determined according to the maximum number of rows, the minimum number of rows and the tilt angle. In this step, the total number of pixels in the above step 85 is divided by the total number of rows or columns where the sampling center is located to obtain the average value of the number of pixels. The average value is multiplied by the sine value of the tilt angle to obtain the width of the target main area, and the length of the target main area is obtained by multiplying the difference between the maximum number of rows and the minimum number of rows by the cosine value of the tilt angle;

[0082] If the tilt angle is less than or equal to 45 degrees, the width of the ship target to be detected is determined according to the average value of the number of pixels and the tilt angle, and the length of the ship target to be detected is determined according to the maximum number of columns, the minimum number of columns and the tilt angle. In this step, the total number of pixels in the above step 85 is divided by the total number of rows or columns where the sampling center is located to obtain the average value of the number of pixels. The average value is multiplied by the sine value of the tilt angle to obtain the width of the target main area, and the length of the target main area is obtained by multiplying the difference between the maximum number of rows and the minimum number of rows by the cosine value of the tilt angle.

[0083] Step 90, determining the detection result according to the first preset size, the second preset size, and the width and length of the target body area. In this step, it is determined whether the width and length of the target body area are greater than or equal to the width and length of the first preset size, and whether the width and length of the target body area are less than or equal to the width and length of the second preset size. If so, the target body area is the detection result, if not, it is considered to be a false target.

[0084] The present invention performs a convolution preprocessing operation before detecting the SAR image, which can not only reduce the number of holes in the SAR image, but also increase the amplitude of the ship target, making the ship target easier to detect.

[0085] By combining the brightness distribution and spatial distribution of the SAR image, the present invention can not only avoid the problem of inaccurate detection results caused by the inability to distinguish between the detection target and the background in the bright domain, but also accurately detect the spatial position of the ship and improve the accuracy of the detection results.

[0086] The present invention adopts a method of taking an average value after multiple tests, which not only avoids the randomness and uncertainty of the ship detection results, but also improves the accuracy of the detection results, ensures the efficiency of the detection results, and can eliminate detection errors to a certain extent.

[0087] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0088] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification.

[0089] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A SAR image ship detection method based on arbitrary direction morphological analysis, characterized in that: The following steps are involved: Step 10, constructing a convolution kernel, and convolving the convolution kernel with the original SAR image of the ship target to be detected to obtain a preprocessed image; wherein the size of the convolution kernel is an average of a first preset size and a second preset size; the first preset size is the size of the smallest ship target on the SAR image, and the second preset size is the size of the largest ship target on the SAR image; Step 20, uniformly sampling the preprocessed image to obtain amplitude data, and binarizing the preprocessed image according to the amplitude data and a K distribution model to obtain a binary image; Step 30, marking the areas in the binary image where the pixel value is 1, to obtain a marked image having a plurality of marked areas; Step 40, determining a plurality of first regions in a plurality of marked regions according to the second preset size and the size of the convolution kernel, and then determining a plurality of preliminary detection regions according to the average amplitude of each of the first regions and the average amplitude and amplitude standard deviation of the plurality of first regions; Step 50, determining a background window of each preliminary detection area in each preliminary detection area according to the size of the convolution kernel and the pixel with the largest amplitude in each preliminary detection area; Step 60, determining the target main body area according to the average amplitude of the background window and the amplitude of each pixel point in the preliminary detection area; Step 70, calculating the center of the target body area according to the zero-order moment and the first-order moment of the target body area; Step 80, sampling from the center of the target body area to equally spaced positions on both sides, and determining the sampling center of each sampling position, and determining the width and length of the target body area according to the sampling center and the center of the target body area; Step 90: Determine a detection result according to the first preset size, the second preset size, and the width and length of the target body area.

2. The method for ship detection in SAR images by arbitrary direction morphological analysis according to claim 1, characterized in that: The step 20 comprises: Step 21, uniformly sampling the preprocessed image to obtain a sampled image; Step 22, taking the first N amplitudes that appear most frequently in the sampled image as statistical data of the K distribution model; wherein N is an integer greater than or equal to 2; Step 23, calculating a threshold value based on the statistical data, the K distribution model and a preset false alarm probability; Step 24, setting the values ​​of the pixels in the preprocessed image whose amplitude is less than the threshold value to 0, and setting the values ​​of the pixels whose amplitude is greater than or equal to the threshold value to 1, to obtain a binary image.

3. The method for ship detection in SAR images by arbitrary direction morphological analysis according to claim 1, characterized in that: The step 30 comprises: Step 31, constructing an initial connected region labeling matrix based on any pixel point with a value of 1 in the binary image as a starting point; the initial connected region labeling matrix is ​​recorded as the i-th result labeling matrix; Step 32, performing expansion processing on the i-th result labeling matrix with the starting point as the center to obtain the i-th expanded labeling matrix; wherein i is an integer greater than or equal to 1; Step 33, multiplying the i-th expansion label matrix by the binary image to obtain the i+1-th result label matrix; Step 34, if the (i+1)th result marking matrix is ​​the same as the (i)th result marking matrix, then the marking of the area where the starting point is located is completed, and the marked area is obtained; If the i+1th result labeling matrix is ​​different from the i-th result labeling matrix, the i+1th result labeling matrix is ​​expanded to obtain the i+1th expanded labeling matrix, and the i+1th expanded labeling matrix is ​​multiplied by the binary image to obtain the i+2th result labeling matrix, and this cycle is repeated until the i+nth result labeling matrix is ​​the same as the i+n-1th result labeling matrix, and a labeling area is obtained; n is an integer greater than or equal to 1; Step 35, setting the values ​​of the pixels in the area in the binary image corresponding to the marked area to 0; Step 36, based on a pixel point with a value of 1 in the binary image except the area corresponding to the marked area as the starting point, execute steps 32 to 35 until the values ​​of the pixel points in all areas of the binary image are 0, and use the last resulting marking matrix as the marked image; the marked image has multiple marked areas.

4. The method for ship detection in SAR images by arbitrary direction morphological analysis according to claim 1, characterized in that: The step 40 comprises: Step 41, setting the amplitude of the pixel points of the marked areas whose size is larger than the sum of the second preset size and the size of the convolution kernel among the multiple marked areas to 0, and each of the remaining marked areas is regarded as a first area; Step 42, calculating the average amplitude of each first region, the average amplitude of all first regions, and the amplitude standard deviation of all first regions; Step 43, among the multiple first regions, the amplitude of the pixel points in the first region whose difference between the average amplitude of each first region and the average amplitude of all the first regions is equal to three times the amplitude standard deviation is set to 0, and each of the remaining first regions is used as a preliminary detection region.

5. The method for ship detection in SAR images by arbitrary direction morphological analysis according to claim 1, characterized in that: The step 50 comprises: In each preliminary detection area, the pixel with the largest amplitude is taken as the center and expanded to the area with the same size as the convolution kernel to obtain a target window, and the area in the preliminary detection area except the target window is used as the background window.

6. The method for ship detection in SAR images by arbitrary direction morphological analysis according to claim 1, characterized in that: The step 60 comprises: Step 61, judging whether each pixel belongs to a pixel of a ship target to be detected according to the average amplitude of the background window and the amplitude of each pixel in the preliminary detection area; Step 62, if it does not belong, then the amplitude of the pixel point is set to 0, and the pixel point belonging to the ship target to be detected is the target main area.

7. The method for ship detection in SAR images by arbitrary direction morphological analysis according to claim 1, characterized in that: The step 70 comprises: Step 71, calculating the zero-order moment and the first-order moment of the target main area; Step 72: Calculate the center of the target body area according to the zero-order moment and the first-order moment.

8. The method for ship detection in SAR images by arbitrary direction morphological analysis according to claim 1, characterized in that: The step 80 comprises: Step 81, sampling from the center of the target main area to equally spaced positions on both sides; Step 82, taking the pixel point with the largest amplitude at each sampling position as the sampling center; Step 83, calculating the slope of the line between each of the sampling centers and the center of the target main area; Step 84, calculating an average value of a plurality of the slopes and an inclination angle corresponding to the average value of the slope; Step 85, if the tilt angle is greater than 45 degrees, then calculate the total number of pixel points with amplitude in the row where the sampling center is located and the maximum number of rows and the minimum number of rows with amplitude; If the tilt angle is less than or equal to 45 degrees, the total number of pixel points with amplitude in the column where the sampling center is located and the maximum number of columns and the minimum number of columns with amplitude are calculated; Step 86, if the tilt angle is greater than 45 degrees, determining the width of the target body area according to the average value of the number of pixels and the tilt angle, and determining the length of the target body area according to the maximum number of rows and the minimum number of rows and the tilt angle; If the tilt angle is less than or equal to 45 degrees, the width of the target body area is determined according to the average value of the number of pixels and the tilt angle, and the length of the target body area is determined according to the maximum number of columns, the minimum number of columns and the tilt angle.

Images