A method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images

Abstract

The present invention belongs to the field of image processing technology, and discloses a method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images, comprising the following steps: reading SAR images; roughly extracting floating ball cage-type sea surface aquaculture farms; precisely extracting floating ball cage-type sea surface aquaculture farms; and exporting the extraction results of floating ball cage-type sea surface aquaculture farms in the form of image files. Based on the backscattering intensity characteristics and spatial distribution characteristics of floating ball cage-type sea surface aquaculture farms in SAR images, the present invention innovatively proposes a "noise point removal method based on neighborhood highlight point distribution density calculation" and a "morphological optimization method based on spatial scale transformation", and innovatively constructs a two-step extraction method of "farm coarse extraction + farm fine extraction", and realizes accurate and rapid automated extraction of floating ball cage-type sea surface aquaculture farm areas through combined processing processes such as automatic threshold binarization, discrete noise point removal, morphological expansion, hole filling and morphological optimization.

Description

Technical Field

[0001] The invention belongs to the technical field of image processing, and in particular relates to a method for extracting a floating ball hanging cage type sea surface aquaculture farm based on SAR images. Background Art

[0002] Offshore farms refer to the cultivation of various aquatic products in specific sea areas by artificially placing fishery facilities and utilizing the natural marine environment. At present, they have become a strategic choice for major maritime countries. my country's offshore farms are mainly distributed in the coastal and offshore waters of the Yellow Sea, the East China Sea and the South China Sea. They are one of my country's important marine pillar industries and the main direction of my country's fishery development. The effective supervision of offshore farms is of great significance. Traditional offshore aquaculture monitoring often has problems such as difficulty, high cost and low efficiency. The remote sensing monitoring method of synthetic aperture radar (SAR) has the advantages of large observation range, low cost, all-day, all-weather, fast monitoring speed and high data accuracy. It is suitable for large-scale offshore farm monitoring. Through SAR remote sensing monitoring means, the spatial distribution and area of ​​offshore farms can be comprehensively and accurately obtained, which can provide a scientific basis for the planning and management of the aquaculture industry, improve the efficiency of supervision, and help to strengthen the supervision of offshore aquaculture activities. At the same time, it helps to reasonably layout the aquaculture area, optimize the aquaculture structure and improve the aquaculture efficiency. In addition, remote sensing monitoring of offshore aquaculture farms can timely discover the distribution of offshore aquaculture farms and their changes, prevent aquaculture facilities from becoming obstacles to maritime traffic, and thereby eliminate safety hazards, maintain the maritime navigation environment, and ensure smooth and safe navigation of ships.

[0003] Floating cage-type sea surface raft aquaculture is one of the main marine aquaculture methods in my country. The floating balls on the water are made of white or black rubber, with a diameter of about 30 cm and a spacing of about 75 cm. They are arranged in a non-continuous interval pattern within the layout area. The diameter of the underwater cage is about 20 cm and the length is 4 to 6 meters. Since the floating balls exposed to the sea surface of the floating cage-type sea surface aquaculture farm are small and easily affected by seawater conditions, seawater impurities and SAR imaging, the sea surface background noise in the SAR image is large, and the floating ball bright spot target is relatively blurred. How to accurately extract the aquaculture area based on the locally concentrated floating ball bright spot targets in the SAR image has high technical difficulty, and there is no public and effective extraction method yet. Summary of the invention

[0004] The purpose of the present invention is to provide a method for extracting buoy-cage type sea surface aquaculture farms based on SAR images. Based on the backscattering intensity characteristics (the backscattering intensity of the sea surface buoy target is stronger than that of the sea surface background area) and spatial distribution characteristics (a single aquaculture farm is distributed in polygonal blocks as a whole, and the buoy targets on the water are arranged at regular intervals) of the buoy-cage type sea surface aquaculture farms in SAR images, the present invention innovatively proposes a "noise point removal method based on neighborhood highlight point distribution density calculation" and a "morphological optimization method of spatial scale transformation", and innovatively constructs a two-step extraction method of "aquaculture farm coarse extraction + aquaculture farm fine extraction", and realizes accurate and rapid automatic extraction of buoy-cage type sea surface aquaculture farm areas through a combination of processing processes such as automatic threshold binarization, discrete noise point removal, morphological expansion, hole filling and morphological optimization.

[0005] The technical solution adopted by the present invention is a method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images, comprising the following steps:

[0006] S1, read SAR image;

[0007] S2, crude extraction floating ball cage type sea surface aquaculture farm, the specific steps include:

[0008] S21, automatically threshold binarizes the SAR image read by S1;

[0009] S22, removing noise from the image after binarization processing;

[0010] S23, dilation processing is performed on the image after noise removal;

[0011] S24, filling holes in the floating ball cage type sea surface aquaculture farm area in the image after expansion processing;

[0012] S3, fine extraction floating ball hanging cage sea surface aquaculture farm, the specific steps include:

[0013] S31, morphological optimization of the image of the floating ball cage type sea surface aquaculture farm roughly extracted in S2;

[0014] S32, performing morphological opening operation on the image after morphological optimization;

[0015] S33, performing Gaussian filtering on the image processed by the morphological opening operation;

[0016] S34, removing small areas of non-floating ball cage-type sea surface aquaculture farms from the image processed by Gaussian filtering, and finally obtaining the extraction result of floating ball cage-type sea surface aquaculture farms;

[0017] S4, exporting the extraction results of the floating ball cage type sea surface aquaculture farm in the form of an image file.

[0018] Furthermore, in S21, the specific steps of automatically performing threshold binarization on the SAR image read in S1 include:

[0019] S211, for the SAR image read in S1 , normalize all pixel values ​​to the range of [0, 255];

[0020] S212, adaptively calculate the optimal segmentation threshold by the maximum inter-class variance method, and Perform binarization to obtain the highlighted pixel part, where the highlighted pixel is marked as 1 and the other pixels are marked as 0, and the automatic threshold binary image is obtained. .

[0021] Furthermore, in S22, the noise point removal method based on the neighborhood highlight point distribution density calculation is used to remove noise from the binarized image, and the specific steps include:

[0022] S221, for binary images Any pixel in , the pixel is in the binary image The value in is , calculate the pixel in Highlight point distribution density within the neighborhood , the calculation formula is:

[0023] ;

[0024] in, n is the neighborhood radius; x For pixels in the binary image The number of rows in y For pixels in the binary image The number of columns in ; i For pixels in the binary image The number of rows in the index; j For pixels in the binary image The column number index in ;

[0025] S222, Initialization and Binary Image Binary images of the same size , set the threshold , the range is [0, 1], for binary images Any pixel in ,like:

[0026] and ,

[0027] In the binary graph The corresponding pixel is marked as 1, otherwise the pixel is marked as 0;

[0028] S223, traversing binary graph The above calculation and judgment are performed for each pixel in the binary image Assignment in.

[0029] Furthermore, in S23, the specific steps of dilating the image after noise removal are: selecting a radius of l The circular structural element is used to dilate the binary image obtained by S22 using the morphological expansion method. After dilation, a binary image is obtained. .

[0030] Furthermore, in said S24, the hole filling is performed on the floating ball cage type sea surface aquaculture farm area in the image after the expansion processing, and the specific steps are as follows: based on the binary image obtained in S23 , the flood filling algorithm is used to fill the isolated hole areas in the connected block areas with pixel value 1, and the binary image is obtained. .

[0031] Furthermore, in S31, the image of the floating ball cage type sea surface aquaculture farm extracted in S2 is morphologically optimized by a morphological optimization method based on spatial scale transformation. The specific steps are: setting the sampling coefficient S , and then for the binary graph conduct S times downsampling and S Upsample by times to get a binary image , edge smoothing is completed.

[0032] Furthermore, in the step S32, the specific steps of performing morphological opening operation on the image after morphological optimization are: selecting The rectangular structural elements of the binary graph Perform morphological erosion and morphological dilation operations to obtain a binary image ;

[0033] in, u is the length of the rectangular structural element; v is the width of the rectangular structure element.

[0034] Furthermore, in S33, the specific steps of performing Gaussian filtering on the image after the morphological opening operation are: The filter kernel size and Gaussian kernel standard deviation in both X and Y directions , for binary images Perform Gaussian filtering to obtain a binary image ;

[0035] in, k is the length / width of the filter kernel.

[0036] Furthermore, in said S34, the specific steps of removing the non-floating ball cage type sea surface aquaculture farm small area from the image after Gaussian filtering are as follows: the pixel number threshold of a single floating ball cage type sea surface aquaculture farm object is obtained by combining the actual resolution of the image, and the pixel number threshold of the single floating ball cage type sea surface aquaculture farm object is obtained by statistical binary image. The number of pixels contained in each independent pixel block in which all spatially connected pixel values ​​are 1 is compared with the pixel number threshold one by one, and the pixel blocks with pixel numbers less than the pixel number threshold are assigned a value of 0 as a whole to eliminate the extraction results. The non-buoyant cage-type sea surface aquaculture farm targets in the image include noise patches, small fishing boats, and other offshore facilities, and a binary image is obtained. .

[0037] The beneficial effects of the present invention are:

[0038] (1) A method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images is proposed, which can automatically complete the extraction of floating ball cage-type sea surface aquaculture farms;

[0039] (2) An innovative noise point removal method based on the calculation of the distribution density of neighborhood highlight points is proposed, which can effectively remove the discrete noise in the extraction results of floating ball cage-type sea surface aquaculture farms;

[0040] (3) An innovative morphological optimization method based on spatial scale transformation is proposed, which can effectively optimize the irregular morphological structure of the extraction results of floating ball cage-type sea surface aquaculture farms;

[0041] (4) It has good accuracy and can effectively extract the polygonal distribution range of floating ball cage-type sea surface aquaculture farms, which is basically consistent with the actual distribution of floating ball cage-type sea surface aquaculture farms;

[0042] (5) It has high processing efficiency. For a SAR image with a pixel size of 480×512, the processing time is less than 3 seconds (8-core processor, 1.60GHz, 12GB of running memory). BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the overall process of the present invention.

[0044] Figure 2 Schematic diagram of SAR image of a floating ball cage type sea surface aquaculture farm target in Example 1 of the present invention .

[0045] Figure 3 The binary image after the automatic threshold binarization processing by the maximum inter-class variance method in Example 1 of the present invention is .

[0046] Figure 4 The binary image after noise removal by the noise point removal method based on the neighborhood highlight point distribution density calculation in Example 1 of the present invention is .

[0047] Figure 5 The binary image after dilation processing by the morphological dilation method in Example 1 of the present invention .

[0048] Figure 6 This is a binary image of the floating ball cage type sea surface aquaculture farm area after the holes are filled by the flood filling algorithm in Example 1 of the present invention. .

[0049] Figure 7 The binary image after the spatial scale transformation morphological optimization processing in Example 1 of the present invention .

[0050] Figure 8 The binary image after morphological opening operation in Example 1 of the present invention .

[0051] Fig. 9 The binary image after Gaussian filtering in Example 1 of the present invention .

[0052] Fig.10 This is a binary image after the small area of ​​non-buoy cage-type sea surface aquaculture farm is eliminated in Example 1 of the present invention. .

[0053] Fig.11 This is a schematic diagram of the final extraction result and the original SAR image superposition of the buoy cage-type sea surface aquaculture farm in Example 1 of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solution and advantages of the present invention clearer and more understandable, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0055] A method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images, such as Figure 1 As shown, the following steps are included:

[0056] S1, read SAR image.

[0057] Read SAR images containing floating ball cage-type sea surface aquaculture farm targets .

[0058] S2, crude extraction floating ball cage type sea surface aquaculture farm.

[0059] This process is based on the read SAR image , the maximum inter-class variance method is used to automatically perform threshold image binarization, the noise point removal method based on the calculation of the neighborhood highlight point distribution density is used to remove discrete noise, the morphological expansion method is used for expansion processing, and the flood filling algorithm is used for hole filling to obtain the rough extraction results of each buoy cage-type sea surface aquaculture farm object as the input for the next step of buoy cage-type sea surface aquaculture farm fine extraction. The specific steps include:

[0060] S21, automatically threshold binarizes the SAR image read by S1, and the specific steps include:

[0061] S211, for the SAR image read in S1 , normalize all pixel values ​​to the range of [0, 255];

[0062] S212, the optimal segmentation threshold is adaptively calculated by the maximum inter-class variance method (Otsu) for the read SAR image Perform binarization to obtain the highlighted pixel part, where the highlighted pixel is marked as 1 and the other pixels are marked as 0, and the automatic threshold binary image is obtained. .

[0063] S22, removing noise from the image after the binarization process.

[0064] Since the SAR image is read The noise in the image is more obvious, and the highlight noise points are also identified as highlight points after binarization processing, but such highlight points are scattered and randomly distributed in the region, while the highlight pixels in the floating ball cage-type sea surface aquaculture farm area are locally concentrated. Therefore, the present invention innovatively proposes a "noise point removal method based on neighborhood highlight point distribution density calculation" to remove spatially discrete highlight noise pixel points. The specific steps include:

[0065] S221, for binary images Any pixel in , the pixel is in the binary image The value in is , calculate the pixel in Highlight point distribution density within the neighborhood , the calculation formula is:

[0066] ;

[0067] in, n is the neighborhood radius; x For pixels in the binary image The number of rows in y For pixels in the binary image The number of columns in ; i For pixels in the binary image The number of rows in the index; j For pixels in the binary image The column number index in ;

[0068] S222, Initialization and Binary Image Binary images of the same size , set the threshold , the range is [0, 1], for binary images Any pixel in ,like:

[0069] and ,

[0070] In the binary graph The corresponding pixel is marked as 1, otherwise the pixel is marked as 0;

[0071] S223, traversing binary graph The above calculation and judgment are performed for each pixel in the binary image Assignment in.

[0072] S23, dilation processing is performed on the image after noise removal.

[0073] In order to connect the unconnected local floating ball cage type sea surface aquaculture farm highlight points and block targets and construct a single floating ball cage type sea surface aquaculture farm object corresponding to the actual situation, a radius of l The circular structural element is used to dilate the binary image obtained by S22 using the morphological expansion method. The expansion process is performed to convert the local concentrated distribution of highlight pixel sets into a single floating ball cage type sea surface aquaculture farm object, and the binary image is obtained after processing. .

[0074] S24, filling holes in the floating ball suspended cage type sea surface aquaculture farm area in the image after expansion processing.

[0075] Since a single floating ball cage-type sea surface aquaculture farm object usually presents a polygonal block structure that is continuous in local space, the binary map obtained based on S23 , the flood fill algorithm is used to fill the isolated hole areas (corresponding to the local connected pixel set with pixel value 0) in the connected block area with pixel value 1 (corresponding to the local connected pixel set with pixel value 1). The pixels in the 0-value hole area in the area block where all the pixel values ​​are 1 are assigned a value of 1 as a whole, and a binary image is obtained. .

[0076] S3, fine extraction floating ball hanging cage type sea surface aquaculture farm.

[0077] Based on the rough extraction results of floating ball cage-type sea surface aquaculture farms, this process further optimizes the extraction results of floating ball cage-type sea surface aquaculture farms based on morphological optimization based on spatial scale transformation, morphological opening operation, Gaussian filtering and elimination of small areas of non-floating ball cage-type sea surface aquaculture farms, and finally obtains the refined extraction results of floating ball cage-type sea surface aquaculture farms that conform to the actual distribution form. The specific steps include:

[0078] S31, morphological optimization of the image of the floating ball cage type sea surface aquaculture farm roughly extracted in S2.

[0079] Since the actual floating ball cage type sea surface aquaculture farm objects usually show relatively regular polygonal block space features, in order to further obtain the macroscopic structural characteristics of the floating ball cage type sea surface aquaculture farm extraction results, this paper innovatively proposes a "morphological optimization method based on spatial scale transformation" to optimize the irregular morphological structure of the floating ball cage type sea surface aquaculture farm extraction results: setting the sampling coefficient S , and then for the binary graph conduct S Downsampling and S Subsampling times to get a binary image , edge smoothing is completed.

[0080] S32, performing morphological opening processing on the image after morphological optimization.

[0081] In order to smooth the contour of the extraction result of the floating ball cage type sea surface aquaculture farm, disconnect the narrow neck and eliminate the small protruding area, the The rectangular structure element of the binary graph Perform morphological opening operation, that is, select The rectangular structural elements of the binary graph Perform morphological erosion and morphological dilation operations to obtain a binary image ;

[0082] in, u is the length of the rectangular structural element; v is the width of the rectangular structure element.

[0083] S33, performing Gaussian filtering on the image processed by the morphological opening operation.

[0084] In order to further remove the discrete noise spots in the extraction results of floating ball cage type sea surface aquaculture farms and smooth the contours of the extraction results of floating ball cage type sea surface aquaculture farms, the The filter kernel size and Gaussian kernel standard deviation in the X (horizontal) and Y (horizontal) directions , for binary images Perform Gaussian filtering to obtain a binary image ;

[0085] in, k is the length / width of the filter kernel.

[0086] S34, removing small areas of non-floating ball cage-type sea surface aquaculture farms from the image processed by Gaussian filtering, and finally obtaining the extraction result of the floating ball cage-type sea surface aquaculture farms.

[0087] According to the actual remote sensing image data statistics, the area of ​​​​the floating ball cage-type sea surface aquaculture farm is generally larger than 100 square meters. Based on this feature and combined with the actual resolution of the image, the pixel number threshold of a single floating ball cage-type sea surface aquaculture farm object is obtained. The number of pixels contained in each independent pixel block in which all spatially connected pixel values ​​are 1 is compared with the pixel number threshold one by one, and the pixel blocks with pixel numbers less than the pixel number threshold are assigned a value of 0 as a whole to eliminate the extraction results. The non-buoyant cage-type sea surface aquaculture farm targets in the image include noise patches, small fishing boats, other offshore facilities, etc., and a binary image is obtained. , which is the final extraction result of the floating ball cage type sea surface aquaculture farm.

[0088] S4, exporting the extraction results of the floating ball cage type sea surface aquaculture farm in the form of an image file.

[0089] The extraction results of floating ball cage type sea surface aquaculture farm (binary image ) as an image file.

[0090] Example 1

[0091] In this embodiment, based on the Ka-band (K-above band) SAR image covering a certain sea area in my country (the area contains the target of the floating ball cage type sea surface aquaculture farm), the floating ball cage type sea surface aquaculture farm extraction method based on SAR image of the present invention is used to realize the extraction of the floating ball cage type sea surface aquaculture farm in the image, such as Figure 1 As shown, the specific steps include:

[0092] S1, read SAR image.

[0093] In MATLAB software, the SAR image file containing the target of the floating ball cage-type sea surface aquaculture farm is read through coding to obtain the SAR image ,like Figure 2 shown.

[0094] S2, crude extraction floating ball cage type sea surface aquaculture farm, the specific steps include:

[0095] S21, automatically threshold binarizes the SAR image read in S1.

[0096] Through MATLAB coding, for the read SAR image , normalize all pixel values ​​to the range of [0, 255]; then adaptively calculate the optimal segmentation threshold by the maximum inter-class variance method, and Perform binarization to obtain the highlighted pixel part, where the highlighted pixel is marked as 1 and the other pixels are marked as 0, and the automatic threshold binary image is obtained. ,like Figure 3 shown.

[0097] S22, removing noise from the image after the binarization process.

[0098] Through MATLAB coding, the noise point removal method based on neighborhood highlight point distribution density calculation proposed in the present invention is used to remove the binary image. Calculate the distribution density of highlight points in the 5×5 neighborhood of any pixel in the image, set the density threshold to 0.4, and initialize the image with the binary value. Binary images of the same size , the binary image In the binary image, the pixel with a distribution density of highlight points greater than 0.4 and a pixel value of 1 is (like Figure 4 ) will assign the corresponding pixel value to 1, and the remaining pixels will be assigned to 0.

[0099] S23, dilation processing is performed on the image after noise removal.

[0100] Through MATLAB coding, a circular structural element with a radius of 3 is selected and the binary image obtained in S22 is morphologically expanded using the method of morphological expansion. After dilation, a binary image is obtained. ,like Figure 5 shown.

[0101] S24, filling holes in the floating ball suspended cage type sea surface aquaculture farm area in the image after expansion processing.

[0102] Binary image obtained based on S23 through MATLAB coding , using flood fill algorithm for binary graph The pixels in the 0-value hole area in the connected block area with all pixel values ​​1 are assigned a value of 1 as a whole, and a binary image is obtained. ,like Figure 6 shown.

[0103] S3, fine extraction floating ball hanging cage sea surface aquaculture farm, the specific steps include:

[0104] S31, morphological optimization is performed on the image of the floating ball cage-type sea surface aquaculture farm roughly extracted in S2, and edge smoothing is completed.

[0105] Through MATLAB coding, the morphological optimization method based on spatial scale transformation proposed in the present invention is used, and the sampling coefficient is set to 4. Perform 4 times downsampling and 4 times upsampling to obtain a binary image (like Figure 7 ), and complete edge smoothing.

[0106] S32, performing morphological opening processing on the image after morphological optimization.

[0107] Through MATLAB coding, a 5×5 rectangular structural element is selected to perform binary image Perform morphological opening operation to obtain a binary image ,like Figure 8 shown.

[0108] S33, performing Gaussian filtering on the image processed by the morphological opening operation.

[0109] Through MATLAB coding, a 3×3 filter kernel with a Gaussian kernel standard deviation of 3 in both the X (horizontal direction) and Y (horizontal direction) directions is used to filter the binary image. Perform Gaussian filtering to obtain a binary image ,like Fig. 9 shown.

[0110] S34, removing small areas of non-floating ball cage-type sea surface aquaculture farms from the image processed by Gaussian filtering, and finally obtaining the extraction result of the floating ball cage-type sea surface aquaculture farms.

[0111] Through MATLAB coding, set the pixel number threshold to 1000 and calculate the statistical binary map The number of pixels contained in each independent pixel block in which all spatially connected pixel values ​​are 1 is compared with the pixel number threshold one by one, and the pixel blocks with pixel numbers less than the pixel number threshold are assigned a value of 0 as a whole to obtain a binary map ,like Fig.10 As shown, the binary image This is the final extraction result of the floating ball cage type sea surface aquaculture farm. Fig.11 It is a schematic diagram of the superposition of the original SAR image and the extraction result of the floating ball cage type sea surface aquaculture farm of the present invention.

[0112] S4, exporting the extraction results of the floating ball cage type sea surface aquaculture farm in the form of an image file.

[0113] In MATLAB software, the extraction results of floating ball cage type sea surface aquaculture farm (binary image) were converted into ) as an image file.

[0114] In this embodiment, the total time consumed for extracting and processing the floating ball cage-type sea surface aquaculture farm target of the SAR image (480×512 pixel size) is 2.87 seconds (8-core processor, 1.60 GHz, running memory 12 GB).

Claims

1. A method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images, characterized in that: The following steps are involved: S1, read SAR image; S2, crude extraction floating ball cage type sea surface aquaculture farm, the specific steps include: S21, automatically threshold binarizes the SAR image read by S1; S22, using a noise point removal method based on neighborhood highlight point distribution density calculation to remove noise from the binarized image, the specific steps include: S221, for binary images Any pixel in , the pixel is in the binary image The value in is , calculate the pixel in Highlight point distribution density within the neighborhood , the calculation formula is: ； in, n is the neighborhood radius; x For pixels in the binary image The number of rows in y For pixels in the binary image The number of columns in ; i For pixels in the binary image The number of rows in the index; j For pixels in the binary image The column number index in ; S222, Initialization and Binary Image Binary images of the same size , set the threshold , the range is [0, 1], for binary images Any pixel in ,like: and , In the binary graph The corresponding pixel is marked as 1, otherwise the pixel is marked as 0; S223, traversing binary graph The above calculation and judgment are performed for each pixel in the binary image Assignment in S23, dilation processing is performed on the image after noise removal; S24, filling holes in the floating ball cage type sea surface aquaculture farm area in the image after expansion processing; S3, fine extraction floating ball hanging cage sea surface aquaculture farm, the specific steps include: S31, morphological optimization of the image of the floating ball cage type sea surface aquaculture farm roughly extracted in S2; S32, performing morphological opening operation on the image after morphological optimization; S33, performing Gaussian filtering on the image processed by the morphological opening operation; S34, removing small areas of non-floating ball cage-type sea surface aquaculture farms from the image processed by Gaussian filtering, and finally obtaining the extraction result of floating ball cage-type sea surface aquaculture farms; S4, exporting the extraction results of the floating ball cage type sea surface aquaculture farm in the form of an image file.

2. The method for extracting floating ball suspended cage type sea surface aquaculture farms based on SAR images according to claim 1, characterized in that: In S21, the specific steps of automatically performing threshold binarization on the SAR image read in S1 include: S211, for the SAR image read in S1 , normalize all pixel values ​​to the range of [0, 255]; S212, adaptively calculate the optimal segmentation threshold by the maximum inter-class variance method, and Perform binarization to obtain the highlighted pixel part, where the highlighted pixel is marked as 1 and the other pixels are marked as 0, and the automatic threshold binary image is obtained. .

3. The method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images according to claim 1, characterized in that: In S23, the specific steps of dilating the image after noise removal are: selecting a radius of l The circular structural element is used to dilate the binary image obtained by S22 using the morphological expansion method. After dilation, a binary image is obtained. .

4. The method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images according to claim 3 is characterized in that: In S24, the hole filling is performed on the floating ball cage type sea surface aquaculture farm area in the image after the expansion process. The specific steps are: based on the binary image obtained in S23 , the flood filling algorithm is used to fill the isolated hole areas in the connected block areas with pixel value 1, and the binary image is obtained. .

5. The method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images according to claim 4 is characterized in that: In the above S31, the image of the floating ball cage type sea surface aquaculture farm extracted in S2 is morphologically optimized by a morphological optimization method based on spatial scale transformation. The specific steps are: setting the sampling coefficient S , and then for the binary graph conduct S times downsampling and S Upsample by times to get a binary image , edge smoothing is completed.

6. The method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images according to claim 5 is characterized in that: In S32, the specific steps of performing morphological opening operation on the image after morphological optimization are: The rectangular structural elements of the binary graph Perform morphological erosion and morphological dilation operations to obtain a binary image ; in, u is the length of the rectangular structural element; v is the width of the rectangular structure element.

7. The method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images according to claim 6 is characterized in that: In S33, the specific steps of performing Gaussian filtering on the image after the morphological opening operation are: The filter kernel size and Gaussian kernel standard deviation in both X and Y directions , for binary images Perform Gaussian filtering to obtain a binary image ; in, k is the length / width of the filter kernel.

8. The method for extracting floating ball cage-type sea surface aquaculture farms based on SAR images according to claim 7 is characterized in that: In S34, the specific steps of removing the non-floating ball cage type sea surface aquaculture farm small area from the image after Gaussian filtering are as follows: the pixel number threshold of a single floating ball cage type sea surface aquaculture farm object is obtained by combining the actual resolution of the image, and the pixel number threshold of the single floating ball cage type sea surface aquaculture farm object is obtained by statistical binary image. The number of pixels contained in each independent pixel block in which all spatially connected pixel values ​​are 1 is compared with the pixel number threshold one by one, and the pixel blocks with pixel numbers less than the pixel number threshold are assigned a value of 0 as a whole to eliminate the extraction results. The non-buoyant cage-type sea surface aquaculture farm targets in the image include noise patches, small fishing boats, and other offshore facilities, and a binary image is obtained. .

Images