A ship detection method for PolSAR images based on PolSAR-SIFT key points

Abstract

The present invention discloses a PolSAR image ship detection method based on PolSAR-SIFT key points, including: collecting multiple original PolSAR images of the same scene in different polarization modes; obtaining the polarization covariance matrix of the image according to the multiple PolSAR images; constructing a PolSAR-Harris response function image using the polarization covariance matrix and extracting all key points in the current scene to form a key point coordinate set; obtaining a preselected target area according to the key point coordinate set, and screening out one or more target areas through a change indication parameter; performing binary segmentation on the SPAN image of the target area to obtain a binary image, and completing ship detection using the binary image. The present invention defines a gradient based on exponentially weighted Riemann distance for PolSAR data, extends the SAR-SIFT key point detection algorithm for single-channel SAR data to PolSAR data, introduces a change indication parameter when obtaining the ship target area, and uses the parameter to screen out potential target areas on the sea surface, thereby reducing the false alarm rate.

Description

Technical Field

[0001] The invention belongs to the field of radar technology, and in particular relates to a PolSAR image ship detection method based on PolSAR-SIFT key points, which can be used for ship detection in PolSAR data ocean scenes. Background Art

[0002] my country is a maritime power with vast territorial waters and a long coastline. Ships are the main tools for marine fisheries, transportation and other activities. Monitoring ships at sea can effectively manage fisheries and marine transportation, and can also perceive the situation at sea in advance. Traditional ship monitoring methods such as VMS (Vessel Monitoring System), VTS (Vessel Traffic Service) and AIS (Automatic Identification System) can usually only monitor cooperative targets and are far from meeting actual needs.

[0003] The research on radar imaging technology began in the 1950s and has made rapid progress in the following sixty years. SAR (Synthetic Aperture Radar), as an active microwave sensor, has the characteristics of all-weather, all-day, and high resolution. It has unique advantages in responding to emergencies and fishery monitoring. Therefore, using SAR images for ship detection is of great significance for monitoring ships at sea. Today, SAR has become an important means of ship detection, and a number of classic SAR image ship detection algorithms have emerged.

[0004] However, for small ships, their radar echo intensity is weak, and the contrast between small ships and sea clutter in SAR images is low. It is difficult for ship detection algorithms based on SAR images to detect such ship targets. PolSAR (Polarimetric Synthetic Aperture Radar) images contain not only the radar echo intensity information of the target, but also the polarization information of the target, which can further distinguish ships from sea clutter. Therefore, ship detection methods based on PolSAR images have also attracted widespread attention. For example, detectors based on optimization technology: Optimal Polarimetric Detector (OPD), Polarimetric Whitening Filter (PWF), etc.; detectors based on scattering mechanism: Reflection Symmetry Detector (RSD), Polarimetric Notch Filter (PNF), etc.; detectors based on neighborhood information: Ship detector based on superpixel-level scattering mechanism, etc.

[0005] There are literatures that apply the SIFT (Scale-invariant feature transform) key point detection algorithm for optical images to ship detection in PolSAR images, which can detect larger ships. Since the scattering characteristics of small ships are different from those of large ships, this method misses more small ships. Moreover, due to the coherent imaging mechanism of the SAR system, speckle noise will appear in the SAR image, which increases the difficulty of key point extraction. Therefore, directly applying the traditional SIFT key point detection algorithm to PolSAR images will affect the performance of key point detection. Summary of the invention

[0006] In order to solve the above problems existing in the prior art, the present invention provides a PolSAR image ship detection method based on PolSAR-SIFT key points. The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0007] The present invention provides a PolSAR image ship detection method based on PolSAR-SIFT key points, comprising:

[0008] S1: Collect multiple original PolSAR images of the same scene with different polarization modes;

[0009] S2: Obtaining a polarization covariance matrix of an image according to the plurality of PolSAR images;

[0010] S3: constructing a PolSAR-Harris response function image using the polarization covariance matrix and extracting all key points in the current scene to form a key point coordinate set;

[0011] S4: obtaining a pre-selected target area according to the key point coordinate set, and screening out one or more target areas by changing the indication parameter;

[0012] S5: performing binary segmentation on the SPAN image of the target area to obtain a binary image, and using the binary image to complete ship detection.

[0013] In one embodiment of the present invention, the S2 includes:

[0014] S21: Using three PolSAR images I HH ,I HV and I VV The elements with the same coordinates in construct a set of polarization scattering vectors k(x,y)=[I HH (x,y),I HV (x,y),I VV (x,y)] T , where I HH (x,y),I HV (x,y) and I VV (x,y) corresponds to image I HH ,I HV and I VV The complex scattering coefficient at the (x,y) coordinate;

[0015] S22: construct a polarization covariance matrix C corresponding to the pixel point (x, y) according to the polarization scattering vector k(x, y):

[0016]

[0017] Among them, <·> represents the ensemble average, k1=I HH (x,y), k2=I HV (x,y), k3=I VV (x,y).

[0018] In one embodiment of the present invention, S3 includes:

[0019] S31: According to the polarization covariance matrix C and the exponential weighting operator, the exponential weighted polarization covariance matrix MC of the four directions of up, down, left and right at each pixel point (a, b) in the image is obtained. u,α , MC d,α , MCl,α and MC r,α :

[0020]

[0021] Among them, C(a+x,b+y) represents the polarization covariance matrix corresponding to the pixel point (a+x,b+y), α represents the scale parameter in the exponential weighting operator, R represents the real number domain, and R - represents the field of negative real numbers, R + represents the field of positive real numbers.

[0022] S32: Using the exponentially weighted polarization covariance matrix MC u,α , MC d,α , MC l,α and MC r,α Construct PolSAR-Harris response function image;

[0023] S33: Extract key points in the current scene using the PolSAR-Harris response function image.

[0024] In one embodiment of the present invention, the S32 includes:

[0025] S321: Using the exponentially weighted polarization covariance matrix MC u,α , MC d,α , MC l,α and MC r,α Get the exponentially weighted Riemann distance RD in the vertical direction v,α and the exponentially weighted Riemann distance RD in the horizontal direction h,α :

[0026]

[0027]

[0028] Where tr(·) represents the trace of the matrix;

[0029] S322: Using the exponentially weighted Riemann distance RD in the vertical direction v,α and the exponentially weighted Riemann distance RD in the horizontal direction h,α Define the vertical gradient G of the image RDv,α and the horizontal gradient G of the image RDh,α :

[0030] G RDv,α =log(RD v,α ),G RDh,α =log(RD h,α );

[0031] S323: Using the vertical gradient GRDv,α and the horizontal gradient G RDh,α Define the PolSAR-Harris matrix C PSH , and using the PolSAR-Harris matrix C PSH Get the PolSAR-Harris response function image R PSH :

[0032]

[0033] R PSH (x,y,α)=det(C PSH (x,y,α))-d·tr(C PSH (x,y,α) 2

[0034] Among them, det(·) means to find the determinant of the matrix, d is the empirical parameter, A standard deviation of Gaussian convolution kernel.

[0035] In one embodiment of the present invention, the S33 includes:

[0036] Select the PolSAR-Harris response function image R PSH All local maximum points in the are taken as pre-selected key points, and then the pre-selected key points are screened according to a preset key point threshold, and points greater than the key point threshold are retained as final key points.

[0037] In one embodiment of the present invention, the S4 includes:

[0038] S41: According to the key point coordinate k i Get the pre-selected target area p i , the pre-selected target area is a square area centered on the key point;

[0039] S42: Calculate the current pre-selected area p i The covariance matrix C of each pixel in (m,n) With the current pre-selected target area p i The covariance matrix C of the central pixel center The similarity measure s KL (C center ,C (m,n) ):

[0040]

[0041] Where h is the adjustment parameter, (m,n) is the coordinate of the pixel in the current pre-selected area;

[0042] S43: The similarity measurement between each pixel and the central pixel is calculated according to the corresponding pixel position to form a similarity measurement image I s (p i ), and according to the similarity measurement image I s (p i ) Construct the change indication parameter P of the current pre-selected area v (p i ):

[0043]

[0044] Among them, std[·] represents the standard deviation of the elements in the calculated image, and mean[·] represents the mean of the elements in the calculated image;

[0045] S44: All pre-selected areas p i The change of the indicator parameter P v (p i ) is compared with the set change indication parameter threshold, and the change indication parameter P is selected v (p i ) greater than the change indication parameter threshold constitutes the final target area set S fa {p fj ,j=1,...,M}.

[0046] In one embodiment of the present invention, the S5 includes:

[0047] S51: Calculate the target area p j SPAN image:

[0048] SPAN(p fj )={|I HH (c,r)| 2 +2|I HV (c,r)| 2 +|I VV (c,r)| 2 , c∈[x j -L,x j +L],r∈[y j -L,y j +L]}

[0049] Among them, SPAN(p fj ) represents the target area p fj The polarization total power image composed of the corresponding pixel positions of the polarization total power calculated at the mid-coordinate point (c, r);

[0050] S52: Compare the pixel value in the polarization total power image with a set polarization total power threshold, and the position where the pixel value is greater than the polarization total power threshold is the position of the ship.

[0051] Another aspect of the present invention provides a storage medium storing a computer program for executing the steps of the PolSAR image ship detection method based on PolSAR-SIFT key points described in any one of the above embodiments.

[0052] Another aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, it implements the steps of the PolSAR image ship detection method based on PolSAR-SIFT key points as described in any one of the above embodiments.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] 1. When acquiring key points, the PolSAR image ship detection method based on PolSAR-SIFT key points of the present invention defines a gradient based on exponentially weighted Riemann distance for PolSAR data, and extends the SAR-SIFT key point detection algorithm for single-channel SAR data to PolSAR data.

[0055] 2. The PolSAR image ship detection method of the present invention searches for areas with angular structures on the sea surface as potential ship target areas through a key point detection algorithm, and introduces a change indication parameter based on regional information when obtaining the final ship target area. This parameter is used to screen out potential target areas on the sea surface, thereby reducing false alarms.

[0056] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a flow chart of a PolSAR image ship detection method based on PolSAR-SIFT key points provided by an embodiment of the present invention;

[0058] Figure 2 These are three original PolSAR images obtained in different polarization modes of the same scene provided by an embodiment of the present invention;

[0059] Figure 3 is a schematic diagram of 3×3 neighborhood elements around a current pixel point (x, y) provided by an embodiment of the present invention;

[0060] Figure 4 yes Figure 2 A marker image of a ship target in the scene;

[0061] Figure 5The PolSAR image ship detection method based on PolSAR-SIFT key points according to the embodiment of the present invention is used to detect Figure 2 Binary image obtained by detecting ships in the scene;

[0062] Figure 6 The existing PNF method is used to Figure 2 Binary image obtained by ship detection in the scene;

[0063] Figure 7 The existing RS method is used to Figure 2 Binary image obtained by detecting ships in the scene. DETAILED DESCRIPTION

[0064] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, a PolSAR image ship detection method based on PolSAR-SIFT key points proposed by the present invention is described in detail below in combination with the accompanying drawings and specific implementation methods.

[0065] The above and other technical contents, features and effects of the present invention are clearly presented in the following detailed description of the specific implementation modes in conjunction with the accompanying drawings. Through the description of the specific implementation modes, the technical means and effects adopted by the present invention to achieve the predetermined purpose can be more deeply and specifically understood. However, the attached drawings are only for reference and explanation purposes and are not used to limit the technical solutions of the present invention.

[0066] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants are intended to cover non-exclusive inclusion, so that an article or device including a series of elements includes not only those elements, but also other elements that are not explicitly listed. In the absence of more restrictions, the elements defined by the statement "including one..." do not exclude the existence of other identical elements in the article or device including the elements.

[0067] See also Figure 1 , Figure 1 1 is a flow chart of a PolSAR image ship detection method based on PolSAR-SIFT key points provided by an embodiment of the present invention. The PolSAR image ship detection method includes:

[0068] S1: Collect multiple original PolSAR images of the same scene in different polarization modes.

[0069] Specifically, three different PolSAR images I of the same scene are collected. HH ,I HV and I VV , these three PolSAR images I HH ,I HV and I VV These are images of the same scene captured by antennas with different polarization modes and have the same size, such as Figure 2 As shown, Figure 2 are three original PolSAR images obtained by different polarization modes of the same scene provided by an embodiment of the present invention, wherein FIG. (a) is image I HH , Figure (b) is image I HV , Figure (c) is image I VV .

[0070] S2: Obtaining a polarization covariance matrix of an image according to the multiple PolSAR images.

[0071] Specifically, using three PolSAR images I HH ,I HV and I VV The elements with the same coordinates in construct a set of polarization scattering vectors k(x,y)=[I HH (x,y),I HV (x,y),I VV (x,y)] T , where I HH (x,y),I HV (x,y) and I VV (x,y) corresponds to image I HH ,I HV and I VV The complex scattering coefficient at the (x, y) coordinate. Later, for convenience of representation, the form of k(x, y) is simplified to k = [k1, k2, k3] T ,k1,k2,k3 are respectively HH (x,y),I HV (x,y),I VV (x, y) one-to-one correspondence, that is, k1 = I HH (x,y), k2=I HV (x,y), k3=I VV (x,y).

[0072] It should be noted that, since each pixel point of the image corresponds to a polarization scattering vector, the number of polarization scattering vectors formed is equal to the number of pixels of the PolSAR image.

[0073] Then, the polarization covariance matrix C corresponding to the pixel point (x, y) is constructed according to the polarization scattering vector k(x, y):

[0074]

[0075] Here, <·> represents the set average. In this embodiment, the elements of the 3×3 neighborhood around the current pixel are selected as a set. Figure 3 , Figure 3 is a schematic diagram of 3×3 neighborhood elements around the current pixel point (x, y) provided by an embodiment of the present invention. Specifically, The calculation formulas for other elements in the polarization covariance matrix C refer to <|k1| 2 > and The calculation process of is not repeated here. The polarization covariance matrix C(x,y) corresponding to the coordinate (x,y) can be obtained from the above formula.

[0076] Furthermore, the polarization covariance matrix corresponding to each pixel in the image can be obtained.

[0077] S3: construct a PolSAR-Harris response function image using the polarization covariance matrix and extract all key points in the current scene to form a key point coordinate set.

[0078] Specifically, S3 includes:

[0079] S31: Use the polarization covariance matrix C and the exponential weighting operator to obtain the exponentially weighted polarization covariance matrix MC in the four directions of up, down, left and right for each pixel in the image u,α , MC d,α , MC l,α , MC r,α .

[0080] Taking a given coordinate (a, b) as an example, the exponentially weighted polarization covariance matrix MC in the four directions of up, down, left, and right is u,α (a,b),MC d,α (a,b),MC l,α (a,b),MC r,α (a,b) are:

[0081]

[0082] Among them, C(a+x,b+y) represents the polarization covariance matrix corresponding to the pixel (a+x,b+y), α represents the scale parameter in the exponential weighting operator, R represents the real number domain, and R - represents the field of negative real numbers, R + represents the field of positive real numbers.

[0083] S32: Using the exponentially weighted polarization covariance matrix MC u,α , MC d,α , MC l,α and MC r,α Construct the PolSAR-Harris response function image.

[0084] Specifically, we first use the exponentially weighted polarization covariance matrix MC in the above four directions u,α , MC d,α , MC l,α , MC r,α Get the exponentially weighted Riemann distance RD in the vertical direction at the current coordinate v,α and the exponentially weighted Riemann distance RD in the horizontal direction h,α :

[0085]

[0086]

[0087] Here, tr(·) represents the trace of the matrix.

[0088] Then, the exponentially weighted Riemann distance RD in the vertical direction is used v,α and the exponentially weighted Riemann distance RD in the horizontal direction h,α Define the vertical gradient G at the current coordinate of the image RDv,α and the horizontal gradient G RDh,α :

[0089] G RDv,α =log(RD v,α ),G RDh,α =log(RD h,α )

[0090] Further, using the vertical gradient G RDv,α and the horizontal gradient G RDh,α Define the PolSAR-Harris matrix C PSH , and using the PolSAR-Harris matrix C PSH Get the PolSAR-Harris response function image R PSH :

[0091]

[0092] R PSH (x,y,α)=det(C PSH (x,y,α))-d·tr(C PSH (x,y,α) 2

[0093] Wherein, det(·) represents the determinant of the matrix, and d is a parameter set based on experience. In this example, d=0.04 is set. A standard deviation of The Gaussian convolution kernel is set to α=2.51 in this example.

[0094] It should be noted that the vertical gradient and horizontal gradient corresponding to each coordinate point in the image can be obtained by using the above process, and the PolSAR-Harris response function image R can be obtained according to the vertical gradient and horizontal gradient corresponding to each coordinate point PSH The corresponding coordinate values ​​in , so all the values ​​together form the PolSAR-Harris response function image R according to the corresponding coordinate positions PSH .

[0095] S33: Extract key points in the current scene using the PolSAR-Harris response function image.

[0096] First, select the PolSAR-Harris response function image R PSH All local maximum points in the are used as pre-selected key points. The local maximum point means that in a predetermined area, if the central pixel value is larger than the surrounding pixel values, then the central pixel value is the local maximum. In this embodiment, the predetermined area is a 3×3 area, that is, if the central pixel value is larger than the surrounding 8 pixel values, then the central pixel value is the local maximum. PSH Each pixel value in is used as the central pixel value for judgment, so as to select all local maximum points, and these local maximum points are used as pre-selected key points.

[0097] Then, the pre-selected key points are screened according to the preset key point threshold, and the points greater than the key point threshold are retained as the final key points. The preset key point threshold in this example is the maximum pixel value of the response function image multiplied by 10 -5 .

[0098] After screening, the coordinates of all final key points are used to form a set S k ={k i =(x i ,y i ),i=1,...,N}, where k i is the key point, (x i ,y i ) is the coordinate of the key point, i is the serial number of the key point, and N is the number of key points.

[0099] S4: obtaining a pre-selected target area according to the key point coordinate set, and filtering out the target area through the change indication parameter.

[0100] Specifically, each key point k i Corresponding to a pre-selected target area p i , the pre-selected target area is a key point k i The set S of pre-selected target areas is a square area with a side length of 2L+1 and a center of pa ={p i ,i=1,...N}.

[0101] Calculate the current pre-selected area p i The covariance matrix C of each pixel in (m,n) With the current pre-selected target area p i The covariance matrix C of the central pixel center The similarity measure s KL (C center ,C (m,n) ), where (m,n) is the coordinate of the pixel in the current pre-selected area, and the value range of m is x i -L to x i +L, the value range of n is y i -L to y i +L, similarity measure s KL It is defined by the symmetric Kullback–Leibler divergence, which is expressed as follows:

[0102]

[0103] Here, h is an adjustment parameter, and h=5 is set in this example.

[0104] Next, the similarity measurement of each pixel and the central pixel is calculated according to the corresponding pixel position to form a similarity measurement image I s (p i ), and according to the similarity measurement image I s (p i ) Construct the change indication parameter P of the current pre-selected area v (p i ), whose expression is:

[0105]

[0106] Among them, std[·] represents the standard deviation of the elements in the calculated image, and mean[·] represents the mean of the elements in the calculated image.

[0107] Then, all pre-selected areas p i The change of the indicator parameter P v (pi ) is compared with the set change indication parameter threshold, and the change indication parameter P is selected v (p i ) greater than the change indication parameter threshold constitutes the final target area set S fa {p fj ,j=1,...,M}. In practice, we can i The change of the indicator parameter P v (p i ) range to determine and adjust the size of the change indication parameter threshold.

[0108] S5: Perform binary segmentation on the selected target area to complete ship detection.

[0109] Specifically, calculate the set S of the above final target area fa Each target region p fj The SPAN (polarimetric total power) image:

[0110] SPAN(p fj )={|I HH (c,r)| 2 +2|I HV (c,r)| 2 +|I VV (c,r)| 2 , c∈[x j -L,x j +L],r∈[y j -L,y j +L]}

[0111] Among them, SPAN(p fj ) represents the target area p fj The polarization total power calculated at each coordinate point in the image is composed of the polarization total power image composed of the corresponding pixel position. In other words, the above formula is used to calculate the current target area p fj The total polarization power of each pixel (c, r) in the current target area p fj The total polarization power of all points together constitutes the target area p fj Then, the elements in the SPAN image are compared with the set polarization total power threshold, and the positions greater than the threshold are set to 1, and the positions less than the threshold are set to 0, to obtain the binary image I d In this embodiment, the final binary image I d The position where the value is 1 is the location of the ship, thus completing the ship detection.

[0112] In summary, the embodiment of the present invention defines an exponentially weighted average polarization covariance matrix for PolSAR image data using a polarization covariance matrix and an exponentially weighted average operator, and then defines a gradient operator based on the PolSAR image using the matrix and the Riemann distance. The gradient operator is combined with the SAR-Harris matrix and the SAR-Harris response function in the traditional SAR-SIFT key point detection algorithm to define a PolSAR-Harris matrix and a PolSAR-Harris response function for PolSAR images, and by setting the scale parameter in the weighted average operator, a PolSAR-Harris matrix and a PolSAR-Harris response function image for extracting key points are generated, and then a set of local maximum points is screened out in the PolSAR-Harris response function image as a set of pre-selected key points, and then the local maximum points are screened by a set threshold, and the set of points greater than the threshold is used as the final set of key points.

[0113] The effect of the PolSAR image ship detection method based on PolSAR-SIFT key points according to an embodiment of the present invention is further illustrated by experiments below.

[0114] (1) Experimental scenario: The image data used in the present invention is PolSAR images collected at Yokohama Port in Japan, such as Figure 2 As shown, Figure (a), Figure (b), and Figure (c) are three original PolSAR images obtained with different polarization modes for the same scene.

[0115] (2) Experimental content:

[0116] Experiment 1: Using the method proposed in the embodiment of the present invention to Figure 2 The test results are as follows Figure 5 As shown; Experiment 2: Using the existing PNF (polarimetric notch filter, polarization notch filter) detection method Figure 2 The test results are as follows Figure 6 As shown; Experiment 3: Using the existing RS (reflection symmetric, reflection symmetric detector) detection method Figure 2 The test results are as follows Figure 7 shown.

[0117] (3) Experimental results analysis

[0118] from Figure 5 , Figure 6 and Figure 7It can be seen that compared with the existing PNF detection method and RS detection method, the method in the embodiment of the present invention uses PolSAR-SIFT key points to pre-select the target area, effectively reduces the influence of coherent speckle noise, and introduces change indication parameters, which can detect more accurate ship areas.

[0119] The ship detection results obtained by the above three methods are compared with Figure 4 The detection results of the three methods are quantitatively compared with the ship marks in

[0120] Table 1 Comparison of detection performance of the method of the present invention and other methods

[0121] method Correctly detect the target Number of false alarm targets Method of the present invention 33 0 PNF 33 >3 RS 30 >3

[0122] It can be seen from Table 1 that, compared with the two existing detection methods, the method in the embodiment of the present invention reduces the false alarm rate while ensuring a high detection rate, thereby improving the detection performance.

[0123] In summary, the PolSAR image ship detection method based on PolSAR-SIFT key points in this embodiment defines a gradient based on exponentially weighted Riemann distance for PolSAR data when acquiring key points, and extends the key point detection algorithm for single-channel SAR data to PolSAR data. This method searches for areas with angular structures on the sea surface as potential ship target areas through the key point detection algorithm, and introduces a change indication parameter based on regional information when acquiring the final ship target area. The parameter is used to screen out potential target areas on the sea surface, reducing false alarms.

[0124] Another embodiment of the present invention provides a storage medium, wherein a computer program is stored in the storage medium, and the computer program is used to execute the steps of the PolSAR image ship detection method based on PolSAR-SIFT key points described in the above embodiment. Another aspect of the present invention provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor calls the computer program in the memory, the steps of the PolSAR image ship detection method based on PolSAR-SIFT key points described in the above embodiment are implemented. Specifically, the above integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above software function module is stored in a storage medium, including several instructions for enabling an electronic device (which can be a personal computer, a server, or a network device, etc.) or a processor (processor) to execute some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.

[0125] 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 PolSAR image ship detection method based on PolSAR-SIFT key points, characterized in that: include: S1: Collect multiple original PolSAR images of the same scene with different polarization modes; S2: Obtaining a polarization covariance matrix of an image according to the plurality of PolSAR images; S3: constructing a PolSAR-Harris response function image using the polarization covariance matrix and extracting all key points in the current scene to form a key point coordinate set; S4: obtaining a pre-selected target area according to the key point coordinate set, and screening out one or more target areas by changing the indication parameter; S5: performing binary segmentation on the SPAN image of the target area to obtain a binary image, and using the binary image to complete ship detection, The S3 includes: S31: According to the polarization covariance matrix C and the exponential weighting operator, the exponential weighted polarization covariance matrix MC of the up, down, left and right directions of each pixel point (a, b) in the image is obtained. u,α , MC d,α , MC l,α and MC r,α : Among them, C(a+x,b+y) represents the polarization covariance matrix corresponding to the pixel point (a+x,b+y), α represents the scale parameter in the exponential weighting operator, R represents the real number domain, and R - represents the field of negative real numbers, R + represents the field of positive real numbers; S32: Using the exponentially weighted polarization covariance matrix MC u,α , MC d,α , MC l,α and MC r,α Construct PolSAR-Harris response function image; S33: extracting key points in the current scene using the PolSAR-Harris response function image, The S32 includes: S321: Using the exponentially weighted polarization covariance matrix MC u,α , MC d,α , MC l,α and MC r,α Get the exponentially weighted Riemann distance RD in the vertical direction v,α and the exponentially weighted Riemann distance RD in the horizontal direction h,α : Where tr(·) represents the trace of the matrix; S322: Using the exponentially weighted Riemann distance RD in the vertical direction v,α and the exponentially weighted Riemann distance RD in the horizontal direction h,α Define the vertical gradient G of the image RDv,α and the horizontal gradient G of the image RDh,α : G RDv,α =log(RD v,α ),G RDh,α =log(RD h,α ); S323: Using the vertical gradient G RDv,α and the horizontal gradient G RDh,α Define the PolSAR-Harris matrix C PSH , and using the PolSAR-Harris matrix C PSH Get the PolSAR-Harris response function image R PSH : R PSH (x,y,α)=det(C PSH (x,y,α))-d·tr(C PSH (x,y,α)) 2 Among them, det(·) means to find the determinant of the matrix, d is the empirical parameter, A standard deviation of Gaussian convolution kernel.

2. The PolSAR image ship detection method based on PolSAR-SIFT key points according to claim 1, characterized in that: The S2 includes: S21: Using three PolSAR images I HH ,I HV and I VV The elements with the same coordinates in construct a set of polarization scattering vectors k(x,y)=[I HH (x,y),I HV (x,y),I VV (x,y)] T , where I HH (x,y),I HV (x,y) and I VV (x,y) corresponds to image I HH ,I HV and I VV The complex scattering coefficient at the (x,y) coordinate; S22: construct a polarization covariance matrix C corresponding to the pixel point (x, y) according to the polarization scattering vector k(x, y): Among them, <·> represents the ensemble average, k1=I HH (x,y), k2=I HV (x,y), k3=I VV (x,y).

3. The PolSAR image ship detection method based on PolSAR-SIFT key points according to claim 1, characterized in that: The S33 includes: Select the PolSAR-Harris response function image R PSH All local maximum points in the are taken as pre-selected key points, and then the pre-selected key points are screened according to a preset key point threshold, and points greater than the key point threshold are retained as final key points.

4. The PolSAR image ship detection method based on PolSAR-SIFT key points according to claim 3 is characterized in that: The S4 includes: S41: According to the key point coordinate k i Get the pre-selected target area p i , the pre-selected target area is a square area centered on the key point; S42: Calculate the current pre-selected area p i The covariance matrix C of each pixel in (m,n) With the current pre-selected target area p i The covariance matrix C of the central pixel center The similarity measure s KL (C center ,C (m,n) ): Where h is the adjustment parameter, (m,n) is the coordinate of the pixel in the current pre-selected area; S43: The similarity measurement between each pixel and the central pixel is calculated according to the corresponding pixel position to form a similarity measurement image I s (p i ), and according to the similarity measurement image I s (p i ) Construct the change indication parameter P of the current pre-selected area v (p i ): Among them, std[·] represents the standard deviation of the elements in the calculated image, and mean[·] represents the mean of the elements in the calculated image; S44: All pre-selected regions pi are indicated by change parameters P v (p i ) is compared with the set change indication parameter threshold, and the change indication parameter P is selected v (p i ) greater than the change indication parameter threshold constitutes the final target area set S fa {p fj ,j=1,...,M}.

5. The PolSAR image ship detection method based on PolSAR-SIFT key points according to claim 4, characterized in that: The S5 includes: S51: Calculate the target area p fj SPAN image: SPAN(p fj )={|I HH (c,r)| 2 +2|I HV (c,r)| 2 +|I VV (c,r)| 2 ,c∈[x j -L,x j +L],r∈[y j -L,y j +L]} Among them, SPAN(p fj ) represents the target area p fj The polarization total power image composed of the corresponding pixel positions of the polarization total power calculated at the mid-coordinate point (c, r); S52: Compare the pixel value in the polarization total power image with a set polarization total power threshold, and the position where the pixel value is greater than the polarization total power threshold is the position of the ship.

6. A storage medium, characterized in that: The storage medium stores a computer program, and the computer program is used to execute the steps of the PolSAR image ship detection method according to any one of claims 1 to 5.

7. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program, and when the processor calls the computer program in the memory, the steps of the PolSAR image ship detection method according to any one of claims 1 to 5 are implemented.

Images