A ship target detection method and device based on non-local information

Abstract

The application discloses a ship target detection method and device based on non-local information, and relates to the technical field of ship target detection. The method comprises the following steps: determining a to-be-detected area and a non-local area in which a target pixel is located according to a superpixel segmentation technology; performing similarity calculation on the target pixel in the to-be-detected area and pixels in the non-local area one by one, and determining a non-local area pixel corresponding to a minimum similarity value; determining a non-local information polarization covariance matrix through an inner product formula; determining a polarization whitening filter value of the target pixel and a target detection threshold value, and determining a ship pixel according to the polarization whitening filter value of the target pixel and the target detection threshold value. The method considers the spatial information around the pixel, can significantly overcome the interference of sea clutter under complex sea conditions, can better improve the signal-to-clutter ratio of the ship target, and ensures that the ship target can be accurately detected.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ship target detection, in particular to a ship target detection method and device based on non-local information. BACKGROUND

[0002] At present, the detection of ship targets is mainly completed by establishing the difference between the polarization backscattering characteristics of targets and sea clutter, and constructing ship target detection features. Traditional ship target detection algorithms usually only use the polarization information of a single pixel, that is, directly use the polarization covariance matrix or polarization coherence matrix for ship target detection, without considering the spatial information around the target pixel. Especially when the background is complex or high sea state occurs, they will easily miss the detection of ships, resulting in a high false alarm rate. In fact, the existence of a target pixel is often not isolated, and the context information around it can also help ship detection. Therefore, how to effectively use the spatial information of the pixel to improve the accuracy of ship target detection in complex sea conditions is a problem to be solved. SUMMARY

[0003] The present application aims to: in view of the problem that the existing SAR ship target detection method cannot accurately detect ship targets in complex sea conditions, provide a ship target detection method based on non-local information.

[0004] In a first aspect, the present application provides a ship target detection method based on non-local information, which comprises: determining the target pixel region to be detected according to the superpixel segmentation technology, and removing the target pixel region to be detected from the region where the target is located to determine the non-local region of the target pixel region to be detected; calculating the similarity of each pixel in the non-local region with the target pixel in the target pixel region to be detected, arranging the obtained similarity values from small to large, and determining the non-local region pixel corresponding to the minimum similarity value; determining the non-local information polarization covariance matrix according to the inner product formula of the feature vector of the non-local region pixel and the feature vector of the target pixel; determining the polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix, determining the target detection threshold according to the preset false alarm rate, comparing the polarization whitening filter value of the target pixel with the target detection threshold, and if the polarization whitening filter value of the target pixel is greater than the target detection threshold, the target pixel is a ship pixel.

[0005] In any of the above technical solutions, further, the determination of the non-local information polarization covariance matrix according to the inner product formula of the feature vector of the non-local region pixel and the feature vector of the target pixel comprises:

[0006] According to the polarization feature description vector, the feature vector of the target pixel is combined with the feature vector of the non-local region pixel to construct a non-local information joint use vector, and the non-local information polarization covariance matrix is determined through the inner product formula according to the constructed non-local information joint use vector.

[0007] In any of the technical solutions above, further, the combination of the feature vector of the target pixel with the feature vector of the non-local region pixel to construct a non-local information joint use vector comprises:

[0008] The feature vector of the target pixel is added to the feature vector of the non-local region pixel to obtain a non-local information joint use vector.

[0009] In any of the technical solutions above, further, before the determination of the polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix, the method further comprises:

[0010] A sea clutter region is determined, and an average non-local information polarization covariance matrix of the sea clutter region is calculated.

[0011] In any of the technical solutions above, further, the determination of the polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix comprises:

[0012] The polarization whitening filter value of the target pixel is determined according to the average non-local information polarization covariance matrix of the sea clutter region and the non-local information polarization covariance matrix.

[0013] In any of the technical solutions above, further, the method further comprises that the determination of the target detection threshold according to the preset false alarm rate comprises:

[0014] For pixels in a 100*100 sea clutter region, a sea clutter polarization whitening filter value corresponding to each pixel is calculated;

[0015] A statistical distribution model is established for a plurality of sea clutter polarization whitening filter values by using a mixture Gaussian distribution, wherein a mixture Gaussian distribution model f(x) corresponding to a pixel x in the sea clutter region is as follows:

[0016]

[0017]

[0018] In the formula, K is the number of models in the mixture Gaussian distribution, w k, u k ,σ k are the weight, mean value and standard deviation of the kth Gaussian distribution, respectively, m k(x) represents the kth Gaussian distribution of the pixels in the sea clutter region, k = 1, 2, 3…K, and there are K pixels in the sea clutter region in total;

[0019] According to the preset false alarm rate P fa And the mixed Gaussian distribution model f(x) corresponding to the pixel x in the sea clutter region determines the target detection threshold T, and the calculation formula is as follows:

[0020]

[0021] In any of the above technical solutions, further, the method further comprises:

[0022] The polarized whitening filter value of the target pixel is compared with the target detection threshold, and if the polarized whitening filter value of the target pixel is less than or equal to the target detection threshold, the target pixel is a clutter pixel.

[0023] In a second aspect, the application also provides a ship target detection device based on non-local information, comprising: a first processing module for determining a to-be-detected region where a target pixel is located according to a superpixel segmentation technology, and determining a non-local region of the to-be-detected region by excluding the to-be-detected region from a region where a target is located; a second processing module for performing similarity calculation on the pixels in the non-local region and the target pixel in the to-be-detected region one by one, arranging the obtained similarity values from small to large, and determining a non-local region pixel corresponding to a minimum similarity value; a third processing module for determining a non-local information polarization covariance matrix according to a feature vector of the non-local region pixel and a feature vector of the target pixel through an inner product formula; a fourth processing module for determining a polarized whitening filter value of the target pixel according to the non-local information polarization covariance matrix, determining a target detection threshold according to a preset false alarm rate, comparing the polarized whitening filter value of the target pixel with the target detection threshold, and if the polarized whitening filter value of the target pixel is greater than the target detection threshold, the target pixel is a ship pixel.

[0024] In a third aspect, the application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the program to realize the steps of the ship target detection method based on non-local information according to any of the above.

[0025] In a fourth aspect, the application also provides a non-transitory computer readable storage medium having a computer program stored thereon, and the computer program is executable by a processor to realize the steps of the ship target detection method based on non-local information according to any of the above.

[0026] In a fifth aspect, the present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the steps of any of the above-mentioned ship target detection methods based on non-local information.

[0027] The present application has the following advantages: the technical solution in the present application combines the pixel information to be detected with the non-neighbor pixel information around the pixel to be detected to construct a non-local information polarization covariance matrix, considers the spatial information around the pixel, and more thoroughly applies the polarization scattering information around the target pixel. This application manner can not only significantly overcome the interference of sea clutter in complex sea conditions, but also better improve the signal-to-clutter ratio of the ship target, and ensure that the ship target can be accurately detected. BRIEF DESCRIPTION OF DRAWINGS

[0028] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the following drawings of which:

[0029] Figure 1 is a schematic flowchart of a ship target detection method based on non-local information according to an embodiment of the present application;

[0030] Figure 2 is a schematic block diagram of the range of each window according to an embodiment of the present application;

[0031] Figure 3 is a schematic diagram of the actual matrix calculation process according to an embodiment of the present application;

[0032] Figure 4 is a schematic block diagram of an application scenario according to an embodiment of the present application;

[0033] Figure 5 is a schematic diagram of the detection result of the ship target according to an embodiment of the present application;

[0034] Figure 6 is a structural schematic diagram of some embodiments of a ship target detection device based on non-local information provided by the present application;

[0035] Figure 7 An example of an electronic device is shown in the physical structure schematic diagram. DETAILED DESCRIPTION

[0036] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the present application will be further described in detail below in conjunction with the drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0037] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be practiced without the specific details. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so as not to obscure the present application.

[0038] As shown in Figure 1 The embodiment provides a ship target detection method based on non-local information, and the method comprises the following steps.

[0039] Step 1: determining a to-be-detected region where a target pixel is located according to a superpixel segmentation technology, and removing the to-be-detected region from a region where the target is located to determine a non-local region of the to-be-detected region.

[0040] In some embodiments, the superpixel segmentation technology can be used to obtain the region range where the target pixel is located, and then a 3*3 range region (the 3*3 range region is the region range where the target pixel is located) centered on the target pixel is removed from the region range where the target is located to obtain the non-local region corresponding to the target pixel. The superpixel segmentation technology refers to a process of subdividing a digital image into a plurality of image sub-regions (a collection of pixels) (also known as superpixels), such as a linear iterative clustering method.

[0041] In some embodiments, the to-be-detected region can also be determined according to a preset to-be-detected window, and the non-local region of the to-be-detected region is determined according to a preset non-local window and a preset protection window. The size of the to-be-detected window, the non-local window and the protection window can be set to obtain the range of non-local information required for the to-be-detected target.

[0042] In some embodiments, the to-be-detected target can be a ship in a SAR image. Before determining the non-local region of the to-be-detected target, the type, size and other information of the image can be obtained and judged.

[0043] In some embodiments, the pixels in the to-be-detected region can have one or more.

[0044] Step 2: calculating the similarity of each pixel in the non-local region with the target pixel in the to-be-detected region one by one, arranging the obtained similarity values from small to large, and determining the non-local region pixel corresponding to the minimum similarity value.

[0045] As an example, the superpixel segmentation technology can be used to obtain the region range where the target pixel is located, and then a 3*3 range region centered on c is removed from the region range where the target pixel is located to obtain the non-local region pixel set M of the target pixel, and the similarity of any pixel M(i) in the set with c is calculated.

[0046] As an example, the to-be-detected frame size is 1X1, and the target pixel in the to-be-detected region is one, denoted as target pixel c. By setting three different window sizes of the to-be-detected window, the non-local window, and the protection window, a pixel set M in the non-local region is determined, and the similarity between any pixel M(i) in the set and the target pixel c is calculated.

[0047] The calculated multiple similarity values are arranged from small to large, and the pixel point i corresponding to the minimum similarity value is taken out. The pixel point i is included in the pixel set M in the non-local region.

[0048] The similarity value calculation formula is:

[0049]

[0050] wherein v i and v c are feature vectors of pixel points i and c (pixel c is a target pixel), and the expression of v is as follows:

[0051]

[0052] In the formula, Cxy (x = 1, 2, 3; y = 1, 2, 3) is the corresponding position element of the covariance matrix [C], and the definition of [C] is as follows:

[0053]

[0054] In the formula, k is a straight sequence base, that is:

[0055]

[0056] Step 3: According to the feature vector of the non-local region pixel and the feature vector of the target pixel, the non-local information polarization covariance matrix is determined by the inner product formula.

[0057] The polarization covariance matrix is also called a complex Hermitian matrix, and like the co-polarization scattering matrix, it also contains all the target polarization information measured by the radar. The polarization SAR image processing process is generally based on the polarization covariance matrix and the polarization coherence matrix, and it is the basis for analyzing and processing multi-polarization SAR data.

[0058] In some embodiments, according to the polarization feature description vector, the feature vector of the target pixel and the feature vector of the non-local region pixel can be combined to construct a non-local information joint use vector, and according to the constructed non-local information joint use vector, the non-local information polarization covariance matrix is determined by the inner product formula.

[0059] In some embodiments, the feature vector of the target pixel can be added to the feature vectors of the non-local region pixels to obtain a non-local information joint use vector.

[0060] As an example: after obtaining a series of different similarity values, arrange them from small to large, take out the pixel point i corresponding to the minimum similarity value, and then combine the feature vector v i with the feature vector v c of the pixel c.

[0061] v new = v i + v c Equation (5)

[0062] Take the inner product operation on v new to obtain the non-local information polarization covariance matrix [C] NL , that is

[0063]

[0064] where [C] NL represents the non-local information polarization covariance matrix.

[0065] In one embodiment, as shown in Figure 2 , from top to bottom, the size of the detection window (window size is 1x1), the non-local window (window size is 9x9), and the protection window (window size is 3x3) can be set from left to right. As shown in Figure 3 , the detection window (window size is 1x1), the non-local window (window size is 9x9), and the protection window (window size is 3x3) slide from left to right and calculate the non-local information polarization covariance matrix [C] NLc from top to bottom.

[0066] Step 4, determining the polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix, determining the target detection threshold according to the preset false alarm rate, comparing the polarization whitening filter value of the target pixel with the target detection threshold, and if the polarization whitening filter value of the target pixel is greater than the target detection threshold, the target pixel is a ship pixel.

[0067] In some embodiments, before determining the polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix, it further includes determining a sea clutter region and calculating the average non-local information polarization covariance matrix of the sea clutter region.

[0068] The embodiment can overcome the bottleneck that the conventional polarimetric SAR ship target detection algorithm is prone to miss detection of ships in complex sea conditions. The ship target detection algorithm generated by the embodiment combines the spatial information of the surrounding pixels, uses the correlation polarization characteristics of the polarimetric covariance matrix to construct a ship target detection algorithm in complex sea conditions, and the core is to combine the information of the to-be-detected pixel and the surrounding non-neighbor pixel information to construct a non-local information polarimetric covariance matrix, so that the non-local information is applied in the traditional polarimetric covariance matrix. The ship detector constructed based on the statistical characteristics of the new matrix has strong robustness and improves the ship target detection accuracy.

[0069] In some embodiments, the polarimetric whitening filter value of the target pixel can be determined according to the average non-local information polarimetric covariance matrix of the sea clutter region and the non-local information polarimetric covariance matrix.

[0070] In some embodiments, the target detection threshold can be determined according to a preset false alarm rate, including:

[0071] For the pixels in the 100x100 sea clutter region, the sea clutter polarimetric whitening filter value corresponding to each pixel is calculated;

[0072] The statistical distribution of the plurality of sea clutter polarimetric whitening filter values is modeled using a mixture Gaussian distribution, wherein the mixture Gaussian distribution model f(x) corresponding to the pixel x in the sea clutter region is as follows:

[0073]

[0074]

[0075] In the formula, K is the number of models in the mixture Gaussian distribution, w k ,u k ,σ k are the weight, mean value, and standard deviation of the kth Gaussian distribution, respectively, m k (x) represents the kth Gaussian distribution of the pixels in the sea clutter region, k = 1, 2, 3…K, and there are K pixels in the sea clutter region;

[0076] The target detection threshold T is determined according to a preset false alarm rate P fa and the mixture Gaussian distribution model f(x) corresponding to the pixel x in the sea clutter region, and the calculation formula is as follows:

[0077]

[0078] In some embodiments, the polarimetric whitening filter value of the target pixel is compared with the target detection threshold, and if the polarimetric whitening filter value of the target pixel is less than or equal to the target detection threshold, the target pixel is a clutter pixel.

[0079] As an example, the non-local information polarimetric covariance matrix [C] NL After that, the non-target sea clutter region Z is further selected, and the non-local information polarimetric covariance matrix of all pixels is accumulated, and then the average non-local information polarimetric covariance matrix ∑ is calculated. For a specific pixel c, the corresponding polarimetric whitening filter value is defined as follows:

[0080] PWF = Tr(∑ -1 [C] NLc ) Formula (7)

[0081] In the formula, Tr(·) represents the trace of the matrix, [C] NLc is the non-local information polarimetric covariance matrix of the pixel c.

[0082] After obtaining the PWF value of the to-be-detected target pixel, a detection threshold can be set. If the PWF value is greater than the detection threshold, it is considered that the to-be-detected target pixel is a ship pixel, otherwise it is considered that the pixel is a clutter pixel.

[0083] In some embodiments, the resulting matrix polarimetric scattering characteristics can also be determined, and the polarimetric whitening filter is constructed by using these properties; then by comparing PWFNL with the target detection threshold, whether the to-be-detected target pixel is a real ship target pixel can be judged.

[0084] Figure 4 The intensity values of each pixel calculated by the matrix [C]NLare shown, wherein the larger the value is, the closer to white it is. Obviously, the matrix [C] NLc has a good ability to reflect the polarimetric information of the ship target. Combined with this characteristic, the polarimetric whitening filter value is calculated for ship target detection, and the result is shown in FIG. 6. Figure 5 By comparing with the real scene in FIG. 5, it can be found that Figure 3 all the to-be-detected ship targets in FIG. 6 are accurately detected, which proves the effectiveness of the ship target detection in the complex background in the embodiment. Figure 5

[0085] The steps in the present application can be adjusted, combined and deleted in sequence according to actual needs.

[0086] ​The technical solution in the application determines the to-be-detected region where the target pixel is located according to the superpixel segmentation technology, removes the to-be-detected region from the region where the target is located to determine the non-local region of the to-be-detected region; the pixels in the non-local region are calculated with the target pixel c in the to-be-detected region one by one; the obtained similarity values are arranged from small to large, and the pixel i corresponding to the minimum value is taken; according to the polarized feature description vector, the feature vector of the i element is combined with the feature vector of the c element to construct a non-local information joint use vector; according to the inner product formula, a non-local information guided polarization covariance matrix is proposed; the clutter distribution characteristics of the new matrix are analyzed, and a polarization whitening filter PWFNC is established; according to the preset false alarm rate, a target detection threshold is determined, the polarization whitening filter value of the target pixel is compared with the target detection threshold, the pixel greater than the target detection threshold is identified as a target pixel, and vice versa, a clutter pixel is identified, and the detection of the target is completed. The application realizes the combination of to-be-detected pixel information and its surrounding non-neighbor pixel information in the ship target detection process, the construction of a non-local information polarization covariance matrix, the consideration of the spatial information of the surrounding pixels, and the more thorough application of the polarization scattering information around the target pixel. This application method can not only overcome the interference of sea clutter in complex sea conditions, but also better improve the signal-to-clutter ratio of the ship target, and ensure that the ship target can be accurately detected. In addition, knowing the specific position of the ship target will undoubtedly help people better explore marine resources, protect the marine environment, and manage maritime traffic.

[0087] Please refer to Figure 6 , Figure 6 Some embodiments of the ship target detection device based on non-local information provided by the application are structural schematic diagrams, as an implementation of the method shown in the above figures, the application also provides some embodiments of a ship target detection device based on non-local information. These device embodiments correspond to some method embodiments shown in Figure 1 The device can be applied to various electronic devices.

[0088] As Figure 6As shown, the non-local information based ship target detection device of some embodiments includes a first processing module 601, a second processing module 602, a third processing module 603, and a fourth processing module 604. The first processing module is configured to determine a to-be-detected region in which a target pixel is located according to a superpixel segmentation technique, remove the to-be-detected region from a region in which the target is located, and determine a non-local region of the to-be-detected region. The second processing module is configured to calculate the similarity between each pixel in the non-local region and the target pixel in the to-be-detected region, arrange the obtained similarity values in ascending order, and determine a non-local region pixel corresponding to a minimum similarity value. The third processing module is configured to determine a non-local information polarization covariance matrix according to a feature vector of the non-local region pixel and a feature vector of the target pixel by using an inner product formula. The fourth processing module is configured to determine a polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix, determine a target detection threshold value according to a preset false alarm rate, compare the polarization whitening filter value of the target pixel with the target detection threshold value, and determine that the target pixel is a ship pixel if the polarization whitening filter value of the target pixel is greater than the target detection threshold value.

[0089] In an optional implementation of some embodiments, the third processing module is further configured to combine the feature vector of the target pixel and the feature vector of the non-local region pixel according to a polarization feature description vector to construct a non-local information joint use vector, and determine the non-local information polarization covariance matrix according to the constructed non-local information joint use vector by using the inner product formula.

[0090] In an optional implementation of some embodiments, the third processing module is further configured to add the feature vector of the target pixel and the feature vector of the non-local region pixel to obtain the non-local information joint use vector.

[0091] In an optional implementation of some embodiments, the device further includes a fifth processing module configured to determine a sea clutter region and calculate an average non-local information polarization covariance matrix of the sea clutter region.

[0092] In an optional implementation of some embodiments, the fourth processing module is further configured to determine the polarization whitening filter value of the target pixel according to the average non-local information polarization covariance matrix of the sea clutter region and the non-local information polarization covariance matrix.

[0093] In an optional implementation of some embodiments, the determination of the target detection threshold value according to the preset false alarm rate includes:

[0094] For pixels in a 100x100 sea clutter region, a sea clutter polarization whitening filter value corresponding to each pixel is calculated.

[0095] The multiple sea clutter polarization whitening filter values are modeled by using a Gaussian mixture distribution, wherein a pixel x in the sea clutter region corresponds to a Gaussian mixture distribution model f(x) as follows:

[0096]

[0097]

[0098] wherein K is the number of models in the Gaussian mixture distribution, w k ,u k ,σ k are the weight, mean value, and standard deviation of the kth Gaussian distribution, respectively, m k (x) represents the kth Gaussian distribution of the pixel in the sea clutter region, k = 1, 2, 3…K, and there are K pixels in the sea clutter region;

[0099] According to a preset false alarm rate P fa and the Gaussian mixture distribution model f(x) corresponding to the pixel x in the sea clutter region, a target detection threshold T is determined, and the calculation formula is as follows:

[0100]

[0101] In some optional implementations of the embodiments, the apparatus further comprises a sixth processing module configured to compare the polarization whitening filter value of the target pixel with the target detection threshold, and if the polarization whitening filter value of the target pixel is less than or equal to the target detection threshold, the target pixel is a clutter pixel.

[0102] It can be understood that each module described in the apparatus corresponds to each step in the method described with reference to Figure 1 . Therefore, the operations, features, and advantages described above for the method also apply to the apparatus and the modules contained therein, and will not be repeated here.

[0103] Figure 7 An example of an electronic device is shown in FIG. 1, which is a schematic diagram of the physical structure of an electronic device, such as a mobile phone, a tablet computer, a wearable device, or the like. Figure 7As shown, the electronic device can include a processor 710, a communications interface 720, a memory 730, and a communications bus 740, wherein the processor 710, the communications interface 720, and the memory 730 complete mutual communication through the communications bus 740. The processor 710 can invoke a logic instruction in the memory 730 to execute a ship target detection method based on non-local information, which includes: determining a to-be-detected region where a target pixel is located according to a superpixel segmentation technology, removing the to-be-detected region from a region where the target is located to determine a non-local region of the to-be-detected region; calculating the similarity of pixels in the non-local region with the target pixel in the to-be-detected region one by one, arranging the obtained similarity values from small to large, and determining a non-local region pixel corresponding to a minimum similarity value; determining a non-local information polarization covariance matrix through an inner product formula according to a feature vector of the non-local region pixel and a feature vector of the target pixel; determining a polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix, determining a target detection threshold value according to a preset false alarm rate, comparing the polarization whitening filter value of the target pixel with the target detection threshold value, and if the polarization whitening filter value of the target pixel is greater than the target detection threshold value, the target pixel is a ship pixel.

[0104] In addition, the logic instruction in the memory 730 described above can be implemented in the form of a software function unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the above-mentioned method of various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various program code storage media.

[0105] In another aspect, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer-readable storage medium, and the computer program comprises program instructions, and when the program instructions are executed by a computer, the computer can execute the ship target detection method based on non-local information provided by the above method, and the method comprises the following steps: determining a to-be-detected region in which a target pixel is located according to a superpixel segmentation technology, determining a non-local region of the to-be-detected region by excluding the to-be-detected region from a region in which the target is located; performing similarity calculation on each pixel in the non-local region and the target pixel in the to-be-detected region one by one, arranging the obtained similarity values in ascending order, and determining a non-local region pixel corresponding to a minimum similarity value; determining a non-local information polarization covariance matrix by an inner product formula according to a feature vector of the non-local region pixel and a feature vector of the target pixel; determining a polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix, determining a target detection threshold value according to a preset false alarm rate, comparing the polarization whitening filter value of the target pixel with the target detection threshold value, and regarding the target pixel as a ship pixel if the polarization whitening filter value of the target pixel is greater than the target detection threshold value.

[0106] In still another aspect, the present application also provides a non-transitory computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the ship target detection method based on non-local information provided by the above method, and the method comprises the following steps: determining a to-be-detected region in which a target pixel is located according to a superpixel segmentation technology, determining a non-local region of the to-be-detected region by excluding the to-be-detected region from a region in which the target is located; performing similarity calculation on each pixel in the non-local region and the target pixel in the to-be-detected region one by one, arranging the obtained similarity values in ascending order, and determining a non-local region pixel corresponding to a minimum similarity value; determining a non-local information polarization covariance matrix by an inner product formula according to a feature vector of the non-local region pixel and a feature vector of the target pixel; determining a polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix, determining a target detection threshold value according to a preset false alarm rate, comparing the polarization whitening filter value of the target pixel with the target detection threshold value, and regarding the target pixel as a ship pixel if the polarization whitening filter value of the target pixel is greater than the target detection threshold value.

[0107] The device embodiments described above are only schematic, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment scheme. Those skilled in the art can understand and implement without creative labor.

[0108] Those skilled in the art can clearly understand the implementation of the embodiments by the description of the above embodiments, and the embodiments can be implemented by means of software and necessary universal hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, and the computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the above-mentioned method of each embodiment or some parts of the embodiment.

[0109] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for some technical features thereof; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A ship target detection method based on non-local information, characterized in that, The method comprises: According to the superpixel segmentation technology, a to-be-detected region where the target pixel is located is determined, the to-be-detected region is removed from a region where the target is located, and a non-local region of the to-be-detected region is determined; The pixels in the non-local region are sequentially compared with the target pixel in the to-be-detected region for similarity calculation, the obtained similarity values are arranged in ascending order, and a non-local region pixel corresponding to a minimum similarity value is determined; According to the feature vector of the non-local region pixel and the feature vector of the target pixel, a non-local information polarization covariance matrix is determined through an inner product formula; According to the non-local information polarization covariance matrix, a polarization whitening filter value of the target pixel is determined, a target detection threshold is determined according to a preset false alarm rate, the polarization whitening filter value of the target pixel is compared with the target detection threshold, and if the polarization whitening filter value of the target pixel is greater than the target detection threshold, the target pixel is a ship pixel. 2.The non-local information based SAR image ship target detection method of claim 1, wherein, According to the feature vector of the non-local region pixel and the feature vector of the target pixel, a non-local information polarization covariance matrix is determined through an inner product formula, comprising: According to the polarization feature description vector, the feature vector of the target pixel and the feature vector of the non-local region pixel are combined to construct a non-local information joint use vector, and the non-local information polarization covariance matrix is determined through the inner product formula according to the constructed non-local information joint use vector.

3. The non-local information based ship target detection method of claim 2, wherein, The feature vector of the target pixel and the feature vector of the non-local region pixel are combined to construct a non-local information joint use vector, comprising: The feature vector of the target pixel and the feature vector of the non-local region pixel are added to obtain the non-local information joint use vector.

4. The non-local information based ship target detection method of claim 1, wherein, Before the polarization whitening filter value of the target pixel is determined according to the non-local information polarization covariance matrix, the method further comprises: A sea clutter region is determined, and an average non-local information polarization covariance matrix of the sea clutter region is calculated.

5. The non-local information based ship target detection method of claim 4, wherein, The polarization whitening filter value of the target pixel is determined according to the average non-local information polarization covariance matrix of the sea clutter region and the non-local information polarization covariance matrix. The target detection threshold is determined according to the preset false alarm rate, comprising:

6. The non-local information based ship target detection method of claim 4, wherein, For the pixels in a 100*100 sea clutter region, a sea clutter polarization whitening filter value corresponding to each pixel is calculated; A plurality of sea clutter polarization whitening filter values are statistically modeled by using a mixed Gaussian distribution, wherein a mixed Gaussian distribution model f(x) corresponding to a pixel x in the sea clutter region is as follows: The method further comprises: where K is the number of models in the Gaussian mixture distribution, w k, u k ,σ k are the weight, mean, and standard deviation of the kth Gaussian distribution, respectively, m k (x) represents the kth Gaussian distribution of the pixel in the sea clutter region, k = 1, 2, 3…K, and there are K pixels in the sea clutter region. According to the preset false alarm rate P fa A target detection threshold T is determined according to a mixed Gaussian distribution model f(x) corresponding to the pixel x in the sea clutter region, and a calculation formula of the target detection threshold T is as follows:

7. The non-local information based ship target detection method of claim 1, wherein, The polarization whitening filter value of the target pixel is compared with the target detection threshold, and if the polarization whitening filter value of the target pixel is less than or equal to the target detection threshold, the target pixel is a clutter pixel. Comprise:

8. A non-local information based ship target detection device, characterized by, The first processing module is configured to determine a to-be-detected region where a target pixel is located according to a superpixel segmentation technology, remove the to-be-detected region from a region where the target is located, and determine a non-local region of the to-be-detected region; ​ a second processing module configured to perform similarity calculation on each pixel in the non-local region with a target pixel in the detection region, arrange the obtained similarity values in ascending order, and determine a non-local region pixel corresponding to a minimum similarity value; a third processing module configured to determine a non-local information polarization covariance matrix according to a feature vector of the non-local region pixel and a feature vector of the target pixel through an inner product formula; a fourth processing module configured to determine a polarization whitening filter value of the target pixel according to the non-local information polarization covariance matrix, determine a target detection threshold value according to a preset false alarm rate, compare the polarization whitening filter value of the target pixel with the target detection threshold value, and determine that the target pixel is a ship pixel if the polarization whitening filter value of the target pixel is greater than the target detection threshold value.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps of the ship target detection method based on non-local information according to any one of claims 1 to 7. 10.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the ship target detection method based on non-local information according to any one of claims 1 to 7.

Images