A Wide-Swath Polarimetric SAR Image Ship Target Detection Method, System, Device, Medium and Product

Abstract

The present application discloses a method, system, device, medium and product for ship target detection in wide-swath polarimetric SAR images, which relates to the field of image target detection. The method includes obtaining wide-swath SAR data; preprocessing the wide-swath SAR data to obtain preprocessed data; extracting scattering features from the preprocessed data to obtain multiple SAR images with different scattering features; selecting the SAR image with one of the scattering features for adaptive cropping, and obtaining the position information and cropping grid of each cropped image; using the cropping grid to crop the SAR images with other scattering features to obtain cropped images; performing ship target detection based on the multiple cropped images to obtain ship target detection information; and mapping the ship target detection information to the wide-swath SAR image based on the position information of each cropped image. The present application can adaptively crop images without damaging ship targets, and improve the real-time performance and accuracy of ship target detection in SAR images.

Description

Technical Field

[0001] The present application relates to the field of image target detection, and particularly to a method, system, device, medium and product for ship target detection in wide-swath polarimetric SAR images. Background Art

[0002] Synthetic Aperture Radar (SAR) is an active microwave imaging sensor with the ability to work all day and all weather, and has good adaptability for monitoring the ocean with changeable climate. Full polarimetric SAR can simultaneously receive and transmit electromagnetic waves in multiple polarization states, so that the measured data is no longer just the radar cross section (RCS) of the radar target, but a 2×2 scattering matrix, and the obtained target scattering information is richer. How to effectively use different scattering information for real-time ship target detection is one of the current research hotspots of ship targets. In addition, since the pixels of wide-swath SAR images are too large to input the entire image into the network for detection, how to adaptively crop the image without damaging the ship target is also the difficulty and focus of wide-swath polarimetric SAR image target detection. Summary of the Invention

[0003] The purpose of the present application is to provide a method, system, device, medium and product for ship target detection in wide-swath polarimetric SAR images, which can adaptively crop the image without damaging the ship target and improve the real-time performance and accuracy of ship target detection in SAR images.

[0004] To achieve the above purpose, the present application provides the following solutions:

[0005] In the first aspect, the present application provides a method for ship target detection in wide-swath polarimetric SAR images, including:

[0006] Obtain wide-swath SAR data;

[0007] Preprocess the wide-swath SAR data to obtain preprocessed data;

[0008] Extract scattering features from the preprocessed data to obtain multiple SAR images with different scattering features;

[0009] Select one of the SAR images with a scattering feature for adaptive cropping, and obtain the position information of each cropped picture and the SAR image cropping grid;

[0010] Crop the SAR images with other scattering features based on the cropping grid to obtain multiple cropped pictures with different scattering features of the same-position SAR data;

[0011] Perform ship target detection based on multiple of the cropped images to obtain ship target detection information; the ship target detection information includes ship target categories, ship target bounding boxes, and the position information of the ship targets;

[0012] Map the ship target detection information onto the wide - swath SAR image based on the position information of each cropped image.

[0013] Optionally, select the SAR image with one type of scattering feature for adaptive cropping, and obtain the position information of each cropped image and the SAR image cropping grid, specifically including:

[0014] Use a clustering algorithm to perform clustering on the SAR image with one type of scattering feature to obtain a clustered image; the clustered image includes multiple clustering targets;

[0015] Determine the bounding box position information of each clustering target in the clustered image;

[0016] Generate a cropping grid according to the original size of the wide - swath SAR image and the set cropping size;

[0017] Take a boundary point in the clustered image as the origin, set the boundary line of the cropping grid with the origin as the vertex to 0, and use the lines parallel and adjacent to the boundary line in the cropping grid as cropping lines;

[0018] Use the cropping grid to crop the clustered image, and during the cropping process, determine whether there are clustering targets on the cropping lines;

[0019] If there are clustering targets on the cropping line, traverse all the clustering targets on the cropping line, and determine the position information of the clustering target closest to the boundary line of the cropping grid among these clustering targets;

[0020] Determine the offset based on the position information of the clustering target closest to the boundary line of the cropping grid;

[0021] After offsetting the cropping grid along the cropping line by the offset amount towards the boundary line of the cropping grid, update the cropping grid and complete the cropping;

[0022] If there are no clustering targets on the cropping line, directly complete the cropping using the cropping grid.

[0023] Optionally, perform ship target detection based on multiple of the cropped images to obtain ship target detection information, specifically including:

[0024] Construct a ship target detection network; the ship target detection network includes: a multi - channel convolutional attention network module, a multi - scale convolutional attention network module, and a hybrid encoder module;

[0025] Input multiple pieces of the cropped images and the position information of each cropped image into the ship target detection network to obtain initial ship target detection information; the initial ship target detection information includes multiple prediction box encoding information; the prediction box encoding information includes: predicted target category, predicted bounding box, and predicted target position information;

[0026] Adopt the uncertainty minimization query and selection method to select the first j pieces of prediction box encoding information from the prediction box encoding information of the initial ship target detection information;

[0027] Decode the first j pieces of prediction box encoding information to obtain the ship target detection information.

[0028] Optionally, inputting multiple pieces of the cropped images and the position information of each cropped image into the ship target detection network to obtain initial ship target detection information specifically includes:

[0029] Input multiple pieces of the cropped images and the position information of each cropped image into the multi-channel convolutional attention network module to obtain a first feature;

[0030] Input the first feature into the multi-scale convolutional attention network module to obtain a second feature; the second feature includes a first sub-feature, a second sub-feature, and a third sub-feature;

[0031] Input the second feature into the hybrid encoder module to obtain the initial ship target detection information.

[0032] Optionally, inputting the second feature into the hybrid encoder module to obtain the initial ship target detection information specifically includes:

[0033] In the hybrid encoder module, flatten the third sub-feature into a one-dimensional feature vector;

[0034] Perform multi-head attention mechanism encoding on the one-dimensional feature vector to achieve intra-scale feature interaction;

[0035] Convert the multi-head attention mechanism encoded one-dimensional feature vector into a two-dimensional vector;

[0036] Perform bidirectional feature fusion on the first sub-feature, the second sub-feature, and the two-dimensional vector and then splice and output to obtain the initial ship target detection information.

[0037] Optionally, set a multi-branch convolutional structure RepBlock module in the hybrid encoder module; use the multi-branch convolutional structure RepBlock module to perform bidirectional feature fusion on the first sub-feature, the second sub-feature, and the two-dimensional vector.

[0038] In a second aspect, the present application provides a wide-swath polarimetric SAR image ship target detection system, which is used to implement the wide-swath polarimetric SAR image ship target detection method provided above; the wide-swath polarimetric SAR image ship target detection system includes:

[0039] A polarimetric SAR data acquisition module, which is used to acquire wide-swath SAR data;

[0040] A wide-swath SAR image ship target adaptive cropping module, which is connected to the polarimetric SAR data acquisition module, and is used to preprocess the wide-swath SAR data to obtain preprocessed data, extract scattering features from the preprocessed data to obtain multiple SAR images with different scattering features, obtain cropped pictures, and acquire the position information of the cropped pictures;

[0041] An SAR ship target real-time detection module, which is connected to the wide-swath SAR image ship target adaptive cropping module, and is used to detect ship targets based on the cropped pictures to obtain ship target detection information, and is used to map the ship target detection information to the wide-swath SAR image based on the position information of each cropped picture; the ship target detection information includes the ship target category, the ship target bounding box, and the position information of the ship target.

[0042] In a third aspect, the present application provides a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the steps of a wide-swath polarimetric SAR image ship target detection method provided above.

[0043] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of a wide-swath polarimetric SAR image ship target detection method provided above are implemented.

[0044] In a fifth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of a wide-swath polarimetric SAR image ship target detection method provided above are implemented.

[0045] According to the specific embodiments provided by the present application, the following technical effects are disclosed in the present application:

[0046] The present application provides a method, system, device, medium and product for detecting ship targets in wide - swath polarimetric SAR images. By extracting scattering features from SAR data, SAR images with different scattering features can be obtained, and thus richer scattering information can be acquired. Adaptive cropping of SAR images with different scattering features can adaptively crop the images without damaging the ship targets. Based on multiple cropped images with different scattering features and the position information of each cropped image, ship target detection is performed to obtain ship target detection information, and the ship target detection information is mapped onto the wide - swath SAR image, which can improve the real - time performance and accuracy of ship target detection in SAR images. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1 It is an application environment diagram of a method for detecting ship targets in wide - swath polarimetric SAR images according to an embodiment of the present application;

[0049] Figure 2 It is a schematic flowchart of a method for detecting ship targets in wide - swath polarimetric SAR images provided by an embodiment of the present application;

[0050] Figure 3 It is a schematic flowchart of pre - processing and adaptive cropping provided by an embodiment of the present application;

[0051] Figure 4 It is a schematic flowchart of ship target detection provided by another embodiment of the present application;

[0052] Figure 5 It is a schematic diagram of the structure of a multi - channel convolutional attention network module provided by an embodiment of the present application;

[0053] Figure 6 It is a schematic diagram of the structure of a multi - scale attention network module provided by an embodiment of the present application;

[0054] Figure 7 It is a schematic diagram of the structure of a hybrid encoder module provided by an embodiment of the present application;

[0055] Figure 8 It is a schematic diagram of the module functions of a wide - swath polarimetric SAR image ship target detection system provided by an embodiment of the present application;

[0056] Figure 9 It is a schematic diagram of the structure of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0057] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.

[0058] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0059] The ship target detection method for wide-swath polarimetric SAR images provided by the embodiments of the present application can be applied to an application environment as Figure 1 shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send wide-swath SAR data to the server 104. After receiving the wide-swath SAR data, the server 104 preprocesses the wide-swath SAR data to obtain preprocessed data. Scattering feature extraction is performed on the preprocessed data to obtain multiple SAR images with different scattering features; one of the SAR images with a scattering feature is adaptively cropped, and the position information of each cropped picture and the SAR image cropping grid are obtained; based on the cropping grid, the SAR images with other scattering features are cropped to obtain multiple cropped pictures with different scattering features of the SAR data at the same position. Ship target detection is performed based on multiple cropped pictures with different scattering features to obtain ship target detection information. The ship target detection information is mapped to the wide-swath SAR image based on the position information of each cropped picture. The server 104 can feedback the wide-swath SAR image mapped with the ship target detection information to the terminal 102. In addition, in some embodiments, the ship target detection method for wide-swath polarimetric SAR images provided by the present application can also be implemented by the server 104 or the terminal 102 alone. For example, the terminal 102 can directly perform ship target detection on the wide-swath SAR image, or the server 104 can obtain the wide-swath SAR image from the data storage system and perform ship target detection on the wide-swath SAR image.

[0060] Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0061] In an exemplary embodiment, as Figure 2 shown, a method for detecting ship targets in wide-swath polarimetric SAR images is provided. This method is executed by a computer device, and specifically can be executed alone by a computer device such as a terminal or a server, or jointly executed by a terminal and a server. In the embodiment of the present application, taking this method applied to Figure 1 the server 104 in

[0062] as an example, the following steps 200 to step 206 are included.

[0063] Step 200: Obtain wide-swath SAR data.

[0064] Step 201: Preprocess the wide-swath SAR data to obtain preprocessed data.

[0065] Step 202: Extract scattering features from the preprocessed data to obtain multiple SAR images with different scattering features.

[0066] Step 203: Adaptively crop the SAR image with one type of scattering feature, and obtain the position information of each cropped image and the SAR image cropping grid;

[0067] Step 204: Crop the SAR images with other scattering features based on the cropping grid to obtain multiple cropped images with different scattering features of the SAR data at the same position.

[0068] Step 205: Detect ship targets based on multiple cropped images to obtain ship target detection information. The ship target detection information includes ship target categories, ship target bounding boxes, and the position information of ship targets.

[0069] Step 206: Map the ship target detection information to the wide-swath SAR image based on the position information of each cropped image.

[0070] Implementing the above steps 200 to step 206 can adaptively crop the image without damaging the ship target, and improve the real-time performance and accuracy of ship target detection in SAR images.

[0071] In another exemplary embodiment of the present application, since the picture data (i.e., wide-swath SAR image) generated from wide-swath SAR imaging data is usually very large and cannot be directly input into the detection network for object detection, image cropping is required before object detection. Most of the existing cropping methods crop by a fixed size, such as 512*512, but these existing cropping methods will damage the detection target, resulting in missed detection or false detection. To solve this problem, in step 203 of the present application, the SAR image is adaptively cropped to avoid damaging the detection target. In the actual application process, the implementation process of step 203 can be described as follows:

[0072] Step 2031: Use a clustering algorithm to perform clustering on the SAR image of one type of scattering feature to obtain a clustered image. The clustered image includes multiple clustering targets.

[0073] Step 2032: Determine the border position information of each clustering target in the clustered image.

[0074] Step 2033: Generate a cropping grid according to the original size of the wide-swath SAR image and the set cropping size.

[0075] Step 2034: Take a boundary point in the clustered image as the origin, set the boundary line of the cropping grid with the origin as the vertex to 0, and use the line parallel and adjacent to the boundary line in the cropping grid as the cropping line.

[0076] Step 2035: Crop the clustered image using the cropping grid, and during the cropping process, determine whether there are clustering targets on the cropping line.

[0077] Step 2036: If there are clustering targets on the cropping line, traverse all the clustering targets on the cropping line and determine the position information of the clustering target closest to the boundary line of the cropping grid among these clustering targets.

[0078] Step 2037: Determine the offset based on the position information of the clustering target closest to the boundary line of the cropping grid.

[0079] Step 2038: After offsetting the cropping grid along the cropping line by the offset amount towards the boundary line of the cropping grid, update the cropping grid and complete the cropping.

[0080] Step 2039: If there are no clustering targets on the cropping line, directly complete the cropping using the cropping grid.

[0081] In another exemplary embodiment of the present application, since there is certain noise such as sea clutter in the collected wide-swath SAR image, preprocessing is required to improve the image quality in order to minimize the impact of the noise on object detection. Taking the example of adaptive cropping along the x-axis and y-axis of the position information respectively, such as Figure 3As shown, the specific implementation processes of steps 202 and 203 can be described as follows:

[0082] The collected wide - swath SAR image is complex data. To segment the image, the complex data is converted into intensity data to generate a grayscale image. The target area is white with a pixel value of 255, the background area is black with a pixel value of 0, and the size of the grayscale image is (n, m), where n represents the length and m represents the width.

[0083] After generating the grayscale image, there may still be Gaussian noise, etc. in the grayscale image. For example, if there are individual background pixels 0 in the target area, or individual target pixels 255 in the background area, it will affect the region continuity. Therefore, morphological opening and closing operations are used to process the image to remove noise and interference in the grayscale image and improve the image quality.

[0084] Using a clustering algorithm, obtain the border positions of each clustering target in the denoised grayscale image. For example, (P x1 ,P y1 ,P x2 ,P y2 ). For example, using clustering algorithms such as k - means, cluster the target area and calculate the border position information of each clustering target. Among them, (P x1 ,P y1 ) is the upper - left point coordinate of the denoised grayscale image, and (P x2 ,P y2 ) is the lower - right point coordinate of the denoised grayscale image.

[0085] Generate a cropping grid according to the size of the originally obtained wide - swath SAR image and the preset cropping size. During the process of generating the cropping grid, if there are decimals, round down.

[0086] Taking the denoised grayscale image as the origin, define the coordinates of the cropping grid. Among them, the coordinates of the cropping grid can be defined as: The purpose of this processing is to adjust the cropping coordinates later to achieve adaptive cropping.

[0087] Define the left - most and top - most boundary lines of the cropping grid as 0, then let i, j = 1, and start cropping from the second line of the grid.

[0088] When performing vertical cropping using the cropping grid, judge from left to right. For example, if there are clustering targets on the second vertical cropping line of the cropping grid, find all the clustering targets on the cropping line and calculate the left - most coordinate x min of these clustering targets. Subsequently, shift all the second coordinate points x a , a ∈ [i, n / S] to the last coordinate point of the cropping grid to the left by xa -x min , that is, subtract x from each cropping grid coordinate one by one a -x min . Repeat this process until there are no cropping clustering targets in the vertical direction. Among them, n represents the original size of the image, and S represents the cropping size of the image.

[0089] Furthermore, the cropping idea in the horizontal direction is the same as that in the vertical direction. In the horizontal direction, judgments are made sequentially from top to bottom to ensure that the cropping clustering targets are not cropped in the horizontal direction. For the specific process, see Figure 3 as shown, and details will not be elaborated here.

[0090] In another exemplary embodiment of the present application, due to using different decomposition features (i.e., the cropped pictures obtained above), richer scattering information can be obtained. By analyzing various scattering information, it is beneficial to improve the detection accuracy and robustness of ship targets. Based on this, in step 205 of the present application, a reasoning network can be used to detect ship targets in multiple cropped pictures. Specifically, as Figure 4 shown, the implementation process of step 204 can be described as:

[0091] (1) Construct a ship target detection network (i.e., a reasoning network). The ship target detection network includes: a multi-channel convolutional attention network module, a multi-scale convolutional attention network module, and a hybrid encoder module.

[0092] (2) Input multiple cropped pictures and the position information of each cropped picture into the ship target detection network to obtain initial ship target detection information. The initial ship target detection information includes multiple prediction box information. The prediction box information includes: predicted target category, predicted bounding box, and predicted target position information. Among them:

[0093] (2-1) Input multiple cropped pictures and the position information of each cropped picture into the multi-channel convolutional attention network module to obtain a first feature. For example, as Figure 5 shown, by processing each decomposition feature (i.e., the cropped picture) through the CBS1 module, and then using the channel attention mechanism to modify the input different scattering features through the attention mechanism. Figure 5 In, the processing flow of the CBS1 module is: Conv1×1s1-->BN-->Silu. Among them, Conv1×1 is a 1×1 convolution. s1 is the step size 1. BN is short for BatchNormalization, indicating batch normalization. Silu is the Silu activation function.

[0094] (2-2) Input the first feature into the multi-scale convolutional attention network module to obtain a second feature. The second feature includes a first sub-feature, a second sub-feature, and a third sub-feature.

[0095] In the actual application process, due to the large differences in the sizes of detection targets, such as an aircraft carrier being dozens or even hundreds of times larger in volume than a fishing boat, a multi-scale attention network module is adopted to extract features of different scales. In this embodiment, a three-layer multi-scale convolutional attention network module (MSANet1, MSANet2, and MSANet3) is used to extract features of different depths, and the features of different depths are the first sub-feature, the second sub-feature, and the third sub-feature, denoted as S5, S4, and S3 respectively.

[0096] Such as Figure 6 As shown, in MSANet, first, a large-scale feature convolution kernel K1 is used to capture large target features. The large-scale feature convolution kernel K1 represents a large convolution kernel, such as 7*7, with a large receptive field. A small-scale feature convolution kernel K2 is used to capture small target features. The small-scale feature convolution kernel K2 is, for example, 3*3, that is, it has a relatively small receptive field. Subsequently, through a spatial attention mechanism, MSANet adaptively adjusts the spatial attention according to the needs of the detection target.

[0097] (2-3) Input the second feature into the hybrid encoder module to obtain the initial information for ship target detection. As Figure 7 In the structure of the hybrid encoder module shown, first, the third sub-feature S5 is flattened into a one-dimensional feature vector using formula (1). Then, the one-dimensional feature vector is encoded using the multi-head attention mechanism with formula (2) to achieve intra-scale feature interaction. Subsequently, the one-dimensional feature vector after multi-head attention mechanism encoding is converted into a two-dimensional vector F. Finally, the first sub-feature S3, the second sub-feature S4, and the two-dimensional feature vector F are fused bidirectionally and concatenated for output using formula (3). For example, a multi-branch convolution structure, the RepBlock module, is used to deeply fuse features of different layers.

[0098] Among them, the relevant processing formulas of the hybrid encoder module are expressed as:

[0099] Q = K = V = Flatten(S5) (1)

[0100] F = Reshape(Attn(Q, K, V)) (2)

[0101] Output = Fusion({S3, S4, F}) (3)

[0102] In the formula, Flatten represents the flattening process, Q, K, and V all represent one-dimensional feature vectors, Reshape(Attn()) represents multi-head attention mechanism encoding, Fusion represents bidirectional feature fusion and concatenation, and Output represents the output.

[0103] (3) Select the first j predicted box encoding information from the predicted box encoding information of the initial information of ship target detection by using the method of uncertainty minimization query and selection.

[0104] (4) Decode the first j predicted box encoding information to obtain the ship target detection information.

[0105] In another exemplary embodiment of the present application, in the above step 204, a certain number of encoded predicted boxes can be generated according to the result of the hybrid encoder module. To improve the pertinence of the encoded predicted boxes, as Figure 4 shown, the method of uncertainty minimization query and selection can be used to select the first j encoded predicted boxes from i encoded predicted boxes, thereby improving the detection speed.

[0106] Decode the encoded predicted boxes, and obtain the detected target category, border and position information.

[0107] Based on the same inventive concept, the embodiment of the present application also provides a wide-swath polarimetric SAR image ship target detection system for implementing the wide-swath polarimetric SAR image ship target detection method involved above. The implementation solutions provided by this system to solve problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the wide-swath polarimetric SAR image ship target detection system provided below can refer to the limitations on the wide-swath polarimetric SAR image ship target detection method in the above text, and will not be repeated here.

[0108] In an exemplary embodiment, as Figure 8 shown, a wide-swath polarimetric SAR image ship target detection system is provided, including:

[0109] A polarimetric SAR data acquisition module, configured to acquire wide-swath SAR data;

[0110] A wide-swath SAR image ship target adaptive cropping module, connected to the polarimetric SAR data acquisition module, configured to preprocess the wide-swath SAR data to obtain preprocessed data, extract scattering features from the preprocessed data to obtain multiple SAR images with different scattering features, obtain cropped pictures, and acquire the position information of the cropped pictures.

[0111] An SAR ship target real-time detection module, connected to the wide-swath SAR image ship target adaptive cropping module, configured to perform ship target detection based on the cropped pictures to obtain ship target detection information, and configured to map the ship target detection information to the wide-swath SAR image based on the position information of each cropped picture. The ship target detection information includes the ship target category, the ship target border and the position information of the ship target.

[0112] As an alternative implementation, the polarimetric SAR data acquisition module acquires polarimetric SAR data from the SAR sensor, and further obtains a wide-swath SAR image.

[0113] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal, and its internal structure diagram can be as Figure 9 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store ship target detection data. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a method for detecting ship targets in a wide-swath polarimetric SAR image.

[0114] Those skilled in the art can understand that Figure 9 the structure shown in

[0115] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps in the above method embodiments.

[0116] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0117] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0118] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0119] The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0120] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0121] In this text, specific examples are used to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A ship target detection method for wide-band polarimetric SAR images, characterized in that: The wide-band polarimetric SAR image ship target detection method comprises: Acquire wide-band SAR data; Preprocessing the wide-width SAR data to obtain preprocessed data; Extracting scattering features from the preprocessed data to obtain a plurality of SAR images with different scattering features; A SAR image with one scattering feature is selected for adaptive cropping, and the position information of each cropped image and the SAR image cropping grid are obtained, which specifically includes: clustering the SAR image with one scattering feature using a clustering algorithm to obtain a clustered image; the clustered image includes multiple clustered targets; the frame position information of each clustered target in the clustered image is determined; a cropping grid is generated according to the original size of the wide-band polarization SAR image and the set cropping size; a boundary point in the clustered image is taken as the origin, a boundary line of the cropping grid with the origin as the vertex is set to 0, and a line in the cropping grid that is parallel to and adjacent to the boundary line is cut off; as a cropping line; using the cropping grid to crop the cluster image, and during the cropping process, determining whether there is a cluster target on the cropping line; if there is a cluster target on the cropping line, traversing all cluster targets on the cropping line, and determining the position information of the cluster target closest to the boundary line of the cropping grid among the cluster targets; determining an offset based on the position information of the cluster target closest to the boundary line of the cropping grid; after offsetting the cropping grid along the cropping line toward the boundary line of the cropping grid by the offset, updating the cropping grid and completing the cropping; if there is no cluster target on the cropping line, directly using the cropping grid to complete the cropping; SAR images of other scattering features are cropped based on the cropping grid to obtain multiple cropped images of SAR data with different scattering features at the same location; Based on the multiple cropped images, ship target detection is performed to obtain ship target detection information, specifically including: constructing a ship target detection network; the ship target detection network includes: a multi-channel convolutional attention network module, a multi-scale convolutional attention network module and a hybrid encoder module; inputting the multiple cropped images and the position information of each cropped image into the ship target detection network to obtain initial ship target detection information; the initial ship target detection information includes multiple prediction box encoding information; the prediction box encoding information includes: predicted target category, predicted border and predicted target position information; using an uncertain minimization query and selection method, selecting the first j prediction box encoding information from the prediction box encoding information of the initial ship target detection information; decoding the first j prediction box encoding information to obtain the ship target detection information; the ship target detection information includes the ship target category, the ship target border and the position information of the ship target; The ship target detection information is mapped onto a wide-width polarization SAR image based on the position information of each of the cropped images.

2. The ship target detection method of wide-band polarimetric SAR image according to claim 1, characterized in that: Inputting the plurality of cropped images and the position information of each cropped image into the ship target detection network to obtain initial information of ship target detection specifically includes: Inputting the plurality of cropped images and the position information of each of the cropped images into the multi-channel convolutional attention network module to obtain a first feature; Inputting the first feature into the multi-scale convolutional attention network module to obtain a second feature; the second feature includes a first sub-feature, a second sub-feature and a third sub-feature; The second feature is input into the hybrid encoder module to obtain the initial information of the ship target detection.

3. The ship target detection method of wide-band polarimetric SAR image according to claim 2, characterized in that: Inputting the second feature into the hybrid encoder module to obtain the initial information of ship target detection specifically includes: In the hybrid encoder module, flattening the third sub-feature into a one-dimensional feature vector; The one-dimensional feature vector is encoded by a multi-head attention mechanism to achieve intra-scale feature interaction; Convert the one-dimensional feature vector encoded by the multi-head attention mechanism into a two-dimensional vector; The first sub-feature, the second sub-feature and the two-dimensional vector are bidirectionally fused and spliced ​​to output, so as to obtain the initial information of the ship target detection.

4. The ship target detection method of wide-band polarimetric SAR image according to claim 3 is characterized in that: A multi-branch convolution structure RepBlock module is set in the hybrid encoder module; A multi-branch convolution structure RepBlock module is used to perform bidirectional feature fusion on the first sub-feature, the second sub-feature and the two-dimensional vector.

5. A wide-band polarimetric SAR image ship target detection system, characterized in that: The wide-band polarimetric SAR image ship target detection system is used to implement the wide-band polarimetric SAR image ship target detection method according to any one of claims 1 to 4; the wide-band polarimetric SAR image ship target detection system comprises: Polarimetric SAR data acquisition module, used to acquire wide-band SAR data; A wide-width SAR image ship target adaptive cropping module is connected to the polarization SAR data acquisition module and is used to preprocess the wide-width SAR data to obtain preprocessed data, extract scattering features from the preprocessed data to obtain a plurality of SAR images with different scattering features, adaptively crop the SAR image to obtain a cropped image, and obtain position information of the cropped image; The SAR ship target real-time detection module is connected to the wide-width SAR image ship target adaptive cropping module, and is used to perform ship target detection based on the cropped image to obtain ship target detection information, and is used to map the ship target detection information to the wide-width polarization SAR image based on the position information of each cropped image; the ship target detection information includes the ship target category, the ship target border and the position information of the ship target.

6. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the wide-band polarimetric SAR image ship target detection method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for detecting ship targets in wide-band polarimetric SAR images according to any one of claims 1 to 4 is implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for detecting ship targets in wide-band polarimetric SAR images according to any one of claims 1 to 4 is implemented.

Images