Clustering-based multi-prototype SAR image out-of-domain data post-detection method and device

Through the cluster-based multi-prototype SAR image extra-domain data detection method, the distance judgment between the feature prototype obtained by clustering and the target feature vector is solved, and the problem of low extra-domain data recognition efficiency in SAR images in the prior art is achieved, and more efficient and stable target recognition.

CN119048909APending Publication Date: 2024-11-29NAT UNIV OF DEFENSE TECH
View PDF 0 Cites 6 Cited by

Patent Information

Application Number
CN202411169200.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

Existing deep learning models are difficult to effectively identify extradomain data in SAR image processing, especially in open world scenarios, resulting in inefficient and poor performance of target recognition.

Method used

Using the multi-prototype SAR image extradomain data post-detection method based on clustering, the feature data set is extracted through pre-trained classification neural network, and multiple clusters are obtained by hierarchical clustering. The cluster center is used as feature prototypes, and the target feature vector is used with The distance between feature prototypes is judged to determine whether the image is extraterritorial data.

Benefits of technology

This method can accurately detect extradomain data, improve the accuracy of object detection, enhance the reliability of object recognition in an open environment, avoid the limitation of a single prototype representing each category, and improve the model's characterization ability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048909A_ABST
    Figure CN119048909A_ABST
Patent Text Reader

Abstract

The invention relates to a clustering-based multi-prototype SAR image out-of-domain data post-detection method and device, and the method comprises the steps: carrying out the feature extraction of all SAR sample images in a training data set through a feature extraction part of a pre-trained classification neural network, obtaining a feature data set, and carrying out the detection of the out-of-domain data of a multi-prototype SAR image according to a preset prototype number, performing hierarchical clustering based on the feature data set to obtain a plurality of clusters, taking the cluster center of each cluster as a feature prototype, and performing feature extraction on the SAR image to be subjected to target classification by using a feature extraction part of a pre-trained classification neural network to obtain target features, and judging according to the distance between the feature vector of the target feature and each feature prototype to obtain a detection result that the SAR image is out-of-domain data or in-domain data. By adopting the method, the out-of-domain data can be accurately and quickly detected, so that the reliability of the classification model in practical application is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of SAR image target recognition, and in particular to a clustering-based multi-prototype SAR image out-of-domain data post-detection method and device. Background Art

[0002] Synthetic Aperture Radar (SAR) has all-weather and all-time observation capabilities and is widely used in agriculture, industry and other fields. With the improvement of SAR imaging resolution, Automatic Target Recognition (ATR) has become the most important technology in SAR image processing in the fields of airport management, military reconnaissance and maritime surveillance.

[0003] However, existing deep learning models are based on the independent and identically distributed (iid) assumption, which means that the training data and test data come from the same distribution. However, this assumption may not hold in practice, for example, new categories that have never been seen before may appear in the test data or the data distribution may change due to changes in imaging conditions. Therefore, it is crucial to achieve out-of-distribution (OOD) data detection for SAR targets in open-world scenarios, which can reliably identify in-distribution (ID) categories and reject OOD data. Summary of the invention

[0004] Based on this, it is necessary to provide a clustering-based multi-prototype SAR image out-of-domain data post-detection method and device that can accurately identify out-of-domain data in response to the above technical problems.

[0005] A clustering-based multi-prototype SAR image out-of-domain data post-detection method, characterized in that the method comprises:

[0006] Acquire a training data set, wherein the training data set includes SAR sample images of targets of different categories;

[0007] Using a feature extraction part in a pre-trained classification neural network to extract features from all SAR sample images in the training data set, to obtain a feature data set;

[0008] According to a preset number of prototypes, hierarchical clustering is performed based on the feature data set to obtain a plurality of clusters, and the cluster center of each cluster is used as the feature prototype;

[0009] Acquire a SAR image for target classification, and use the feature extraction part of the pre-trained classification neural network to extract features from the SAR image to obtain target features;

[0010] A judgment is made based on the distance between the feature vector of the target feature and each feature prototype to obtain a detection result of whether the SAR image is out-of-domain data or in-domain data.

[0011] In one embodiment, the target categories in the training data set are known target categories, and there is at least one feature prototype corresponding to each known target category.

[0012] In one embodiment, clustering based on the feature data set adopts any one of the following methods: K-Means method, hierarchical clustering method, DNSCAN clustering method, mean shift clustering method or spectral clustering method.

[0013] In one embodiment, clustering the feature data set to obtain a plurality of clusters according to a preset number of prototypes specifically includes:

[0014] Calculate according to the feature data set to obtain a corresponding distance matrix;

[0015] Perform hierarchical clustering according to the distance matrix to obtain a dendrogram;

[0016] The dendrogram is cut to obtain a plurality of clusters.

[0017] In one embodiment, the cluster center of each cluster is calculated using the following formula:

[0018]

[0019] In the above formula, μ i represents the cluster center of the i-th cluster, C i represents the i-th cluster, z j Represents the jth feature of the feature dataset.

[0020] In one embodiment, the judging based on the distance between the feature vector of the target feature and each feature prototype includes:

[0021] Calculate the distance between the feature vector of the target feature and each feature prototype, and arrange all feature prototypes in increasing order according to the calculated distance;

[0022] The feature prototype ranked first and the feature vector of the target feature are judged by using out-of-domain data detection decision to determine whether the SAR image is out-of-domain data. If it is out-of-domain data, target recognition is rejected;

[0023] If it is in-domain data, the target type in the SAR image is determined to be the target category corresponding to the feature prototype arranged in the first place, or the classification neural network is used for classification.

[0024] In one embodiment, the distance between the feature vector of the target feature and each feature prototype is calculated using the following formula:

[0025]

[0026] In the above formula, u i represents the cluster center of the i-th cluster, z * A feature vector representing the target feature.

[0027] In one embodiment, the out-of-domain data detection decision is:

[0028] When the distance between the feature vector of the target feature and the feature prototype ranked first is greater than or equal to a preset threshold, the SAR image is judged to be out-of-domain data;

[0029] When the distance between the feature vector of the target feature and the feature prototype ranked first is less than a preset threshold, the SAR image is judged as in-domain data, and the target category is determined to be the target category corresponding to the feature prototype ranked first, or the classification neural network is used for classification.

[0030] In one embodiment, the out-of-domain data detection decision is expressed as:

[0031] 1{||z * -μ (1) || 2 ≥λ}

[0032] In the above formula, z * The feature vector representing the target feature, μ (1) represents the feature prototype ranked first, λ represents the preset threshold, and 1{·} represents the indicator function.

[0033] The present application also provides a clustering-based multi-prototype SAR image out-of-domain data post-detection device, the device comprising:

[0034] A training data set acquisition module is used to acquire a training data set, wherein the training data set includes SAR sample images of targets of different categories;

[0035] A feature data set acquisition module is used to extract features from all SAR sample images in the training data set using a feature extraction part in a pre-trained classification neural network to obtain a feature data set;

[0036] A feature prototype obtaining module, used for clustering the feature data set to obtain a plurality of clusters according to a preset number of prototypes, and taking the cluster center of each cluster as the feature prototype;

[0037] A target feature extraction module is used to obtain a SAR image to be classified, and to extract features from the SAR image using a feature extraction part in the pre-trained classification neural network to obtain target features;

[0038] The in-domain and out-domain data judgment module is used to judge according to the distance between the feature vector of the target feature and each feature prototype, and obtain the detection result of whether the SAR image is out-domain data or in-domain data.

[0039] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0040] Acquire a training data set, wherein the training data set includes SAR sample images of targets of different categories;

[0041] Using a feature extraction part in a pre-trained classification neural network to extract features from all SAR sample images in the training data set, to obtain a feature data set;

[0042] According to a preset number of prototypes, hierarchical clustering is performed based on the feature data set to obtain a plurality of clusters, and the cluster center of each cluster is used as the feature prototype;

[0043] Acquire a SAR image for target classification, and use the feature extraction part of the pre-trained classification neural network to extract features from the SAR image to obtain target features;

[0044] A judgment is made based on the distance between the feature vector of the target feature and each feature prototype to obtain a detection result of whether the SAR image is out-of-domain data or in-domain data.

[0045] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0046] Acquire a training data set, wherein the training data set includes SAR sample images of targets of different categories;

[0047] Using a feature extraction part in a pre-trained classification neural network to extract features from all SAR sample images in the training data set, to obtain a feature data set;

[0048] According to a preset number of prototypes, hierarchical clustering is performed based on the feature data set to obtain a plurality of clusters, and the cluster center of each cluster is used as the feature prototype;

[0049] Acquire a SAR image for target classification, and use the feature extraction part of the pre-trained classification neural network to extract features from the SAR image to obtain target features;

[0050] A judgment is made based on the distance between the feature vector of the target feature and each feature prototype to obtain a detection result of whether the SAR image is out-of-domain data or in-domain data.

[0051] The above-mentioned clustering-based multi-prototype SAR image out-of-domain data post-detection method and device, by using the feature extraction part of the pre-trained classification neural network to extract features from all SAR sample images in the training data set, obtain a feature data set, then hierarchically cluster the feature data set according to the preset number of prototypes to obtain multiple clusters, and use the cluster center of each cluster as the feature prototype, and use the feature extraction part of the pre-trained classification neural network to extract features from the SAR image to be classified as a target to obtain target features, and judge according to the distance between the feature vector of the target feature and each feature prototype to obtain the detection result of whether the SAR image is out-of-domain data or in-domain data. This method can accurately detect out-of-domain data to improve the accuracy of target detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flowchart of a post-detection method of out-of-domain data of multiple prototype SAR images based on clustering in one embodiment;

[0053] Figure 2 A schematic diagram of the motivation of an out-of-domain data detection method based on dynamic multiple prototypes in one embodiment;

[0054] Figure 3 It is a structural block diagram of a post-detection device for out-of-domain data of multiple prototype SAR images based on clustering in one embodiment;

[0055] Figure 4 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0057] In view of the problem that in the prior art, when using the post-hoc method to classify SAR image targets, especially for the detection of out-of-domain data, the efficiency is low and the detection performance is poor, in this application, Figure 1 As shown, a clustering-based multi-prototype SAR image out-of-domain data post-detection method is provided, comprising the following steps:

[0058] Step S100: acquiring a training data set, wherein the training data set includes SAR sample images of targets of different categories.

[0059] Step S110, using the feature extraction part of the pre-trained classification neural network to extract features from all SAR sample images in the training data set to obtain a feature data set.

[0060] Step S120 , hierarchical clustering is performed based on the feature data set according to the preset number of prototypes to obtain multiple clusters, and the cluster center of each cluster is used as the feature prototype.

[0061] Step S130, obtaining a SAR image for target classification, and using a feature extraction part in a pre-trained classification neural network to extract features from the SAR image to obtain target features.

[0062] Step S140 , judging based on the distance between the feature vector of the target feature and each feature prototype, and obtaining a detection result of whether the SAR image is out-of-domain data or in-domain data.

[0063] In this embodiment, the focus is on detecting unknown targets in SAR images, that is, out-of-domain data. Since SAR target classification can generally only identify known categories, that is, target category features that have appeared when the algorithm is trained, but in practice, unknown categories often appear. At this time, the use of existing target recognition methods will identify unknown targets as known categories, thereby making wrong decisions. At the same time, when encountering unknown category targets, no clear warning can be given, indicating that the target in the SAR image does not belong to any category of known targets. This will not ensure its reliability and security in many scenarios that require warnings. Therefore, the out-of-domain data detection proposed in this application is actually a process step before target recognition, which aims to detect out-of-domain data. When the result is out-of-domain data, subsequent target recognition is rejected. When the detection result is in-domain data, the target category recognition work is further performed.

[0064] In the existing post-detection algorithms, the out-of-domain data detection method of optical images is often used in the target recognition task of SAR images. However, due to the differences between SAR images and optical images, this type of algorithm cannot achieve good recognition results on SAR images. In view of this point, re-examining the differences between optical and SAR images shows that SAR images reflect the electromagnetic scattering characteristics of the target and have a low signal-to-noise ratio. Given the characteristics of SAR images, the inter-class differences are small and the intra-class variations are large. Using a single prototype to represent each category significantly limits the representation ability of the model. However, this simplified method of using a single prototype to represent each category fails to take into account a large amount of intra-class variations, resulting in the existing detection algorithms being difficult to distinguish out-of-domain data from in-domain data, that is, known target categories and unknown target categories.

[0065] In this embodiment, a dynamic multi-prototype SAR out-of-domain data detection method is proposed, where dynamic means that the number of prototypes for each category is not fixed, but is determined by the characteristics of each target category, such as Figure 2 shown.

[0066] In step 100, first, a feature prototype of each known target category is obtained through a training data set. The training data set includes SAR sample images of multiple different target categories, and each target category includes multiple SAR sample images. The target categories appearing in the training data set are known target categories. In practice, when the target category in the SAR image to be detected does not belong to any known target category, it is an unknown target category, that is, out-of-domain data.

[0067] In step S110, a feature extraction part of a pre-trained classification neural network is used to extract features from all sample images in the training data set. The feature extraction part of the classification neural network can use a neural network encoder.

[0068] In step S120, hierarchical clustering is performed based on the feature data set according to the preset number of prototypes to obtain multiple clusters, specifically including: calculating according to the feature data set to obtain the corresponding distance matrix, then performing hierarchical clustering according to the distance matrix to obtain a dendrogram, and finally cutting the dendrogram to obtain multiple clusters.

[0069] In this embodiment, the means of clustering based on the feature data set may include K-Means method, hierarchical clustering method, DNSCAN clustering method, mean shift clustering method or spectral clustering method, any of which may be used to implement clustering. It is understandable that other clustering means may also be used.

[0070] In this embodiment, since only the total number of prototypes is given and the prototypes of each target type are not specified, multiple clusters will be dynamically obtained after the feature clustering in step S120. At this time, one or more corresponding clusters will be obtained according to the class difference of each target category, and the center of each cluster represents the corresponding feature prototype. At this time, there are one or more feature prototypes corresponding to each known target category.

[0071] In this embodiment, the cluster center of each cluster is calculated using the following formula:

[0072]

[0073] In the above formula, μ i represents the cluster center of the i-th cluster, C i represents the i-th cluster, z j Represents the jth feature of the feature dataset.

[0074] Different from the idea of ​​using the distance between training samples and test samples for target recognition in the prior art, this method uses the distance between the test sample, i.e. the sample to be detected, and each feature prototype for target recognition, which makes this method more efficient than the prior art. At the same time, it can be seen from the cluster center calculation method shown above that the feature prototype is the average value of the clustering cluster, which makes it more stable.

[0075] After obtaining feature prototypes of multiple known target categories through training data sets, they can be applied in practice to detect targets in SAR images. It should be noted here that one target category does not correspond to one prototype, but to at least one prototype, which depends on the intra-class distinction of the target.

[0076] In step S130, the feature extraction part of the pre-trained classification neural network is first used to extract features from the SAR image to be classified to obtain target features.

[0077] In step S140, the target category is judged according to the distance between the feature vector of the target feature and each feature prototype, including: firstly calculating the distance between the feature vector of the target feature and each feature prototype, and arranging all feature prototypes in increasing order according to the calculated distance, and then judging whether the SAR image is out-of-domain data by using out-of-domain data detection decision with the feature prototype arranged in the first place and the feature vector of the target feature, and rejecting the recognition if it is out-of-domain data. If it is in-domain data, further target recognition is performed to determine that the target type in the SAR image is the target category corresponding to the feature prototype arranged in the first place, or the target classification can be directly performed by using a classification network according to the target feature.

[0078] In this embodiment, the distance between the feature vector of the target feature and each feature prototype is calculated using the following formula:

[0079]

[0080] In the above formula, u i represents the cluster center of the i-th cluster, z * Feature vector representing the target features.

[0081] Furthermore, the eigenvector is calculated using the following formula:

[0082]

[0083] In the above formula, φ(x * ) represents the feature extraction part φ of the classification neural network according to the SAR image x * The target features are obtained by feature extraction.

[0084] In this embodiment, the corresponding feature prototypes are arranged in ascending order according to the calculated distance between the feature vector of the target feature and each feature prototype. In this way, the feature prototype arranged first is the prototype closest to the target feature in the SAR image. Then, the out-of-domain data detection decision is used to judge the out-of-domain data, including: when the distance between the feature vector of the target feature and the feature prototype arranged in the first position is greater than or equal to a preset threshold, the SAR image is judged to be out-of-domain data, when the distance between the feature vector of the target feature and the feature prototype arranged in the first position is less than the preset threshold, the SAR image is judged to be in-domain data, and the target category in the SAR image can be determined by the target category corresponding to the feature prototype arranged in the first position.

[0085] It should be noted here that this method focuses on the detection of out-of-domain data, not the identification of target types. Therefore, after obtaining the detection results through this method, the category corresponding to the nearest feature prototype can be used as the target category to achieve target identification, and other methods can be used for subsequent target identification. At that time, when it is detected as out-of-domain data through this method, the target category will no longer be identified in the future.

[0086] Furthermore, the out-of-domain data detection decision is expressed as:

[0087] 1{||z * -μ (1) || 2 ≥λ}

[0088] In the above formula, z * The feature vector representing the target feature, μ (1)represents the first-ranked feature prototype, λ represents the preset threshold, and 1{·} represents the indicator function.

[0089] Specifically, if the distance between the target feature and the nearest feature prototype is greater than or equal to a preset threshold, this means that the feature is not close to any prototype of a known target category, and then it can be determined that the current SAR image to be detected is out-of-domain data.

[0090] Specifically, if the distance between the target feature and the nearest feature prototype is less than a preset threshold, this indicates that the feature belongs to the nearest feature prototype, and at this time, it can be determined that the current SAR image to be classified is in-domain data.

[0091] The above-mentioned clustering-based multi-prototype SAR image out-of-domain data post-detection method uses the feature extraction part of the pre-trained classification neural network to extract features from all SAR sample images in the training data set to obtain a feature data set, and then clusters the feature data set according to the preset number of prototypes to obtain multiple clusters, and uses the cluster center of each cluster as the feature prototype, and uses the feature extraction part of the pre-trained classification neural network to extract features from the SAR image to be classified to obtain target features, and judges according to the distance between the feature vector of the target feature and each feature prototype to obtain the SAR image as out-of-domain data. If it is in-domain data, the target category can be further determined. This method uses the distance between the prototype and the sample to be detected as the OOD score function, which has lower storage requirements than the existing KNN algorithm, higher computational efficiency, and more stable performance. In addition, this method also has the advantages of no distribution assumption, OOD-independent, easy to use and model-independent, and can accurately detect out-of-domain data to improve the reliability of target recognition in an open environment.

[0092] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0093] In one embodiment, Figure 3As shown, a clustering-based multi-prototype SAR image out-of-domain data post-detection device is provided, comprising: a training data set acquisition module 200, a feature data set acquisition module 210, a feature prototype acquisition module 220, a target feature extraction module 230 and an in-domain and out-of-domain data judgment module 240, wherein:

[0094] The training data set acquisition module 200 is used to acquire a training data set, wherein the training data set includes SAR sample images of targets of different categories;

[0095] A feature data set obtaining module 210 is used to extract features from all SAR sample images in the training data set using a feature extraction part in a pre-trained classification neural network to obtain a feature data set;

[0096] A feature prototype obtaining module 220 is used to perform hierarchical clustering based on the feature data set to obtain multiple clusters according to a preset number of prototypes, and use the cluster center of each cluster as the feature prototype;

[0097] The target feature extraction module 230 is used to obtain a SAR image to be subjected to target classification, and to extract features from the SAR image using the feature extraction part of the pre-trained classification neural network to obtain target features;

[0098] The in-domain and out-domain data judgment module 240 is used to judge the target category according to the distance between the feature vector of the target feature and each feature prototype, and determine whether the target category in the SAR image is an unknown category target. If it is a known category target, the target category is determined.

[0099] The specific definition of the out-of-domain data post-detection device for multiple prototype SAR images based on clustering can be found in the definition of the out-of-domain data post-detection method for multiple prototype SAR images based on clustering, which will not be repeated here. Each module in the above-mentioned out-of-domain data post-detection device for multiple prototype SAR images based on clustering can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0100] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 4As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. 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 and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network 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, a post-detection method for out-of-domain data of multi-prototype SAR images based on clustering is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse, etc.

[0101] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure 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 a different arrangement of components.

[0102] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0103] Acquire a training data set, wherein the training data set includes SAR sample images of targets of different categories;

[0104] Using a feature extraction part in a pre-trained classification neural network to extract features from all SAR sample images in the training data set, to obtain a feature data set;

[0105] According to a preset number of prototypes, hierarchical clustering is performed based on the feature data set to obtain a plurality of clusters, and the cluster center of each cluster is used as the feature prototype;

[0106] Acquire a SAR image for target classification, and use the feature extraction part of the pre-trained classification neural network to extract features from the SAR image to obtain target features;

[0107] A judgment is made based on the distance between the feature vector of the target feature and each feature prototype to obtain a detection result of whether the SAR image is out-of-domain data or in-domain data.

[0108] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0109] Acquire a training data set, wherein the training data set includes SAR sample images of targets of different categories;

[0110] Using a feature extraction part in a pre-trained classification neural network to extract features from all SAR sample images in the training data set, to obtain a feature data set;

[0111] According to a preset number of prototypes, hierarchical clustering is performed based on the feature data set to obtain a plurality of clusters, and the cluster center of each cluster is used as the feature prototype;

[0112] Acquire a SAR image for target classification, and use the feature extraction part of the pre-trained classification neural network to extract features from the SAR image to obtain target features;

[0113] A judgment is made based on the distance between the feature vector of the target feature and each feature prototype to obtain a detection result of whether the SAR image is out-of-domain data or in-domain data.

[0114] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and 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-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0115] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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, they should be considered to be within the scope of this specification.

[0116] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A clustering-based out-of-domain data post-detection method for multi-prototype SAR images, characterized in that: The method comprises: Acquire a training data set, wherein the training data set includes SAR sample images of targets of different categories; Using a feature extraction part in a pre-trained classification neural network to extract features from all SAR sample images in the training data set to obtain a feature data set; According to a preset number of prototypes, hierarchical clustering is performed based on the feature data set to obtain a plurality of clusters, and the cluster center of each cluster is used as the feature prototype; Acquire a SAR image for target classification, and use the feature extraction part of the pre-trained classification neural network to extract features from the SAR image to obtain target features; A judgment is made based on the distance between the feature vector of the target feature and each feature prototype to obtain a detection result of whether the SAR image is out-of-domain data or in-domain data.

2. The method for post-detection of out-of-domain data of multiple prototype SAR images according to claim 1, characterized in that: The target categories in the training data set are known target categories, corresponding to at least one feature prototype of each known target category.

3. The method for post-detection of out-of-domain data of multiple prototype SAR images according to claim 2, characterized in that: Clustering based on the feature data set adopts any one of the following methods: K-Means method, hierarchical clustering method, DNSCAN clustering method, mean shift clustering method or spectral clustering method.

4. The method for post-detection of out-of-domain data of multiple prototype SAR images according to claim 3, characterized in that: The step of clustering the feature data set according to the preset number of prototypes to obtain a plurality of clusters specifically includes: Calculate according to the feature data set to obtain a corresponding distance matrix; Perform hierarchical clustering according to the distance matrix to obtain a dendrogram; The dendrogram is cut to obtain a plurality of clusters.

5. The method for post-detection of out-of-domain data of multiple prototype SAR images according to claim 3, characterized in that: The cluster center of each cluster is calculated using the following formula: In the above formula, μ i represents the cluster center of the i-th cluster, C i represents the i-th cluster, z j Represents the jth feature of the feature dataset.

6. The method for post-detection of out-of-domain data of multiple prototype SAR images according to claim 3, characterized in that: The judging according to the distance between the feature vector of the target feature and each feature prototype includes: Calculate the distance between the feature vector of the target feature and each feature prototype, and arrange all feature prototypes in increasing order according to the calculated distance; The feature prototype ranked first and the feature vector of the target feature are judged by using out-of-domain data detection decision to determine whether the SAR image is out-of-domain data. If it is out-of-domain data, target recognition is rejected; If it is in-domain data, the target type in the SAR image is determined to be the target category corresponding to the feature prototype arranged in the first place, or the classification neural network is used for classification.

7. The method for post-detection of out-of-domain data of multiple prototype SAR images according to claim 6, characterized in that: The distance between the feature vector of the target feature and each feature prototype is calculated using the following formula: In the above formula, u i represents the cluster center of the i-th cluster, z * A feature vector representing the target feature.

8. The method for post-detection of out-of-domain data of multiple prototype SAR images according to claim 6, characterized in that: The out-of-domain data detection decision is: When the distance between the feature vector of the target feature and the feature prototype ranked first is greater than or equal to a preset threshold, the SAR image is judged to be out-of-domain data; When the distance between the feature vector of the target feature and the feature prototype ranked first is less than a preset threshold, the SAR image is judged as in-domain data, and the target category is determined to be the target category corresponding to the feature prototype ranked first, or the classification neural network is used for classification.

9. The method for post-detection of out-of-domain data of multiple prototype SAR images according to claim 8, characterized in that: The out-of-domain data detection decision is expressed as: 1{||of * -μ (1) ||2≥λ} In the above formula, z * The feature vector representing the target feature, μ (1) represents the feature prototype ranked first, λ represents the preset threshold, and 1{·} represents the indicator function.

10. A clustering-based multi-prototype SAR image out-of-domain data post-detection device, characterized in that: The device comprises: A training data set acquisition module is used to acquire a training data set, wherein the training data set includes SAR sample images of targets of different categories; A feature data set acquisition module is used to extract features from all SAR sample images in the training data set using a feature extraction part in a pre-trained classification neural network to obtain a feature data set; A feature prototype obtaining module, used for clustering the feature data set to obtain a plurality of clusters according to a preset number of prototypes, and taking the cluster center of each cluster as the feature prototype; A target feature extraction module is used to obtain a SAR image to be classified, and to extract features from the SAR image using a feature extraction part in the pre-trained classification neural network to obtain target features; The in-domain and out-domain data judgment module is used to judge according to the distance between the feature vector of the target feature and each feature prototype, and obtain the detection result of whether the SAR image is out-domain data or in-domain data.

Citation Information

Cited By

  • SAR image out-of-domain data detection method and device based on dynamic multiple prototypes

    CN119048910A

  • SAR image domain external data detection method and device based on dynamic multi-prototype

    CN119048910B

  • SAR imaging processing chip multi-computing-unit efficient clustering interconnection method

    CN121093026A

  • Method, device and equipment for few-sample SAR target recognition and medium

    CN121640286A

  • Method and apparatus for few-shot sar target recognition, and medium

    CN121640286B