Radar radiation source individual open set identification method and device based on dynamic characteristic distance, and medium

Abstract

The invention discloses a radar radiation source individual open set recognition method and device based on a dynamic feature distance and a medium, and the method comprises the steps: extracting a feature vector through a convolutional neural network model, and mapping the feature vector of a target to a feature space through t-SNE, and setting a feature space distance threshold based on a known target recall rate to perform open set identification. By means of the method, the appropriate distance threshold value can be independently set for each type of known targets, and therefore refined open set recognition is achieved. And meanwhile, different recall rate thresholds can be set according to the importance of different targets, so that the recognition effect is more targeted.

Description

Technical Field

[0001] This application belongs to the field of radar reconnaissance technology, and in particular relates to a method, device and medium for identifying individual open sets of radar radiation sources based on dynamic characteristic distance. Background Technology

[0002] Convolutional neural networks can use the softmax layer to output the probability of each known target class. Traditional open-set recognition methods use the maximum output probability of known targets as the confidence score, and use a confidence score threshold to define unknown targets. However, targets from radar sources of the same type have a high degree of similarity, such as... Figure 1 The confidence scores for known and unknown targets are largely overlapping, making it difficult to identify unknown targets using only the confidence scores.

[0003] To address this issue, scholars have proposed the OpenMax layer. The OpenMax layer replaces the traditional softmax layer and, by incorporating Extreme Value Theory (EVT) and the Weibull distribution, recalibrates the model's output rate. This allows for the detection and rejection of unknown class samples while maintaining the ability to classify known classes. However, in practical applications, the similarity between known and unknown targets varies. Using this method, a threshold can only be set for all known targets, failing to achieve refined recognition.

[0004] Table 1 Openmax Confusion Matrix

[0005] Table 2. Recognition performance of each class in Openmax.

[0006] As shown in Tables 1 and 2, Category 1 has a high similarity to the unknown target, while the other categories have a low similarity to the unknown category. When the threshold is set in the optimal recognition range for low similarity, the recognition effect of the high similarity category is not ideal. Summary of the Invention

[0007] The purpose of this application is to overcome the problems of existing technologies by disclosing a method, device, and medium for identifying individual radar radiation sources based on dynamic characteristic distance. This method can independently set appropriate distance thresholds for each type of known target, thereby achieving refined open-set identification. Furthermore, different recall thresholds can be set according to the importance of different targets, making the identification effect more targeted.

[0008] On the one hand, the objective of this application is achieved through the following technical solution: A method for identifying individual open sets of radar radiation sources based on dynamic characteristic distance, the method comprising: Feature vectors are extracted using a convolutional neural network model, and t-SNE is used to map the target's feature vectors to the feature space. Open set recognition is then performed by setting a distance threshold in the feature space based on the known target recall rate.

[0009] According to a preferred embodiment, the process of extracting feature vectors using a convolutional neural network model includes: Features of the input data are extracted through convolution, the features are reduced in dimensionality through pooling, the nonlinear problem is fitted through activation functions, the data is normalized through batch normalization, and the known target probability is obtained through a fully connected layer.

[0010] According to a preferred embodiment, an overall loss function suitable for open set recognition is constructed, which enables the classifier to classify known samples and separate known targets from unknown targets.

[0011] According to a preferred embodiment, the overall loss function includes: cross-entropy loss, triplet loss, and center loss.

[0012] According to a preferred embodiment, the cross-entropy loss is a loss function used for classification tasks, configured to measure the difference between the probability distribution predicted by the model and the true label distribution.

[0013] According to a preferred embodiment, the center loss is a loss function used for deep feature learning that makes the feature embeddings of each class close to its class center, thereby enhancing intra-class compactness. The central loss increases the clustering of similar samples by imposing constraints at the feature level, thereby helping the model learn more discriminative features. This is achieved by minimizing the Euclidean distance between sample features and their corresponding class centers.

[0014] According to a preferred embodiment, the triplet loss optimizes the distance relationship of triples through comparative learning of triples, thereby enhancing the feature embedding capability of the model, constraining the distance between samples, optimizing the relative distance of features, and thus enhancing discriminability.

[0015] According to a preferred embodiment, the open set recognition process includes: mapping the activation vectors output by the convolutional neural network into the feature space and clustering them, calculating the cluster center of each known class, and in the testing phase, thresholding the test signal based on the trained model and feature centers to obtain the recognition result of the test signal. That is, the center point is determined using known target samples obtained from training, and open set recognition is achieved by setting a dynamic feature distance threshold based on the center point and target recall rate during testing.

[0016] On the other hand, this application also discloses: An electronic device includes: at least one processor; and a memory communicatively connected to said at least one processor; The memory stores instructions that can be executed by the at least one processor, and the at least one processor executes the instructions stored in the memory to perform the aforementioned method.

[0017] On the other hand, this application also discloses: A computer-readable storage medium for storing instructions that, when executed, cause the aforementioned method to be implemented.

[0018] The aforementioned main solution and its various further alternative solutions can be freely combined to form multiple solutions, all of which are solutions that can be adopted and are claimed in this application. Those skilled in the art, after understanding the solution of this application, will realize that there are many combinations based on the prior art and common general knowledge, all of which are technical solutions to be protected in this application, and will not be exhaustively listed here.

[0019] The beneficial effects of this application are: In the process of identifying known and unknown samples of radar signals collected in actual use, this application achieves good separation of different individual targets in the feature space. By utilizing dynamic feature distance, the accuracy of identifying unknown individual targets can reach over 85%, and the recall rate can reach over 98%.

[0020] This application relates to an optimized method for identifying individual radiation sources, which can solve the problem of difficulty in identifying location targets in traditional identification methods, improve the generalization of handling unknown targets, and improve the identification effect of unknown targets. Attached Figure Description

[0021] Figure 1 These are schematic diagrams illustrating the information sharing between different radar models and the same radar model. Figure 2 This is a flowchart illustrating the method of this application; Figure 3 This is a schematic diagram of the feature extraction module; Figure 4 This is a schematic diagram of a convolutional neural network structure; Figure 5 This is a schematic diagram of the overall loss function; Figure 6 This is a schematic diagram illustrating the results of open set recognition using feature distance; Figure 7 This is a schematic diagram of the equipment structure in this application. Detailed Implementation

[0022] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0023] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0024] In the description of this application, it should also be noted that, unless otherwise expressly specified and limited, the terms "set up," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0025] Example 1: refer to Figures 2 to 5 As shown, this embodiment discloses a radar radiation source individual open set identification method based on dynamic feature distance. The radar radiation source individual open set identification method includes: extracting feature vectors through a convolutional neural network model, mapping the feature vectors of the target to the feature space using t-SNE, and performing open set identification based on a feature space distance threshold set according to the known target recall rate.

[0026] Preferably, this application uses a classic convolutional neural network to extract radiation source features. It extracts features from the input data through convolution, reduces the dimensionality of the features through pooling to improve the model's generalization ability, fits nonlinear problems through activation functions, normalizes the data through batch normalization to accelerate model convergence, and obtains the known target probability through a fully connected layer.

[0027] Preferably, an overall loss function suitable for open set recognition is constructed. This overall loss function enables the classifier to classify known samples and separate known and unknown targets. The overall loss function includes: cross-entropy loss, triplet loss, and center loss. Cross-entropy loss optimizes the classification boundary, triplet loss optimizes the sample problem, and center loss optimizes intra-class clustering, thereby obtaining the total loss. Intra-class clustering and inter-class separation of samples are achieved by using the joint loss of cross-entropy, center, and triplet loss.

[0028] Specifically, commonly used loss functions in deep learning include classification loss and clustering loss. Classification loss enables the classifier to classify known samples. Common classification losses include minimum mean squared error (MES) loss and cross-entropy loss. This application chooses cross-entropy loss as the classification loss. Cross-entropy loss is a loss function used for classification tasks, measuring the difference between the probability distribution predicted by the model and the true label distribution.

[0029] Cross-entropy loss is based on the concept of cross-entropy in information theory, and its goal is to make the class probabilities output by the model as close as possible to the distribution of the true labels. In multi-class classification tasks, cross-entropy loss is often used in conjunction with the softmax activation function to transform the model's raw output into probabilities.

[0030] Cross-entropy loss has advantages such as simplicity, effectiveness, direct optimization of classification accuracy, ease of computation, and easy gradient propagation. However, it only focuses on whether the classification is correct and does not directly optimize the distribution of features in space. This may result in features not being compactly distributed in space or having insufficient separability, making it unable to independently complete open-set recognition tasks for unknown targets.

[0031] To optimize the spatial distribution of features and facilitate the separation of known and unknown targets, this application introduces center loss and triplet loss. Center loss is a loss function used in deep feature learning, designed to make the feature embeddings of each class closer to its class center, thereby enhancing intra-class compactness. Center loss constrains at the feature level, increasing the clustering of similar samples, potentially helping the model learn more discriminative features. It achieves this by minimizing the Euclidean distance between sample features and their corresponding class centers.

[0032] Triple loss optimizes the distance relationship between triples (anchor, positive, negative) through comparative learning of triples, thereby enhancing the model's feature embedding ability, constraining the distance between samples, optimizing the relative distance of features, and enhancing discriminability.

[0033] Preferably, the open-set recognition process includes: mapping the activation vectors output by the convolutional neural network into the feature space and clustering them; calculating the cluster center for each known class; and, during the testing phase, thresholding the test signal based on the trained model and feature centers to obtain the recognition result for the test signal. That is, using the known target samples obtained through training to determine the center points, and setting a dynamic feature distance threshold based on the center points and target recall rate during testing to achieve open-set recognition.

[0034] This application achieves good separation of different individual targets in the feature space during the identification of known and unknown samples of radar signals collected in actual operation. Utilizing dynamic feature distance, it can achieve an accuracy rate of over 85% and a recall rate of over 98% for identifying unknown individual targets. This application belongs to an optimized method for identifying individual radiation sources, which can solve the problem of difficulty in identifying location targets in traditional identification methods, improve the generalization of handling unknown targets, and enhance the identification effect of unknown targets.

[0035] Table 3 Optimal Distance Confusion Matrix

[0036] like Figure 6 As shown in Table 3, the method of this application achieved good recognition results for unknown targets in open-set recognition tests of actual radar signal individuals. The recognition accuracy for unknown targets reached over 85%, and the recognition recall rate reached over 98%.

[0037] Example 2 like Figure 7 As shown, based on Embodiment 1, this embodiment also discloses an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; the specific connection medium between the processor and the memory is not limited in this embodiment of the invention.

[0038] Figure 7 The example used is the connection between the processor and memory via a bus. The bus... Figure 7 The connections between other components are indicated by thick lines and are for illustrative purposes only, not as limiting information. Buses can be divided into address buses, data buses, control buses, etc., but for ease of representation, [the specific bus type is not shown here]. Figure 7 The processor is represented by a single thick line, but this does not imply that there is only one bus or one type of bus. Alternatively, a processor can also be called a controller; there are no restrictions on the name.

[0039] In this embodiment, the memory stores instructions executable by the at least one processor. By executing the instructions stored in the memory, the at least one processor performs the method described in Embodiment 1. The processor can implement... Figure 7 The functions of each module in the device shown.

[0040] The processor is the control center of the device. It can connect to various parts of the control device through various interfaces and lines. By running or executing instructions stored in memory and calling data stored in memory, it can monitor the device's various functions and process data, thereby enabling overall monitoring of the device.

[0041] In an alternative design, the processor may include one or more processing units. The processor may integrate an application processor and a modem processor, wherein the application processor primarily handles the operating system, user interface, and applications, while the modem processor primarily handles wireless communication. It is understood that the modem processor may also not be integrated into the processor. In some embodiments, the processor and memory may be implemented on the same chip; in some embodiments, they may also be implemented separately on separate chips.

[0042] The processor can be a general-purpose processor, such as a CPU, digital signal processor, application-specific integrated circuit, field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the radar radiation source individual open set identification method based on dynamic characteristic distance disclosed in the embodiments of this invention can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0043] Memory, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory can include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic memory, magnetic disk, optical disk, etc. Memory is any other medium capable of carrying or storing desired program code in the form of instructions or data structures, and accessible by a computer, but is not limited thereto. In embodiments of the present invention, memory can also be a circuit or any other device capable of implementing storage functions, used to store program instructions and / or data.

[0044] By designing and programming the processor, the code corresponding to the radar radiation source individual open set identification method based on dynamic feature distance described in the foregoing embodiments can be embedded into the chip, thereby enabling the chip to execute the steps of the method described in the foregoing embodiments during runtime. How to design and program the processor is a technique well known to those skilled in the art, and will not be elaborated here.

[0045] Example 3 Based on Embodiment 1, this embodiment also discloses: a computer-readable storage medium for storing instructions that, when executed, cause the method described in Embodiment 1 to be implemented.

[0046] In some alternative embodiments, the present invention also provides that various aspects of the radar emitter individual open set identification method based on dynamic characteristic distance can also be implemented in the form of a program product, which includes program code that, when the program product is run on a device, causes the control device to perform the steps in the radar emitter individual open set identification method based on dynamic characteristic distance according to various exemplary embodiments of the present invention as described above.

[0047] It should be noted that although several units or sub-units of the apparatus have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the invention, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units. Furthermore, although the operation of the method of the invention is described in a specific order in the drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0048] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0049] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a server, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0050] Program code for performing the operations of this invention can be written using any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0051] In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0052] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0053] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the functions specified in one or more boxes. The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for identifying individual open-set radar radiation sources based on dynamic characteristic distance, characterized in that, The radar radiation source individual open set identification method includes: Feature vectors are extracted using a convolutional neural network model, and t-SNE is used to map the target's feature vectors to the feature space. Open set recognition is then performed by setting a distance threshold in the feature space based on the known target recall rate.

2. The radar radiation source individual open set identification method based on dynamic feature distance as described in claim 1, characterized in that, The process of extracting feature vectors using a convolutional neural network model includes: Features of the input data are extracted through convolution, the features are reduced in dimensionality through pooling, the nonlinear problem is fitted through activation functions, the data is normalized through batch normalization, and the known target probability is obtained through a fully connected layer.

3. The radar radiation source individual open set identification method based on dynamic feature distance as described in claim 1, characterized in that, An overall loss function suitable for open set recognition is constructed. This overall loss function enables the classifier to classify known samples and separate known targets from unknown targets.

4. The radar radiation source individual open set identification method based on dynamic feature distance as described in claim 3, characterized in that, The overall loss function includes: cross-entropy loss, triplet loss, and center loss.

5. The radar radiation source individual open set identification method based on dynamic feature distance as described in claim 4, characterized in that, The cross-entropy loss is a loss function used for classification tasks, configured to measure the difference between the probability distribution predicted by the model and the true label distribution.

6. The radar radiation source individual open set identification method based on dynamic feature distance as described in claim 4, characterized in that, The center loss is a loss function used for deep feature learning, which makes the feature embedding of each class close to its class center, thereby enhancing intra-class compactness. The central loss increases the clustering of similar samples by imposing constraints at the feature level, thereby helping the model learn more discriminative features. This is achieved by minimizing the Euclidean distance between sample features and their corresponding class centers.

7. The radar radiation source individual open set identification method based on dynamic feature distance as described in claim 4, characterized in that, The triplet loss optimizes the distance relationship of triples through comparative learning of triples, thereby enhancing the feature embedding ability of the model, constraining the distance between samples, optimizing the relative distance of features, and thus enhancing discriminability.

8. The radar radiation source individual open set identification method based on dynamic feature distance as described in claim 4, characterized in that, The open set recognition process includes: The activation vectors output by the convolutional neural network are mapped into the feature space and clustered. The cluster center of each known class is calculated. During the testing phase, the test signal is thresholded based on the model obtained from training and the feature centers to obtain the recognition result of the test signal. That is, the center point is determined using known target samples obtained from training, and open set recognition is achieved by setting a dynamic feature distance threshold based on the center point and target recall rate during testing.

9. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions executable by the at least one processor, which executes the instructions stored in the memory to perform the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store instructions that, when executed, cause the method as described in any one of claims 1 to 8 to be implemented.

Images