Radar target identification method suitable for small sample condition

Abstract

The invention discloses a radar target identification method suitable for a small sample condition, which solves the problems of poor generalization ability and unstable performance of a model caused by small sample data in the prior art, and comprises the following steps: obtaining radar data to be classified; training the radar target recognition model by using a similarity constraint method to obtain a trained radar target recognition model; wherein the similarity constraint method is applied to the last high-dimensional layer after feature extraction in a radar target recognition model; inputting the radar data to be classified into the trained radar target recognition model to obtain a recognition result; according to the method, the features are adjusted in the high-dimensional feature space, so that the relationship between the data is better captured, the similarity and the difference between the data can be better understood, the characterization capability and the interpretability of the data are improved, and more powerful support is provided for data analysis and model training.

Description

Technical Field

[0001] The present invention relates to the technical field of radar target recognition, and in particular to a radar target recognition method suitable for a small sample situation. Background Art

[0002] The difficulty of field collection of radar data mainly stems from many factors. First, radar equipment usually needs to be deployed in a specific geographical location, and some areas may be difficult to reach due to complex terrain or harsh environment, resulting in limited data collection. Secondly, weather conditions also have an important impact on radar data collection. Bad weather may cause radar equipment to fail to operate normally or data quality to deteriorate. In addition, the collection of radar data requires advanced technical equipment and professionals, and these resources are not available anytime and anywhere. Based on the above background, the amount of radar data currently available for research is relatively small, which limits the in-depth research and analysis of radar data.

[0003] On the other hand, during the flight, aerial targets will show different postures, which may lead to large differences in high-resolution range images of the same type of aerial targets in different postures. At the same time, different types of aerial targets show high similarity in the same posture, which causes trouble in the identification of aerial targets and leads to misjudgment. With the rapid development of deep learning technology, feature extraction plays an increasingly important role in the field of computer vision and pattern recognition. The quality of high-dimensional feature representation plays a vital role in data analysis and model performance. Therefore, it is of great significance to improve the accuracy of radar target recognition by studying the high-dimensional feature relationship between samples and fully capturing the similarities and differences between data. Summary of the invention

[0004] The present invention provides a radar target recognition method suitable for small sample conditions, thereby solving the problem in the prior art that small sample data causes poor generalization ability and unstable performance of the model, and realizes the adjustment of features in a high-dimensional feature space to better capture the relationship between data, better understand the similarities and differences between data, and improve the representation ability and interpretability of data, thereby providing more powerful support for data analysis and model training.

[0005] The present invention provides a radar target recognition method applicable to a small sample situation, the method comprising:

[0006] Obtain radar data to be classified and determine a radar target recognition model;

[0007] The radar target recognition model is trained using a similarity constraint method to obtain a trained radar target recognition model; wherein a similarity constraint loss function corresponding to the similarity constraint method is constructed using a similarity constraint matrix of a training sample set, and the similarity constraint method is applied to the last high-dimensional layer after feature extraction in the radar target recognition model;

[0008] The radar data to be classified is input into the trained radar target recognition model to obtain a recognition result.

[0009] In a possible implementation, the using a similarity constraint method to train the radar target recognition model to obtain a trained radar target recognition model includes:

[0010] Constructing a similarity constraint loss function and a similarity constraint matrix corresponding to the similarity constraint method;

[0011] Based on the training sample set, an optimized sample set is obtained by adopting a gradient-based optimization method; wherein each sample in the training sample set is a high-dimensional feature;

[0012] A joint loss function of the radar target recognition model is constructed according to the similarity constraint function, and the joint loss function is trained using the optimized sample set to obtain a trained radar target recognition model.

[0013] In a possible implementation, the similarity constraint loss function is expressed as:

[0014]

[0015] Where Y represents the training sample set; () T represents transposition; S represents the similarity constraint matrix; Indicates the second norm; Y T Y represents the cosine distance between samples in the sample set.

[0016] In a possible implementation, the adopting of a gradient-based optimization method to converge the similarity constraint loss function until an iteration condition is met includes:

[0017] The similarity constraint function is iteratively trained according to an iterative process; wherein the iterative process includes:

[0018] Taking a first-order derivative of the similarity constraint loss function to obtain a first-order derivative result;

[0019] According to the learning rate and the first-order derivative result, a sample estimation value corresponding to each sample in the sample set in the current training round is obtained, and according to the sample estimation value corresponding to each sample, an estimation sample set of the current round is obtained;

[0020] The loss value of the current round is calculated according to the current round estimation sample set and the similarity loss function, and it is determined whether the loss value meets the iteration condition. If so, the current round estimation sample set is used as the optimization sample set; if not, the iteration process is repeated according to the current round estimation sample set until the iteration condition is met.

[0021] In a possible implementation, the joint loss function is expressed as:

[0022] L=(1-α)L c +αL s ;

[0023] Among them, L c represents the cross entropy loss function; α represents the loss function weight factor; L s represents the similarity constraint loss function.

[0024] In a possible implementation, inputting the radar data to be classified into a trained radar target recognition model to obtain a recognition result includes:

[0025] Using the radar target recognition model to extract features from the radar data to be classified, to obtain a high-dimensional feature set;

[0026] The radar target recognition model performs similarity constraints on the high-dimensional feature set to obtain an optimal estimated high-dimensional feature set;

[0027] The radar target recognition model recognizes the radar data to be classified according to the optimal estimated high-dimensional feature set to obtain a recognition result.

[0028] One or more technical solutions provided in the present invention have at least the following technical effects or advantages:

[0029] The present invention adopts a similarity constraint method and constructs a similarity constraint matrix in a high-dimensional feature space to realize strongly constrained feature mapping, thereby achieving the goal of manipulating high-dimensional features, improving the characterization ability and interpretability of data, making full use of the relationship between samples, alleviating overfitting or underfitting, and improving the generalization ability of the model, thereby further improving the accuracy of radar target recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A flowchart of a radar target recognition method for a small sample size provided by an embodiment of the present invention;

[0031] Figure 2 A schematic diagram of a similarity constraint matrix provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] The present invention provides a radar target recognition method suitable for small sample conditions, such as Figure 1 As shown, the method includes the following steps S101 to S103.

[0034] S101, obtaining radar data to be classified and determining a radar target recognition model;

[0035] S102, training the radar target recognition model using a similarity constraint method to obtain a trained radar target recognition model; wherein, a similarity constraint loss function corresponding to the similarity constraint method is constructed using a similarity constraint matrix of a training sample set, and the similarity constraint method is applied to the last high-dimensional layer after feature extraction in the radar target recognition model;

[0036] Specifically, in step S102, the radar target recognition model is trained using a similarity constraint method to obtain a trained radar target recognition model, including the following steps S1021 to S1023.

[0037] S1021, constructing a similarity constraint loss function and a similarity constraint matrix corresponding to the similarity constraint method;

[0038] Here, the schematic diagram of the similarity constraint matrix is ​​as follows Figure 2 As shown in , the value of the intersection of the rows and columns of the same target data in the same batch is 1, indicating that the distance between the same data is very close, and the rest is 0, indicating that the distance between heterogeneous data is very far. Figure 2 In the figure, both the horizontal and vertical axes represent targets. When the target pointed to by the horizontal axis has a high similarity with the target pointed to by the vertical axis, it is represented by a yellow square in the figure. If the similarity is low, it is represented by a dark square.

[0039] The similarity constraint loss function is expressed as:

[0040]

[0041] Where Y represents the training sample set; ()T represents transposition; S represents the similarity constraint matrix; Indicates the second norm; Y T Y represents the cosine distance between samples in the sample set. T Y is used to measure the similarity of samples in high-dimensional feature space, and its size is batchsize*batchsize.

[0042] The goal is to make the similarity between samples approach the set similarity constraint matrix S, and obtain the current round estimated sample set Y* through a gradient-based alternating projection optimization method.

[0043] S1022, based on the training sample set, using a gradient-based optimization method to converge the similarity constraint loss function until an iteration condition is met, thereby obtaining an optimized sample set; wherein each sample in the training sample set is a high-dimensional feature;

[0044] Here, a gradient-based optimization method is used to converge the similarity constraint loss function until the iteration conditions are met, including:

[0045] (1) iteratively training the similarity constraint function according to an iterative process; wherein the iterative process includes:

[0046] (1.1) taking a first-order derivative of the similarity constraint loss function to obtain a first-order derivative result;

[0047] (1.2) obtaining a sample estimation value corresponding to each sample in the sample set in the current training round according to the learning rate and the first-order derivative result, and obtaining a current round estimation sample set according to the sample estimation value corresponding to each sample;

[0048] (1.3) Calculating the loss value of the current round based on the current round estimated sample set and the similarity loss function, and judging whether the loss value satisfies the iteration condition. If so, taking the current round estimated sample set as the optimized sample set; if not, repeating the iteration process based on the current round estimated sample set until the iteration condition is met.

[0049] S1023, constructing a joint loss function of the radar target recognition model according to the similarity constraint function, and training the joint loss function using the optimized sample set to obtain a trained radar target recognition model.

[0050] Here, the joint loss function is expressed as:

[0051] L=(1-α)L c +αL s ;

[0052] Among them, Lc represents the cross entropy loss function; α represents the loss function weight factor; L s represents the similarity constraint loss function.

[0053] After integrating the similarity constraint method into the neural network model, a joint loss function is constructed. The weight factor α of the joint loss function ranges from [0,1]. When α is 0, it means that the current training only uses the end-to-end deep neural network; when α is 1, it means that the current training only uses the similarity constraint method; when α is (0,1), it means that the similarity constraint method is integrated into the end-to-end deep neural network.

[0054] The present invention provides a radar target recognition method suitable for small sample conditions, which maintains a strongly constrained feature mapping in a high-dimensional space through a similarity constraint matrix, that is, the feature distances of similar targets at different angles are reduced, and the feature distances of heterogeneous targets at the same angle are increased, thereby achieving the goal of manipulating high-dimensional features.

[0055] S103, inputting the radar data to be classified into the trained radar target recognition model to obtain a recognition result.

[0056] Specifically, in step S103, the radar data to be classified is input into the trained radar target recognition model to obtain a recognition result, including the following steps S1031 to S1033.

[0057] S1031, extracting features from the radar data to be classified using a radar target recognition model to obtain a high-dimensional feature set;

[0058] S1032, the radar target recognition model performs similarity constraints on the high-dimensional feature set to obtain an optimal estimated high-dimensional feature set;

[0059] S1033, the radar target recognition model recognizes the radar data to be classified according to the optimal estimated high-dimensional feature set to obtain a recognition result.

[0060] In a specific embodiment provided by the present invention, the HRRP data used in this example are measured echo data of three types of aerial targets admitted by a certain ISAR system. The data set is first preprocessed, and the training sets of "An-26", "Award" and "Yak-42" are divided into 52, 52 and 44 frames respectively, with 1000 samples per frame. 1%, 10%, 20%, and 30% of the samples are randomly selected from each frame to form the training set of the following small sample experiment, and 15,000 samples are randomly selected from the test set to form the test set of the following small sample experiment.

[0061] In order to avoid the uncertainty of the results caused by random sampling, this example conducted 5 repeated experiments and averaged the experimental results to obtain a more convincing conclusion.

[0062] The following analyzes the impact of the similarity constraint method on HRRP recognition performance from the perspective of CNN (Convolution Neural Network).

[0063] In the small sample experiment based on the CNN model, it can be seen from Table 1 that adding the similarity constraint method to the CNN model has a certain improvement in the recognition accuracy of HRRP, and when the training set is 1% of the sample, the recognition accuracy is improved most significantly, by 2.45%. This shows that the similarity constraint method is a radar target recognition method suitable for small sample conditions.

[0064] Table 1. Small sample ablation experiment based on CNN model

[0065] Training set size Using only CNN models Add similarity constraint method 1％ 82.70％ 85.15％ 10％ 92.08％ 93.11％ 20％ 92.82％ 93.42％ 30％ 91.93％ 93.51％

[0066] The present invention aims to manipulate high-dimensional features through similarity constraints. From the perspective of manifold learning, there will be a manifold space structural relationship in high dimensions. The present invention utilizes this part of information through similarity constraints and adjusts the features in the high-dimensional feature space to better capture the relationship between data, better understand the similarities and differences between data, improve the representation ability and interpretability of data, and thus provide more powerful support for data analysis and model training. Even for small sample data sets, this method combined with an end-to-end deep neural network can achieve good radar target recognition accuracy.

[0067] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. All or part of the present invention can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, mobile communication terminals, multi-processor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0068] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the present invention.

Claims

1. A radar target recognition method suitable for small sample conditions, characterized in that: include: Obtain radar data to be classified and determine a radar target recognition model; The radar target recognition model is trained using a similarity constraint method to obtain a trained radar target recognition model; wherein a similarity constraint loss function corresponding to the similarity constraint method is constructed using a similarity constraint matrix of a training sample set, and the similarity constraint method is applied to the last high-dimensional layer after feature extraction in the radar target recognition model; The radar data to be classified is input into the trained radar target recognition model to obtain a recognition result.

2. The radar target recognition method applicable to small sample conditions according to claim 1, characterized in that: The method of training the radar target recognition model by using the similarity constraint method to obtain a trained radar target recognition model includes: Constructing a similarity constraint loss function and a similarity constraint matrix corresponding to the similarity constraint method; Based on the training sample set, an optimized sample set is obtained by adopting a gradient-based optimization method; wherein each sample in the training sample set is a high-dimensional feature; A joint loss function of the radar target recognition model is constructed according to the similarity constraint function, and the joint loss function is trained using the optimized sample set to obtain a trained radar target recognition model.

3. The radar target recognition method applicable to small sample conditions according to claim 2, characterized in that: The similarity constraint loss function is expressed as: Where Y represents the training sample set; () T represents transposition; S represents the similarity constraint matrix; Indicates the second norm; Y T Y represents the cosine distance between samples in the sample set.

4. The radar target recognition method applicable to small sample conditions according to claim 2, characterized in that: The gradient-based optimization method is used to converge the similarity constraint loss function until an iteration condition is met, including: The similarity constraint function is iteratively trained according to an iterative process; wherein the iterative process includes: Taking a first-order derivative of the similarity constraint loss function to obtain a first-order derivative result; According to the learning rate and the first-order derivative result, a sample estimation value corresponding to each sample in the sample set in the current training round is obtained, and according to the sample estimation value corresponding to each sample, an estimation sample set of the current round is obtained; The loss value of the current round is calculated according to the current round estimation sample set and the similarity loss function, and it is determined whether the loss value meets the iteration condition. If so, the current round estimation sample set is used as the optimization sample set; if not, the iteration process is repeated according to the current round estimation sample set until the iteration condition is met.

5. The radar target recognition method applicable to small sample conditions according to claim 2, characterized in that: The joint loss function is expressed as: L=(1-α)L c +αL s ; Among them, L c represents the cross entropy loss function; α represents the loss function weight factor; L s represents the similarity constraint loss function.

6. The radar target recognition method applicable to small sample conditions according to claim 1, characterized in that: The step of inputting the radar data to be classified into a trained radar target recognition model to obtain a recognition result includes: Using the radar target recognition model to extract features from the radar data to be classified, to obtain a high-dimensional feature set; The radar target recognition model performs similarity constraints on the high-dimensional feature set to obtain an optimal estimated high-dimensional feature set; The radar target recognition model recognizes the radar data to be classified according to the optimal estimated high-dimensional feature set to obtain a recognition result.

Images