CN-119780867-B - Radar target recognition method suitable for small sample condition

CN119780867BCN 119780867 BCN119780867 BCN 119780867BCN-119780867-B

Abstract

The invention discloses a radar target identification method suitable for a small sample condition, which solves the problems of poor generalization capability and unstable performance of a model caused by small sample data in the prior art, and comprises the steps of obtaining radar data to be classified; the method comprises the steps of training a 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 a high-dimensional last layer after feature extraction in the radar target recognition model, radar data to be classified is input into the trained radar target recognition model to obtain a recognition result, and the method realizes adjustment of features in a high-dimensional feature space to better capture the relation between the data, so that the similarity and the difference between the data can be better understood, and the representation capability and the interpretability of the data are improved, thereby providing more powerful support for data analysis and model training.

Inventors

DING JUN
ZHANG LI
WANG PENGHUI
LIU HONGWEI
CHEN BO

Assignees

西安电子科技大学

Dates

Publication Date: 20260512
Application Date: 20241231

Claims (3)

1. A radar target recognition method suitable for use in a small sample situation, comprising: Acquiring radar data to be classified, and determining a radar target recognition model; Training the radar target recognition model by using a similarity constraint method to obtain a trained radar target recognition model, constructing a similarity constraint loss function corresponding to the similarity constraint method by using a similarity constraint matrix of a training sample set, and applying the similarity constraint method to a high-dimensional last layer after feature extraction in the radar target recognition model, wherein: The method for training the radar target recognition model by using the similarity constraint method to obtain a trained radar target recognition model comprises the following steps: Constructing a similarity constraint loss function and a similarity constraint matrix corresponding to the similarity constraint method, wherein the similarity constraint loss function is expressed as: Wherein, the method comprises the steps of, Representing a training sample set; Representation transposition; representing a similarity constraint matrix; representing the binary norms; representing cosine distances between samples in the sample set; Obtaining an optimized sample set by adopting a gradient-based optimization mode based on the training sample set, wherein each sample in the training sample set is a high-dimensional characteristic; Constructing a joint loss function of the radar target recognition model according to the similarity constraint function, and training the joint loss function by utilizing the optimized sample set to obtain a trained radar target recognition model, wherein the joint loss function is expressed as: ; Wherein, the Representing a cross entropy loss function; Representing a loss function weighting factor; Representing a similarity constraint loss function; and inputting the radar data to be classified into the radar target recognition model after training is completed, and obtaining a recognition result.
2. The method for radar target identification in small sample cases according to claim 1, wherein the converging the similarity constraint loss function by using a gradient-based optimization method until an iteration condition is reached, includes: and carrying out iterative training on the similarity constraint function according to an iterative process, wherein the iterative process comprises the following steps: performing first-order derivation on the similarity constraint loss function to obtain a first-order derivation result; 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; And calculating a loss value of the current round according to the current round estimation sample set and the similarity loss function, judging whether the loss value meets an iteration condition, if so, taking the current round estimation sample set as an optimized sample set, and if not, repeating the iteration process according to the current round estimation sample set until the iteration condition is reached.
3. The method for radar target recognition in the case of small samples according to claim 1, wherein the inputting the radar data to be classified into the trained radar target recognition model to obtain the recognition result includes: extracting features of the radar data to be classified by using the radar target recognition model to obtain Gao Weite collection; the radar target recognition model carries out similarity constraint on the high-dimensional feature set to obtain an optimal estimation Gao Weite sign set; and the radar target recognition model recognizes the radar data to be classified according to the optimal estimation Gao Weite symptom to obtain a recognition result.

Description

Radar target recognition method suitable for small sample condition Technical Field The invention relates to the technical field of radar target recognition, in particular to a radar target recognition method suitable for a small sample condition. Background The difficulty in the field acquisition of radar data is mainly due to a number of factors. First, radar devices often need to be deployed in specific geographic locations, while some areas may be difficult to reach due to complex terrain or harsh environments, resulting in limited data acquisition. Secondly, weather conditions also have an important influence on radar data acquisition, and severe weather may cause the radar device to fail to operate normally or the data quality to be degraded. Furthermore, the acquisition of radar data requires advanced technical equipment and professionals, and these resources are not available anytime and anywhere. Based on the background, the amount of radar data currently available for research is relatively small, limiting the in-depth research and analysis of radar data. On the other hand, during flight, the aerial targets may exhibit different poses, which may result in a large difference in high-resolution range profile of the same type of aerial target at different poses. Meanwhile, different types of aerial targets show higher similarity under the same gesture, which causes trouble to the identification of the aerial targets and leads to erroneous judgment. With the rapid development of deep learning technology, feature extraction plays an increasingly important role in the fields of computer vision and pattern recognition, and the quality of high-dimensional feature representation plays a vital role in data analysis and model performance. Therefore, by researching the high-dimensional characteristic relation among samples, the similarity and the difference among the data are fully captured, and the method has important significance for improving the accuracy of radar target identification. Disclosure of Invention The radar target recognition method suitable for the condition of the small sample solves the problems of poor generalization capability and unstable performance of a model caused by the small sample data in the prior art, realizes the adjustment of the characteristics in a high-dimensional characteristic space so as to better capture the relation between the data, can better understand the similarity and the difference between the data, and improves the representation capability and the interpretability of the data, thereby providing more powerful support for data analysis and model training. The invention provides a radar target identification method suitable for a small sample condition, which comprises the following steps: Acquiring radar data to be classified, and determining a radar target recognition model; Training the radar target recognition model by using a similarity constraint method to obtain a trained radar target recognition model, constructing a similarity constraint loss function corresponding to the similarity constraint method by using a similarity constraint matrix of a training sample set, and applying the similarity constraint method to a high-dimensional last layer after feature extraction in the radar target recognition model; and inputting the radar data to be classified into the radar target recognition model after training is completed, and obtaining a recognition result. In one possible implementation manner, the training the radar target recognition model by using the similarity constraint method to obtain a radar target recognition model after training includes: Constructing a similarity constraint loss function and a similarity constraint matrix corresponding to the similarity constraint method; Obtaining an optimized sample set by adopting a gradient-based optimization mode based on the training sample set, wherein each sample in the training sample set is a high-dimensional characteristic; Constructing a joint loss function of the radar target recognition model according to the similarity constraint function, and training the joint loss function by utilizing the optimized sample set to obtain a trained radar target recognition model. In one possible implementation, the similarity constraint loss function is expressed as: wherein Y represents a training sample set, () T represents a transpose, S represents a similarity constraint matrix; Representing the binary norm and Y T Y represents the cosine distance between the samples in the sample set. . In one possible implementation manner, the converging the similarity constraint loss function by adopting a gradient-based optimization manner until an iteration condition is reached includes: and carrying out iterative training on the similarity constraint function according to an iterative process, wherein the iterative process comprises the following steps: performing first-order derivation on the similarity constraint l