Study Of Radar Target Recognition Based On Continual Learning

In the context of today's information age,the function of radar has been continuously improved and has been widely used in various fields.Radar automatic target recognition is a very important technical method in the radar field and has received extensive attention in current research.The early recognition of radar targets uses traditional machine learning methods and cannot be used to mine deep-level feature representations of the target,leading to poor robustness of the method.In recent years,deep learning has become a very popular topic.Deep learning is a multi-layered learning method that transforms raw data into higher-level,more abstract representations.Therefore,it is possible to learn to mine deeper features in the data,thereby improving the reliability and robustness of the target recognition.Deep learning is widely used in various fields because of its superior performance.This paper applies deep learning method to radar target recognition,and enhances the recognition performance of radar targets through its powerful learning ability.With the continuous observation of targets by radars,the amount of data continues to increase.In the context of such large-scale data,it will be too difficult to update the model in real time.This paper will propose a continual learning algorithm to update the model in real time.This paper first introduces the research background and development trend of radar target recognition and deep learning.The main content introduces the traditional radar target recognition method and radar target recognition method based on deep learning.Finally,the radar target recognition method based on continuous learning is introduced.The main content of this article is as follows:1.This paper studies the traditional radar target recognition method.Firstly,the data preprocessing method which is very important in radar target recognition is introduced.Then we introduced several commonly used classifiers in the traditional radar target recognition.Finally,the performance of each classifier in radar target recognition is verified through experiments,and its advantages and disadvantages are summarized and analyzed.2.This paper studies the radar target recognition method based on deep learning.Deep learning uses its powerful non-linear representation ability to understand the deep information of the target so as to achieve good classification performance.This paper uses the convolutional neural network method which is widely used in deep learning to automatically extract the features in the radar target.First of all,we through the introduction of the convolutional neural network structure to understand its working principle.Then,several classic convolutional neural network recognition models are introduced,and the differences between them are compared and analyzed.Finally,the performance of convolutional neural network in radar target recognition is verified by experiments.3.This paper studies a radar target recognition method based on continual learning.In order to solve the problem of real-time updating of model under large-scale data background,this paper adopts a continual learning method based on EWC to realize the real-time updating of model.And introduced a technology to reduce the amount of model training data,known as the coreset of methods,the coreset is a representative sample of the original data.Finally,the performance of the model based on the EWC and the coreset method in the continual learning of the model is verified through experiments.