Narrow Bandwidth Radar Target Recognition Research Based On Deep Learning

Radar target detection and recognition is an important task in radar signal processing,and it plays a very important role in security fields such as detection and monitoring.At this stage,the technology of narrowband radar is mature and the cost is relatively low.In many scenarios,it still cannot be replaced by wideband high resolution radar.Narrowband radar echo itself cannot carry information such as the shape and structure of the target for target detection and recognition.Traditional target detection and recognition methods have problems such as weak environmental adaptability and a sharp increase in the number of false alarms under complex terrain.Therefore,studying the target detection and recognition of narrowband radar has important application significance.This article first introduces the traditional radar target detection and recognition algorithms:CFAR and SVM technology,analyzes the problems and limitations of the current technology,and as a comparison algorithm,compares and compares the detection and recognition results of the model designed.Aiming at the problems of narrow-band radar target detection and recognition,deep learning technology widely used in text,speech and image fields is studied,and the feasibility of introducing deep learning technology into radar target detection and recognition is analyzed.After that,using the range-Doppler image obtained by the radar echo signal processing as the data set,the radar target detection model based on the YOLOv5 network and the SSD network were designed and trained.The test results show that the built detection model Has a good detection rate and false alarm rate,Aiming at the problem of weak generalization ability of the target detection model based on convolutional neural network,finally a data set is constructed based on the range-dimensional demodulation data inputted by MTI processing in signal processing.According to the char acteristics of the radar one-dimensional sequence signal,a method based on gated recurrent neural network-LSTM target recognition model has completed the]recognition task.The results show that the LSTM model can achieve the integration of detection and r ecognition,and can achieve effective target recognition.