Research On Radar Target Recognition

Radar,as a tool of long-range target detection and tracking and location,plays a very important role in human exploration of the surrounding environment.Radar has a very wide range of applications,and has made remarkable achievements in both civil and military fields.Radar target recognition technology is a technology to distinguish the attributes and characteristics of targets by scattering echo analysis.Because of the different types of radar,the detection results may vary greatly,which leads to different branches of research and development direction.Among many branches,the recognition of high-frequency areas is the most valuable,and high-resolution radar mining is the most valuable one.Using broadband signal transmission can obtain high resolution of target in high frequency region,so that imaging of target can be realized directly in radial distance,thus displaying the physical structure of different targets and distinguishing target attributes.In this paper,through the research of radar target recognition technology at home and abroad,the current problems and bottlenecks in the field of radar target recognition are analyzed,the preprocessing method of radar target recognition and the feature extraction technology of radar target signal are studied,and then the classification algorithm of radar target recognition is introduced.The preprocessing methods introduce non-stationary signal,short time Fourier transform and wavelet transform respectively.Feature extraction mainly introduces feature extraction of HRRP target signal based on geometric structure feature,PCA-based feature extraction of target signal and LDA-based feature extraction of target signal.Radar target recognition algorithm focuses on introducing the pattern algorithm based on artificial neural network.Then,the potential geometric structure features of target in time domain echo of high resolution range profile are analyzed.By using statistical method,eight feature quantities reflecting geometric structure information of target from different angles are extracted from time domain echo of high resolution range profile.This paper adopts the idea of multi-feature synthesis and chooses several features to combine to get eight comprehensive features.Through experiments,the principal component analysis method is used to analyze the data of different feature selection of three aircraft,and the results of single feature recognition and comprehensive feature recognition are compared respectively.The experimental results show some of them.Finally,it is found that the effect of comprehensive feature recognition is much better than that of single feature recognition.By selecting three aircrafts for experiment and measuring HRRP data,the principal component analysis(PCA)method in statistics and the comparative analysis of single feature recognition and comprehensive feature recognition are adopted.It is found that the HRRP time domain echoes of radar targets contain abundant features related to geometric structures,such as target size and scattering.It is very important to extract and use these features reasonably for target classification and recognition,such as center position,distribution structure and intensity.In the test experiments of three aircraft,the equivalent scattering center dimension,equivalent target size,entropy,standard deviation,deviation,echo power,irregularity and P5 are selected for single feature analysis,and then the eight different feature terms are fused and recombined through different combinations to carry out comprehensive feature analysis.The experimental results show that entropy and irregularity have the best recognition effect in single feature analysis,and the overall effect of comprehensive feature recognition is obviously better than that of single feature recognition.