Research On Radar Target Recognition In Ballistic Midcourse

Radar Target Recognition(RTR)is one of the key technologies for Ballistic Missile Defense(BMD).With the development of modern military,the research on the radar target recognition method of midcourse ballistic missile has become a hot topic at home and abroad.Based on the ballistic missile target motion modeling and echo modeling,this paper studies the radar target recognition method based on one-dimensional range image,micro-Doppler,and the performance evaluation method of classifier recognition based on confidence.1.Studied the basic characteristics of the middle ballistic target.First,the echo characteristic of the mid-range target is set up,and the method of acquiring the one-dimensional range image being analyzed.Then,a micro-motion model is established,with the micro-Doppler effect of the micromotion target analyzed.It lays the foundation for subsequent mid-range target recognition.2.An object recognition method based on the Adaptive Evolution Particle Swarm Optimization(AEPSO)optimized Support Vector Machine(SVM)is proposed,for the recognition of onedimensional distance image.This method first performs preprocessing and feature extraction on the data of the target;then,uses AEPSO algorithm to optimize the parameters of the SVM with the recognition accuracy of the training set as a fitness function;finally,adopts the optimized SVM to classifying the test samples,and the recognition accuracy up to 97%.3.A micro-Doppler feature extraction method based on time-frequency rearrangement with improved Dijkstra algorithm is proposed,for the recognition of true and false warhead targets.This method first uses time-frequency rearrangement to perform time-frequency analysis on echo signals,then uses the improved Dijkstra algorithm to perform micro-Doppler extraction of time-frequency maps,and finally uses micro-Doppler features under different micro-movements for recognition,and the recognition accuracy up to 96%.4.Based on the confidence theory in statistical learning,an SVM confidence estimation method based on Adaptive Neighborhood Samples(ANS)is proposed.This method estimates the confidence level of indivial samples according to adaptively selecting the neighborhood of the test;ranks the individuals confidence degrees and combines the rejection rate to give the rejection domain.The sample recognition rate after rejection is used as the overall confidence estimate,which the effect is better.