Noise Reduction Method And Subspace Based Noise-robust Recognition Method For Radar HRRP Data

High Resolution Range Profile(HRRP)has received intensive attention from radar automatic target recognition(RATR)field.However,due to some severe measurement conditions,the high Signal-Noise-Ratio(SNR)condition cannot be guaranteed for the measured data.Thus noise robustness becomes an important job in RART field.The existing studies of noise-robust target recognition methods based on HRRP are mostly focused on modification for statistical model.Studies on signal denoising and noise-robust feature extraction are respectively fewer.In this dissertation,we propose a new method on denoising and a new recognition method based on extracting noise-robust feature,both improve the robustness of recognition under low SNR.The main content is as follows.Firstly,we introduce the traditional Compressive Sensing algorithms,which are used to reconstruct signals of high SNR.We find the fact that,due to the influence of external noise,the sparsity of signal is difficult to determine,which leads to the poor performance of traditional Compressive Sensing under low SNR condition.Aiming at the problem of signal denoising for HRRP,we deduce the Bayesian statistical model of HRRP echoes from the target scattering center model theory and propose an improved denoising method based on Bayesian Sparse Decomposition by using the estimated noise power as noise prior and making it fixed,which converts the denoising problem under low SNR condition to reconstruction problem under high SNR condition.In the meantime,we also realize the fast algorithm of Bayesian Sparse Decomposition.Experimental results on measured radar HRRP dataset validate the effective noise reduction and noise robustness of the proposed method.Secondly,we introduce the Minimum Reconstruction Error method based on Principal Component Analysis.By the definition of the distance between subspaces of the same dimension or different dimensions,we then propose a recognition method for HRRP data and analyse the noise robustness of it.Compared with the Minimum Reconstruction Error method,the proposed algorithm has higher recognition accuracy and is more robust to the test noise environment.Experimental results on measured radar HRRP dataset validate the effectiveness of the proposed method.