Research On Sparse Representation Based Radio Frequency Interference Suppression

Ultra wide band synthetic aperture radar (UWB-SAR) operating at VHF/UHF bandcan penetrate the foliage and the surface of the earth to detect the concealed targets. Ithas been widely used in military detection. However, this band is heavily occupied byradio, wireless television, mobie communication and different kinds of dedicatedmicrowave signals, which constitute the RFI to the UWB-SAR. The existence of RFIseverely degrades the image quality and target detection performance of the UWB-SAR.RFI suppression approaches are divided into parametric and nonparametricmethods. Nonparametric methods have low computational complexity and are easy toimplement, which is why they are so widely used in real time UWB-SAR systems,while parameter methods have better RFI suppression performance. With thedevelopment of the electronic and information technique, the electromagneticenvironment is becoming more and more complicated and the RFI in the air presents anobvious nonstationarity. Invariable coefficients sinusoidal wave and autoregressivemodels cannot present nonstationary signals, and variable coefficients models are tooinvoluted to apply to practical data processing. In this thesis, a sparse representationbased RFI suppression method is proposed. It can estimate the components andcoefficients through sparse decomposition algorithms based on selected space so as torepresent nonstationary RFI more accurately. Based on sparse representation theory, themajor work is emboded in three aspects as following:First, analyze the time-frequency distribution of the real RFI by adaptive Gaborexpansion. The UWB-SAR has a long operating time which results a large number ofsamples, dealing with such a large number of data, the traditional MP algorithm is timecostly. Therefore, a refined time-frequency analysis technique with an adaptive Gaborsubdictionary baded MP algorithm is proposed which speed the time-frequency analysisgreatly. The time-frequency analysis reveals the high nonstationarity of the real RFI.Second, propose a RFI suppression approach based on sparse decomposition.Traditional sinusoidal wave model cannot represent time-variant signal accurately, asthere are plenty of atoms in the Gabor reluctant dictionary, this thesis obtain a moreaccurate RFI representation and better performance of RFI suppression by the adaptiveGabor subdictionary baded MP algorithm. And experiment results validate theperformance of the algorithm.Third, realize the target parameter estimation of the sparse scene with RFIsuppression based on compressed sensing(CS). If the RFI and targets are sparse in thescene, we can recover the reflectivity of the targets with RFI suppression by the CStechnique. A two stage OMP algorithm is proposed to suppress the RFI and get therecovery of target reflectivity. Its computation efficiency is one time faster than multi-component dictionary method. A CS imaging frame is given based on the CSbased RFI suppression algorithm. Simulation outcomes demonstrate the validity of CSbased RFI suppression approach and imaging frame.