Target Electromagnetic Feature Extraction And Identification For Ultra Wide-band Radar

Recently, the ultra wide-band radar, as a new system, has been an important development direction of the modern radar system. With the new system, radar target and environment scattering information will be excited more completely in frequency domain. The scattering characteristics in Rayleigh, resonance and optics domains can be measured together and more complete electromagnetic information of radar target and environment can be extracted. The theory investigation of feature extraction and target identification of ultra wide-band radar is of much theoretical and applicable significance in the exploitation of novel mechanisms and methods for target identification, as well as the improvement of the target identification performance in the precise guidance weapon and battlefield surveillance system.In this thesis the theory and technique of conducting target identification based on the new radar system are investigated, and main works are accomplished as follows:Firstly, the feature extraction methods based on the mathematics models of the conducting target's ultra wide-band scattering echo are investigated. The mathematics models of the whole scattering echo, late-time and early-time components are analyzed respectively in this thesis. A uniform scattering model and exact expression are given respectively for the whole scattering echo and its late-time component. Then, in view of the difficult partition to the late- and early-time component of the scattering echo, through an ARMA representation to approximate the whole target scattering echo a pole feature extraction method is proposed with SVD and TLS (Total least square) techniques, which improves the pole extraction precision. Subsequently, based on the uniform scattering model, the characteristic analysis of UWBR conducting target scattering center and feature extraction are investigated respectively. An extraction method of GTD model parameter is proposed, and then based on the proposed method a physical phenomenon that there exists a main scattering component among complicated scattering behaviors of the scattering center of conducting target illuminated by ultra wide-band radar is indicated in the experimental manner. Furthermore, the feasibilities of GTD model for main scattering component extracting and the whole scattering echo modeling are analyzed respectively.Secondly, aiming at the deficiencies in feature extraction based on the model, the construction and extraction techniques based on the sparse property of the whole scattering echo are investigated for the feature without the model restriction. The signal sparse representation with the undercomplete atom dictionary is defined, the binary partition capability of the undercomplete atom dictionary to the signal space is theoretically proved and several important properties of the undercomplete atom dictionary are given, as well as a training algorithm for the undercomplete atom dictionary with least decomposition residual energy. Furthermore, a novel feature for target identification, named as atom dictionary feature is proposed and validated, including an extraction algorithm for the atom dictionary feature and a search algorithm for atom number optimization of the atom dictionary feature.Thirdly, pole feature-based target identification using late-time component of the UWBR conducting target scattering echo is mainly investigated. The cross correlation characteristic of the target late-time component is theoretically analyzed, the condition not producing the new pole component is concluded, and then a new cross-correlation estimator is proposed. According to the new cross-correlation estimator, aiming at the deficiencies in the asymptotically unbiased E-pulse target identification method proposed by Blanco, a method named as MAUAE is proposed. Subsequently, with a view to the deficiencies in the MAUAE method, the superiority and shortcoming of the BE-pulse scheme are theoretically analyzed. The modified BE-pulse identification scheme, engineering realization technique of the E-pulse waveform, and the fast target identification method named as BEGLRT are proposed in succession. As a result, the pole feature-based theory and system sechmes for target identification are established, which provide the pole feature-based target identification in low SNR with the theory basis and elementary engineering realization scheme.Finally, in view of the deficiencies in target identification based on late-time component, the whole scattering echo-based target identification is investigated with the proposed scattering center and atom dictionary features for ultra wide-band radar. Using the likelihood ratio test and matching pursuit technique, the echo reconstruction and target identification methods with the GTD model parameters of the main scattering component of the scattering center and atom dictionary from chapter 2 and 3 as features are proposed in this thesis respectively. Furthermore, based on the calculated and measured transient response by author, the proposed echo reconstruction and target identification methods are validated.