Wideband Electromagnetic Scattering Analysis And Recognition Technique Of Targets

Due to the pressing requirements of information warfare, radar automatic target recognition (RATR) technique has attracted tremendous interests. The radar targets usually adopt stealth and low altitude penetration techniques in modern warfare, which presents new challenges to the radar automatic target recognition technique. Since the wideband radar is considered to be the effective equipment for detecting and identifying these targets, it is in urgent demands to carry out the study of wideband electromagnetic scattering analysis and wideband radar automatic target recognition. This dissertation is supported by the Advanced Defense Research Programs of China and the National Natural Science Foundations of China. The wideband electromagnetic scattering characteristics analysis of radar targets, wideband radar automatic target recognition and superresolution techniques for wideband radar are studied.The marching-on-in-degree (MOD) solver of time domain integral equation is studied for the analysis of transient electromagnetic scattering from perfect electric conductor (PEC). Since many dense matrices need to be stored in this method, the memory requirement and CPU time are very high. The adaptive cross approximation algorithm with singular value decomposition postcompression (ACA-SVD) is introduced into the MOD solver of time domain integral equation. The impedance matrices associated with the well-separated groups are compressed by utilizing their low-rank property. Thus the proposed method can accelerate the traditional MOD method and greatly reduce the memory requirement and matrix-vector product (MVP) time per iteration.The time domain finite element-boundary integral (TD-FEBI) method is studied for the analysis of transient electromagnetic scattering from mixed metal and complex dielectric objects. Since the MOD method has the advantage of unconditionally stability in the late time, the MOD solver of TD-FEBI method is studied in this dissertation. The MOD equations for the finite element part and boundary integral part are derived, repectively. The iterative method and hybrid method are given to solve the matrix equation. In such a manner, the MOD solver of TD-FEBI method is implemented, which provides a new way for stably solving the TD-FEBI equation.The E-pulse based RATR technique is studied. The problem of traditional E-pulse technique is pointed out. A multi-directional E-pulse technique is proposed, which constructs multiple E-pulses for different angle regions of a target. The specific implementation procedure of multi-directional E-pulse technique is demonstrated. The proposed method can dramatically improve the recognition rate of E-pulse based RATR technique.The high range resolution profile (HRRP) based RATR technique is studied. The acquisition method of HRRP is introduced. The feature extraction method based on manifold learning is focused. The existing manifold learning algorithms have two problems, one is not making the best use of the class information, and the other is not considering different contributions to the effect of classification from different training samples. To solve these problems, an adaptive neighborhood preserving discriminant projection method is proposed, which can effectively improve the recognition rate of the HRRP based RATR technique.A superresolution technique for wideband radar based on fast sparse Bayesian learning algorithm is studied. The superresolution problem is transformed to sparse representation problem. Then the fast sparse Bayesian learning algorithm is utilized to solve this problem. A multilevel dynamic dictionary is proposed to improve the accuracy and accelerate the computation. A coherent processing algorithm for multi-band radar data from multiple spatially collocated radars is also studied. The proposed method can effectively fuse the multi-band radar data of wideband radars, thus improve the range resolution.