The High Range Resolution Radar Target Recognition, Feature Extraction Research

Some specific techniques for ATR by a high range resolution radar are researched in feature extraction, which are parts of a project-"Search into Method for HRR Radar Target Feature Extraction Based on Geometric Theory of Diffraction" supported by National Science Foundation of China.The main content is as follows.Chapter 2 discusses the simple scattering-centers model at first. Then both the RELAX algorithm and the revised MRELAX algorithm are employed computing the outfield Radar data in order to compare the feature parameters gained from that. Simulation tests show that the new features gained from those experiments based on the MRELAX algorithm are more concise and clearer than that based on the RELAX algorithm.A fault of the Adaptive Gaussian Classifier used in ATR is uncovered and the influence of the magnitude of radar target range profiles on the classifier based on the Baseline method is studied in chapter 3. During that a novel conception of the factor for magnitude registration is introduced with its formulation presented by the optimization. The computer simulation results show that employing the factor apparently reduces the matching error and improve the performance of the Baseline method.Chapter 4 is in a search for a new way to parametric feature extraction. An idea based on the GTD model is discussed in details on employing matching pursuit algorithm to estimate the parameters of high range resolution radar target. Procedures of feature extraction based on Gabor dictionaries are also given. Simulation result show that the new method is valid, however, there are still much more work left to do in the future about the parameter a related to each scattering center in each range bin.