| With the continuing improvement of modern Radar system, Radar Automatic Target Recognition has come into being and developed, and many new methods were proposed for the increase of the Radar range resolution. Meanwhile, the rapid development of digital technology let ATR come into projects. In this paper, we have proposed some valid methods to solve some problems in ATR field and simulate them based on a lot of measured and simulated data.With low range resolution Radar echo, we just can classify the targets probably and can't get the detail information of the targets, but this is also important to ATR, it can simplify the Radar target recognition system. And we can classify and recognize the group targets precisely with high range resolution Radar echo. In fact, the combination of these two can improve the performance of the system.From the viewpoint of the range resolution, we divide this paper into two main parts: one is the ATR using low range resolution echo. In this part, the characteristic of the fix-wing plane and helicopter and the difference between them are discussed, and two very good feature-abstracting methods are presented, and the simulated classification experiment with Support Vector Machine is finished. The other is the ATR using high range resolution echo, which is the main part of this paper. The characteristics of the high range resolution profile and some methods of feature abstraction and classification in common use are presented in the part. Here we propose the low-filtering thought, namely, there are two methods in detail: direct low filtering in frequency and Mallat tower decomposition algorithm, to decrease the dimensions of the feature vectors. At the same time, the principles of these methods and the simulatedresults are given. At last, a new kernel function classifier training algorithm--KernelMatching Pursuit is discussed, and its advantage is analyzed in performance and operation complexity with the ISAR data.
|
PDF Full Text Request