SAR Image Target Recognition Based On Random Measurements And Mixture Factor Analyzers

Synthetic Aperture Radar (SAR) is one of the active microwave remote sensing equipment with high resolution and all-day work. The technology of target recognition for SAR plays an important role in national defense. However, As an ultra-wideband application, SAR is limited by Nyquist sample theory, and need too high sample rate and too large data volume, so SAR data cannot be tackled quickly by recent algorithm. The new theory of compressive sensing has recently been developed, which indicates that sparse or compressible signals can be recovered with overwhelming probability by less measurements than what Nyquist requires.In this paper, this theory is applied to SAR target recognition field. As we know that the data in MSTAR database lies in the manifold, three innovative points are represented as follows:1. we utilize Mixture factor analysis(MFA) to model the manifold data, a dictionary describing target characters is constructed by MFA parameters which derives a nonparametric method, and then place the framework of sparse representation for the final classification. The experiment result shows a good performance. The algorithm don’t need complicate preprocess for MSTAR data which has different rotated directions and depression angles, so the whole framework is simplified, and is robost to noises.2. Then for the compressive sensing RADAR system, this paper adopts a recognition algorithm based on measurements directly. These measurements are modeled by discovering the information hiding in them for constructing a dictionary to describe the characters on every class exactly. So those measurements for testing can be represented sparsely under this dictionary. After that, those measurements are recognized by a sparse representation classifier, and the experiment shows a good performance for classification.3. At last to satisfy the need of military that the targets can be recognized as fast as possible, a fast recognition method is adopted based on the formal algorithm for the compressive sensing RADAR system. Instead of the time-consuming process of iteration on the formal method, another reconstruction algorithm without iteration is adopted, then the optimal resolution of the reconstruction problem is obtained combining QR decomposition. The speed of operation is increased.This work was supported in part by the National Natural Science Foundation of China (No.60971128,61072106,61173090); the Fundamental Research Funds for the Central Universities (No. JY10000902001).