SAR target recognition based on spectrum feature of optimal Gabor transform

In order to improve synthetic aperture radar(SAR) target recognition´s accuracy, a novel SAR automation recognition(ATR) system is proposed in this paper. The method combines spectrum feature of time-frequency with independent components analysis(ICA) to improve the performance in SAR target recognition. Firstly, according to time frequency band product(TBP), we design a optimal window for Gabor transform(GT) which can improve time frequency spectrum´s aggregation. Secondly, extract optimal time-frequency spectrum´s peak of energy characteristic and get a feature representation of the original image. At last, we utilizing the ICA to process the feature representation and establish training model through support vector machine(SVM). Test images are taken from the MSTAR database. The simulation results shows that the proposed algorithm has a good performance in high accuracy SAR target recognition.