Automatic Building Detection for Multi-Aspect SAR Images Based on the Variation Features

Multi-aspect synthetic aperture radar (SAR) images contain more information available for automatic target recognition (ATR) than from a single view. However, the sensitivity to aspect angles also makes it hard extract and integrate multi-aspect images. In this paper, we propose novel method based on variations features realize building detection in image level. First, get comprehensive description of variation patterns, statistical characteristic variances are derived three representative complementary categories. Then, these obtained fused put K-means classifier prescreening, whose results used as training sets supervised classification later avoid manual labeling. Second, precise performance, finer vector forms by principal component analysis (PCA). The patterns feature vectors explored two different manners correlation fluctuation analyses processed separate support machines (SVMs) after fusion. Finally, independent SVM according maximum probability rule. Experiments conducted airborne data demonstrate robustness effectiveness proposed method, spite significant signature variabilities cluttered background.