基于互补特征层次决策融合的SAR目标识别方法

SAR Target Recognition Based on Hierarchical Decision Fusion of Complementary Features

提出基于互补特征层次决策融合的合成孔径雷达(SAR)目标识别方法,该方法采用主成分分析特征、目标峰值和目标轮廓作为描述SAR图像的特征。三者对于目标的描述具有较强的互补性,从而为目标识别提供更多的鉴别力信息。在决策融合阶段,采用了层次推进的策略。第一级采用主成分分析特征,第二级采用峰值特征,第三级采用轮廓特征进行识别。当前一级得到可靠的识别结果时,下一级则无需进行。采用层次融合策略,大大提高了目标识别的效率,避免了不必要的重复识别过程。为了验证所提方法的有效性,基于MSTAR公共数据集进行了目标识别实验。

This paper proposes a Synthetic Aperture Radar （SAR） target recognition method based on hierarchical decision fusion of complementary features. The Principal Component Analysis （PCA） feature, target peaks and contour are employed as basic features for target recognition, which are complementary with each other and can provide more comprehensive target descriptions. At the stage of decision fusion, the hierarchical decision fusion strategy is used. For target recognition, the first level employs the PCA feature, the second level uses the peaks, and the third level adopts the contour. When the former level recognizes the target with high reliability, the whole recognition process ends and no more classification is needed. Therefore, the efficiency of the SAR target recognition system is significantly improved by avoiding the unnecessary recognition process. To validate the effectiveness of the proposed method, experiments are conducted on public Moving and Stationary Target Acquisition and Recognition （MSTAR） dataset.