提出了一种基于多方位角图像决策融合的合成孔径雷达(SAR)图像目标识别方法。对目标切片图像用二维小波分解和主成分分析提取特征向量, 利用支持向量机对特征向量进行分类, 用贝叶斯方法对目标多幅不同方位角下图像的分类输出进行决策融合, 得到最终类别决策。用MSTAR数据库中3个目标进行识别实验, 实验结果表明, 对3幅以上不同方位角的图像进行决策融合时, 该方法可显著提高目标的正确识别率。该方法是一种有效的SAR图像目标识别方法。

In this paper, new synthetic aperture radar (SAR) image target recognition approach based on multiple views decision fusion is presented. Image chips are represented as feature vectors by 2-D wavelet transformation and principal component analysis algorithm. The feature vectors are classified by using algorithms of support vector machine (SVM). After multiple views of the same vehicle collected at different aspects classified by SVM, the outputs are then fused using Bayesian approach and the final classification decision is generated. Experiments are implemented with three class targets in Moving and Stationary Target Acquisition and Recognition (MSTAR) Program database. Experimental results indicate that there are significant target recognition performance benefits in the probability of correct classification when three or more views are used for decision fu-sion. Therefore, the approach proposed is an effective method for SAR image target recognition.