一种有效的SAR图像自动目标识别方法

摘要

An effective approach for automatic target recognition in SAR(Synthetic Aperture Radar) images is proposed.The SAR targets are segmented by a double CFAR (Constant False Alarm Rate) segmentation method for SAR targets.With the space local features and overall PCA (Prinicpal Component Analysis) features extract from the SAR images,the targets are recognized by the method based on parameter learning and a combination of genetic algorithm optimization on the classifier.Thi9 method does calculation on the original image directly, avoids mass problems of compute depending on the data model.The result shows that the automatic target recognition method based on genetic algorithm performs well to targets T-72 and BMP2.%提出了一种有效的SAR图像中自动目标识别的方法.首先采取双阈值CFAR目标分割算法对SAR图像进行目标分割.通过对SAR图像的空间局部特征和PCA全局特征的提取,在参数学习的基础上,结合了遗传算法进行迭代优化获取分类器,实现SAR图像的自动目标识别.该方法可以直接时原始图像进行计算,避免了基于数据特征计算所带来的问题.实验结果显示,这种基于遗传算法的自动目标识别方法对T-72和BMP2坦克进行识别,获得了较好的识别率.