模糊AdaBoost算法在SAR图像目标识别中的应用

Synthetic Aperture Radar Image Target Recognition Using Fuzzy AdaBoost Algorithm

提出了一种AdaBoost算法的多类别推广方法,并将推广后的算法应用于合成孔径雷达图像目标识别中。针对AdaBoost基本算法只考虑两类分类的情况,对算法进行多类别推广,用"一对一"方法将多类别分类问题分解为多个两类分类问题,用模糊方法对多个两类AdaBoost分类器的输出进行决策判决,得到最终分类结果。将推广后的模糊AdaBoost算法应用于SAR图像目标识别,用MSTAR数据库中3个军事目标进行识别实验。实验结果表明,该算法可有效应用于SAR图像目标识别。与其他分类算法相比较,可获得较高的目标正确识别率。

A method of multi-class classification for extending AdaBoost algorithm is presented and a new algorithm is used on synthetic aperture radar image target recognition. Because the basic AdaBoost algorithm is only suited for two class classification, it needs to be extended for multi-class classification. A multi-class classification problem is decomposed into multiple two class classification problems by using “one against one” method, and a fuzzy algorithm is proposed to give a decision from outputs of multiple two class AdaBoost classifiers. From the end of the fuzzy algorithm, the final classification outcome is obtained. Then the extended fuzzy AdaBoost algorithm on SAR image target recognition is used and three military targets in MSTAR database are experimented. Experimental results show that the extended fuzzy AdaBoost algorithm is effective on SAR image target recognition and a better target correct recog-nition rate is obtained.