针对雷达目标一维距离像(HRRP)识别问题,结合支持向量机(SVM)和DS证据理论提出一种多极化HRRP分类方法—–SDHRRP.该方法通过混淆矩阵获取基分类器之间的距离,从而根据基分类器对不同目标类的分类能力给其赋予不同的可信度.将该可信度值与SVM后验概率结合到DS证据理论的基本概率赋值(BPA)中,以实现SVM和DS证据理论在目标识别中的有效结合.对实测目标数据的实验结果表明,基于分类器可信度得到的BPA能够有效避免证据冲突,SDHRRP方法可以有效降低融合分类的误差率. Aimed at radar target one-dimensional distance image (HRRP) recognition problem, a multi-polarization HRRP classification method, SDHRRP, is proposed by combining support vector machine (SVM) and DS evidence theory. This method obtains the distance between base classifiers by confusion matrix , So as to give different credibility to the classification ability of different target classes according to the classification ability of the base classifier.This credibility value and SVM posterior probability are combined into the basic probability assignment (BPA) of DS evidence theory to realize SVM And DS evidence theory in target recognition.Experimental results on measured target data show that BPA based on the confidence of classifier can effectively avoid evidence conflict and SDHRRP method can effectively reduce the error rate of fusion classification.