改进的ReliefF算法用于雷达距离像目标识别  

Modified ReliefF algorithm for radar HRRP target recognition

作　　者：廖阔[1] 付建胜[1] 杨万麟[1] 

机构地区：[1]电子科技大学电子工程学院,成都610054

出　　处：《电子测量与仪器学报》2010年第9期831-836,共6页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(编号:60772146);国家863高技术研究发展计划(编号:2008AA12Z306)

摘　　要：提出了一种改进的ReliefF算法,并将其用于雷达高分辨距离像(high resolution range profile,HRRP)目标识别。与传统ReliefF算法相比,新算法通过在每类目标中等距离间隔抽取相同数量样本的方式进行权值累积,降低了样本数量及分布差异等因素对特征权值的影响,得到了更稳定有效的特征权值。利用此权值不但可降低特征向量维数,并可对最小距离分类器加权,提高目标识别率。最后,对5种不同飞机实测数据的识别结果表明本算法可达到83%的平均识别率。A modified ReliefF algorithm is proposed in this paper.Meanwhile,it is used in radar target recognitionwith high resolution range profile（HRRP）.Comparing with traditional ReliefF methods,the new algorithm selects thesamples with equal distance interval in each target’s sample set,uses the selected samples to cumulate weight of features.That can decrease the effect of different target’s sample number and distribution.So,more steady feaures’weightvectors could be gotten.The weight can be used to reduce feature dimension and weight minimum distance classifier.The experimental comparisons based on the measured data of five-type aircrafts show that the proposed method effectively improves the classification performance.The average accuracy is 83%.

关 键 词：雷达目标识别 高分辨距离像 RELIEFF算法 特征权值 

分 类 号：TN957.54[电子电信—信号与信息处理]