支持向量机的战场直升机目标分类识别  

Research on Helicopter Target Identification based on Support Vector Machine

作　　者：李京华[1] 许家栋[1] 李红娟[2] 

机构地区：[1]西北工业大学电子信息学院,陕西西安710072 [2]西北工业大学航空学院,陕西西安710072

出　　处：《火力与指挥控制》2008年第1期31-34,共4页Fire Control & Command Control

基　　金：国防重点实验室预研基金资助项目(51454070204HK0320);西工大科技创新基金资助项目(2003CR080001)

摘　　要：对基于支持向量机的战场直升机目标分类识别技术进行了研究,分别将谐波集(HS)频率和不同尺度小波子空间能量作为特征矢量,设计出一种基于支持向量机的直升机目标分类器,并将该分类器与kNN分类器和BP神经网络分类器进行分类对比实验。结果表明两种特征提取方法,都能很好地体现不同声目标之间的差异,SVM分类器相对于其他两种分类器具有更好的分类性能,目标识别率达到96%以上。The helicopter target identification techniques are studied. Two feature extraction approaches employ, respectively, the harmonic sets frequencies and the energies in different scales after the wavelet decomposition as the feature vectors. Support Vector Machine （SVM） classifier is designed in helicopter recognition. Also three classifiers including K-nearest neighborhood classifier, BP neural network classifier and SVM classifier are used to do the comparison experiments of targets classification. Result shows that these two feature extraction approaches can stand for the difference among different acoustic targets well and SVM classifier get better classification performance than other two classifiers. The classification accuracies can reach as high as 96%.

关 键 词：支持向量机 目标识别 特征提取 分类器 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TN957[自动化与计算机技术—控制科学与工程]