基于复合特征及分层特征选择的雷达HRRP识别

Radar HRRP recognition based on hybrid features and multi-stage feature selection scheme

雷达高分辨距离像(highresolutionrangeprofile,HRRP)具有方位敏感性、平移敏感性以及在特征空间高度交叠的特点,采用复合特征可以更好地描述目标特性。利用复合特征,结合分层识别结构提出了一种分层特征选择方法,充分利用了特征信息,简化了识别器结构,使识别运算量大大下降且提高了识别率。基于HRRP的平移不变特征和相关矢量机的计算机仿真实验表明,该方法是有效的。

Radar high resolution range profile (HRRP) is sensitive to the target aspect and time-shift variation, and highly overlapped in feature space between different targets, which determines that hybrid features are suitable to represent the target's property. Based on hybrid features, a multi-stage feature selection and classification scheme is proposed, which utilizes the feature information more efficiently, decreases the computational complexity and improves the classification performance as well. The simulation results based on relevance vector machine (RVM) classifier show the efficiency of the proposed method.