【摘要】 针对目前支持向量机多类分类方法存在的缺点,文中对支持向量机的高斯核函数进行改进,并提出一种结合留一法和单一验证法的参数选择新方案。基于3种雷达目标的HRRP数据,设计了相应的预处理算法,利用改进的SVM分类法用于高分辨距离像的雷达目标识别。从实验目标姿态稳定性、训练集大小稳定性和抗噪能力三个方面验证改进SVM的稳健性。更多还原

【Abstract】 To solve the problems and defects of existing methods in Support Vector Machine(SVM) multiclass classification,an improved Gaussian kernel in SVM is proposed and a new SVM model selection scheme combining Leave-One-Out method with One-Validation method is presented in this paper.Based on the High Resolution Range Profile(HRRP) of three-type target,a preprocessing method is designed,A classification algorithm for HRRP based on this improved SVM is applied.Finally, experimental results prove that the improved SVM classifier has better performance on target-aspect stability,training set-size stability and anti-noise ability than traditional SVM.更多还原