摘要
支持向量机(SVM)具有分类精度高、泛化能力强等优点,已成功应用在雷达目标识别领域。但其性能受多种因素影响。针对低信噪比、分类面混迭、参数选取等问题,文章提出剪辑SVM分类器,通过小波去噪、剪辑、矩阵相似度优选参数等手段有效抑制上述问题的影响。外场实测数据的仿真也表明剪辑SVM的性能优于传统SVM与最近邻分类器。
Support vector machines has a high classification accuracy and strong generalization ability.It is successfully applied in the radar targets recognition.In view of problems such as low SNR,large training samples,classification face aliasing,parameter value choosing,this paper bring forward Edit-SVM classifier.We can restrain the problem mentioned before efficiently through wavelet de-noising,cluster,editing and matrix similarity optimization parameters.The acquired outfield data proved the performance of Edit-SVM is better than traditional SVM and nearest neighbor classifier.
出处
《舰船电子工程》
2010年第4期68-71,165,共5页
Ship Electronic Engineering
关键词
高分辨距离像
目标识别
支持向量机
high resolution range profiles
targets recognition
support vector machines
