摘要
针对传统支持向量机对噪声点敏感问题,提出一种改进的支持向量机。其基本思想是根据样本对分类贡献不同赋予相应的隶属度,贡献大的分配较大的隶属度,贡献小的分配较小的隶属度。与传统支持向量机比较,减小了噪声点对分类的影响,提高了SVM的泛化能力。并将其应用到车型识别中,结果显示该方法的有效性。
A improved support vector machine was proposed against the traditional support vector machine sensitive to yawp. The main idea is giving different subjection degree according to class contribution of data swatch, the larger contribution,the larger membership. Comparison with traditional support vector machine, it minishes the impact of yawp and improves the class ability of support vector machine. The results in vehicle type recognition shows validity of this method.
出处
《计算技术与自动化》
2009年第1期127-130,共4页
Computing Technology and Automation
基金
湖南省教育厅资助科研项目(07C507)
关键词
支持向量机
模糊支持向量机
模糊隶属度
车型识别
support vector machine
fuzzy support vector machine
fuzzy subjection degree
vehicle type recognition
