一种基于近似支撑矢量机(PSVM)的交通目标分类方法

AN APPROACH FOR TRAFFIC OBJECT CLASSIFICATION BASED ON PROXIMAL SUPPORT VECTOR MACHINE

本文介绍了支撑向量机的特点,给出了实际应用中传统支撑矢量机存在的问题。为了克服支撑矢量机算法的不足,引入了一种近似支撑矢量机(PSVM)算法,并将此算法用于交通目标的分类识别。实验结果表明此算法比BP神经网络法准确率高,比传统的SVM法的效率高。

The paper describes the traditional support vector machine （SVM） and some features of it, and discusses some problems in some applications. In order to overcome the defections of traditional SVM, an approach of proximal support vector machine （ PSVM ） is presented. While the approach is applied to traffic object classification, the results show that the approach of PSVM is more accurate than BP nerve network, and more efficient than traditional SVM.