有监督Kohonen网络的车型识别方法  

Research on vehicle type identification method based on supervised Kohonen network

作　　者：纪野 李玉惠[1,2] 李勃[1,2] 杨敏[1,2] 

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650500 [2]云南省智能交通系统工程技术研究中心智能图像处理研究室,云南昆明650500

出　　处：《传感器与微系统》2016年第8期12-14,19,共4页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(61363043)

摘　　要：车型识别已成为智能交通系统研究中的一个重要内容。根据同类车型尺寸特征如长、宽、高特征值都具有相似性特点,利用已有车型数据库,分别使用逆传播(BP)神经网络、支持向量机(SVM)网络、有监督Kohonen网络这三种神经网络分类方法对车型尺寸特征进行分类,得到三个车型识别准确率并进行比较验证。实验结果表明:有监督Kohonen网络的分类方法具有较高的车型识别精度,实验效果明显,车型识别准确率高。Vehicle type identification has become an important research content in intelligent transportation systems. According to the characteristics of size similar vehicle models such as long,width,height features value are similar characteristics,using the existing vehicle models database,respectively using three neural network classification methods,such as BP neural network,SVM network,supervised Kohonen network,to classify features of vehicle type size,obtain identification accuracy rate of three vehicle type and comparison and verification are carried out. Experimental results show that the classification method of supervised Kohonen network has high precision of vehicle identification,the experimental effect is obvious,and the accuracy rate of vehicle type identification is high.

关 键 词：智能交通系统 车型识别 BP神经网络 SVM网络 有监督Kohonen网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]