基于多特征融合的绿色通道车辆检测判定  

Detection method of vehicle in highway green toll lane based on multi-feature fusion

作　　者：张浩[1,2] 刘海明[1,2] 吴春国[1,2] 张艳梅[1,2] 赵天明[1,2] 李寿涛[3] 

机构地区：[1]吉林大学计算机科学与技术学院,长春130012 [2]吉林大学符号计算与知识工程教育部重点实验室,长春130012 [3]吉林大学通信工程学院,长春130022

出　　处：《吉林大学学报（工学版）》2016年第1期271-276,共6页Journal of Jilin University:Engineering and Technology Edition

基　　金：吉林省交通厅重大科技项目(2010-1-4);吉林省交通运输科技推广项目(2013-3-7)

摘　　要：为了保证对高速公路绿色通道车辆进行快速、高效的计算机检测,本文提出了一种基于多特征融合的绿色通道车辆检测判定方法 PGM-OCSVM。该方法首先利用主成分分析法(Principal component analysis,PCA)对学习样本的属性进行特征过滤,然后利用遗传算法(Genetic algorithm,GA)对one-class SVM的两个重要参数——函数核带宽σ和错误接受率v进行自适应调整,增强方法的易用性,最后构建one-class SVM模型学习样本,并对结果进行分类。通过对大数据样本分析,PGM-OCSVM可以有效地完成绿色通道车辆判定任务。目前该方法已经应用于绿色通道车辆检测系统中,取得了良好效果。In order to ensure efficient computerized detection of vehicles passing green toll lane of highway,a detection method,named PGM-OCSVM,is proposed for free toll lane of highway based on multi-feature fusion.First,the Principal Component Analysis（PCA）is used to filter and simplify the sample characteristics.Then,the Genetic Algorithm is applied for adaptive adjustment of bandwidth of kernel function（σ）and false acceptance rate（v）,which are two important parameters of one-class SVM.Finally,a one-class SVM model is constructed to learn the samples and classify the results.Big data analysis demonstrates that the proposed PGM-OCSVM can effectively complete green-vehicle discrimination task.This method has been applied to the vehicle detection system in free toll lane of highway.

关 键 词：计算机应用 支持向量机 多特征融合 单类分类器 绿色通道车辆检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]