改进型AdaBoost算法在车辆检测中的应用  

The application of improved AdaBoost algorithm in vehicle detection

作　　者：刘天键[1] 邱立达[1] 张宁[1] 

机构地区：[1]闽江学院物理学与电子信息工程系,福建福州350121

出　　处：《闽江学院学报》2014年第2期74-79,共6页Journal of Minjiang University

基　　金：福建省教育厅科技项目(JA12263;JB11127);福州市科技计划项目(2013-G-86)

摘　　要：Adaboost算法通过特征选择把一组弱分类器组合成强分类器,并已成功地应用在人脸识别中.遵循Adaboost算法的基本思想,提出在联合空间进行Haar类特征的选择,并把该算法应用在车辆的检测中.使用基于联合特征空间的Adaboost算法,可以改善检测器的可靠性和快速性.实验结果表明与传统的Adaboost算法相比,可以提高分类器的目标击中率和降低虚警率.Adaboost is used to select and combine weak classifiers from a very large pool of weak clas- sifiers and it has been proven to be very successful for detecting faces. We follow the approach and present a new Harr feature selection in joint feature space. It is applied to detect rear views of cars. By using Adaboost in joint feature spaces, a reliable and fast classification is got. Experiment shows this classification perform good hitting rate and litter false positive rate than traditional AdaBoost algorithm.

关 键 词：ADABOOST算法 联合特征空间 分类器 车辆检测 

分 类 号：U491.2[交通运输工程—交通运输规划与管理]