基于AdaBoost和压缩跟踪的前方车辆检测方法  

Detection of Preceding Vehicle Based on AdaBoost and Compressive Tracking

作　　者：谭力凡 吴俊[1,2] 曹立波[1] 廖家才[1] Tan Lifan;Wu Jun;Cao Libo;Liao Jiacai(State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body,Hunan University,Changsha 410082;Hunan Normal University,Changsha 410082)

机构地区：[1]湖南大学汽车车身先进设计制造国家重点实验室,长沙410082 [2]湖南师范大学,长沙410082

出　　处：《汽车技术》2019年第2期21-25,共5页Automobile Technology

基　　金：国家自然科学基金项目(51505137);汽车车身先进设计制造国家重点实验室(湖南大学)开放课题基金(31615007)

摘　　要：为提供准确、实时的前方车辆位置信息,提出了一种基于AdaBoost和压缩跟踪的前方车辆检测方法。首先通过Haar-like特征及AdaBoost算法进行样本训练,获得车辆检测分类器,利用道路纵向测距模型进行感兴趣区域分割和检测窗口多尺度优化以提高检测效率,最后通过压缩跟踪算法增强环境适应性,提高检测准确率。结果表明,该方法在天气良好条件下的平均检测率为93%,每帧耗时26.5 ms,能够为具有避撞功能的车辆智能驾驶系统提供有效决策依据。A detection method of preceding vehicle method was proposed based on AdaBoost and compressive tracking algorithm to provide accurate and real-time vehicle position information.The vehicle detection classifier was obtained through sample training with Haar-like feature and AdaBoost algorithm.Segmentation of interest region and multiscale optimization of the sub-window based on the longitudinal range model were utilized to improve the efficiency of detection.The compressive tracking algorithm was used to enhance environment adaptability and improve detection accuracy.The experimental results show that the average detection accuracy of the proposed method is 93%with a speed of 26.5 ms per frame under good weather conditions,which could provide an effective decision basis for the intelligent driving system with the collision avoidance function.

关 键 词：前方车辆检测 ADABOOST算法 道路纵向测距模型 压缩跟踪 

分 类 号：U471.15[机械工程—车辆工程]