复杂道路环境中前方车辆检测方法  

Front Vehicle Detection Method in Complex Road Environment

作　　者：王战古 高松[1] 邵金菊[1] 于杰 冯鹏航 Wang Zhangu;Gao Song;Shao Jinju;Yu Jie;Feng Penghang(School of Transportation and Vehicle Engineering,Shandong University of Technology,Zibo 255049,P.R. China)

机构地区：[1]山东理工大学,淄博255049

出　　处：《科学技术与工程》2017年第33期112-118,共7页Science Technology and Engineering

基　　金：国家重点研发计划(2016YFD0701101);山东省自然科学基金(ZR2016EL19)资助

摘　　要：为提高前方车辆检测在不同道路环境中的鲁棒性和实时性,提出一种基于支持向量机的多传感器融合前方车辆检测方法。系统工作前利用多传感器数据融合建立雷达坐标与图像坐标的转化关系,以毫米波雷达在各种复杂道路环境中前方障碍物的检测数据为基础,利用支持向量机(SVM)训练分类器构建车辆与非车辆识别系统,最终根据车辆宽高比的统计规律,建立前方车辆识别窗口。道路试验结果表明该方法前方车辆识别准确率为89.2%,单帧图像的处理速度为31 ms。对于不同道路环境中的前方车辆检测表现出了良好的稳定性和准确性,总体性能取得较为显著的提高。In order to improve the robustnes and real-time performance of advanced driving assistant system for front vehicle identification,a new method based on support vector machine for multi sensor fusion is proposed.The relationship between radar coordinates and image coordinates is established by multi sensor data fusion.In order to build vehicle and non-vehicle identification system,support vector machine(SVM)algorithm is used to train proper classifiers by obstacle detection data of millimeter wave radar in different road environment.Finally,the vehicle identification window is established according to the statistical law of vehicle aspect ratio,and then complete front vehicle identification.The road test results show that the accuracy of the algorithm is89.2%,and the processing speed of single frame image is31ms,which shows good adaptability and accuracy in vehicle identification of different road environments,what's more overall performance has been significantly improved.

关 键 词：支持向量机 毫米波雷达 机器视觉 前方车辆检测 

分 类 号：TP212.9[自动化与计算机技术—检测技术与自动化装置]