欢迎访问《汽车安全与节能学报》，
汽车安全与节能学报

汽车安全与节能学报 ›› 2016, Vol. 07 ›› Issue (03): 285-290.DOI: 10.3969/j.issn.1674-8484.2016.03.006

• • 上一篇    下一篇

基于车尾部特征的对车辆的视觉精确定位

谭华春1，周洋1，李克强2，钟智宇1   

  1. 1. 北京理工大学 机械与车辆学院，北京 100081 ；
    2. 清华大学 汽车安全与节能国家重点实验室，北京100084，中国
  • 收稿日期:2015-12-16 出版日期:2016-09-25 发布日期:2016-09-30
  • 作者简介:谭华春（1975—），男（汉），湖南，副教授。E-mail: tanhc@bit.edu.cn
  • 基金资助:

    国家自然科学基金资助项目(61620106002) ； 汽车安全与能源国家重点实验室开放基金(KF14031)

Visual accurate positioning for vehicle based on vehicle rear features

TAN Huachun1, ZHOU Yang1, LI Keqiang2, ZHONG Zhiyu1   

  1. 1. Beijing Institute of Technology, School of Mechanical Engineering, Beijing 100081, China;
    2. State Key Lab of Automotive Safety and Energy, Tsinghua University, Beijing 100084, China
  • Received:2015-12-16 Online:2016-09-25 Published:2016-09-30

PDF

765

摘要：

在车辆防碰撞预警系统中，对道路前方车辆的检测定位是很重要的研究内容。为了实现对车辆的精确定位，为后续的车距测量服务，本文使用已有的车辆视觉检测方法（AdaBoost 和聚合通道特征）对车辆进行初步检测，并利用车辆几何尺寸约束滤除误检结果。在车辆的感兴趣区域内利用车尾部特征（车尾部阴影及车后轮）对车辆进行精确定位。相较于没有添加几何约束的粗定位方法，本文提出的精确定位方法在车辆误检率方面降低了2.28%，在车辆定位的误差均值方面降低了44.58%。实验结果表明：该方法能有效的解决传统车辆检测方法对于车辆定位精确度不高的问题。

关键词: 汽车安全, 防碰撞预警, 精确定位, 车辆视觉检测, 几何尺寸约束, 车辆尾部特征, 聚合通道特征

Abstract:

The vehicle detection and positioning of the road ahead is very important in the vehicle anticollision warning system. To achieve accurate positing of vehicle, and contribute to subsequent vehicle distance measurement, the primary vehicle detection was done by using the existing visual vehicle detection method (AdaBoost and aggregate channel feature). Then, filtering false positive results by using vehicle geometry constraints. At last, the vehicle accurate positioning based on the rear feature of the vehicle (the rear shadow and the rear wheel) in the area-of-interest. Compared to the positioning method without geometric constraint, the proposed method reduces the false detection rate of the vehicle by 2.28%, and the mean error of the vehicle location is reduced by 44.58%. Experimental results show that the method can effectively solve the problem of low vehicle position accuracy of traditional vehicle detection method.

Key words: automobile safety, anti-collision warning system, accurate positioning, visual vehicle detection, geometry constraints, the rear feature of the vehicle, aggregate channel feature