鲁棒的车载摄像头对向车辆检测与跟踪方法  

Robust on-board camera on-coming vehicle detection and tracking method

作　　者：鹿文浩[1] 王生进[1] 

机构地区：[1]清华大学电子工程系,北京100084

出　　处：《清华大学学报（自然科学版）》2013年第11期1509-1514,共6页Journal of Tsinghua University(Science and Technology)

基　　金：国家"八六三"高技术项目(2011AA110402)

摘　　要：在智能交通领域,越来越多的新兴应用场景如移动式的车流量统计系统和电子收费系统需要通过车载设备采集并分析视频数据。因此,一种有效、快速的车载摄像头对向车辆检测与跟踪方法具有重要意义。该文针对车载摄像头对向车辆检测与跟踪问题通过选择与当前路段环境匹配的车辆模式,动态应对对向车辆的表观变化(如在路段或摄像机参数发生变化时视角以及截断情形的改变),同时避免提高算法的单帧检测时间。实验结果表明:该方法能够实时、有效地检测并跟踪对向车辆。Intelligent transportation systems will involve many new applications such as mobile traffic flow monitoring systems and mobile electronic tolling systems which require video analysis methods using on-board systems. Effective on coming vehicle detection and tracking systems using on board cameras will be very important. This study analyzes the various vehicle appearances for on-coming vehicle detection and tracking systems for different road environments, camera parameters, views and truncations of the vehicle. A low computational tracking method is then developed using adaptive vehicle patterns that best fit the current road environment. Tests show that the method efficiently detects and tracks vehicles in on-coming traffic for varied weather conditions.

关 键 词：对向车辆检测 目标跟踪 模板匹配 车流量统计 智能交通 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]