夜间复杂交通场景中的车辆检测和跟踪

发布时间：2018-01-14 23:10

本文关键词：夜间复杂交通场景中的车辆检测和跟踪　出处：《交通运输系统工程与信息》2016年01期 　论文类型：期刊论文

【摘要】：为提高夜间车辆视频检测和跟踪的准确率,提出一种夜间车辆检测和跟踪算法.本算法通过亮斑分割和连通组件匹配来检测和定位车辆前灯,并利用区域跟踪算法对前灯进行跟踪以提高检测准确率.考虑到夜间行车时车辆前灯的显著特征,通过改进Otsu方法以自适应地分割明亮区域,并根据前灯的几何形状、尺寸及位置信息滤除非车灯部分的车辆信息.然后利用前灯的对称性进行前灯的配对和归类;最后采用区域跟踪算法对前灯进行定位和跟踪.实验结果表明,本算法车辆检测平均准确率大于97%,处理速度比已有方法提高15.8%以上.
[Abstract]:In order to improve the accuracy of video detection and tracking, a nocturnal vehicle detection and tracking algorithm is proposed, which detects and locates vehicle headlights by bright spot segmentation and component matching. The region tracking algorithm is used to track the headlights to improve the detection accuracy. Considering the prominent features of the headlights when driving at night, the bright region can be self-adaptively segmented by improving the Otsu method. According to the geometry, size and position of the headlight, the information of the headlight is filtered except the vehicle information of the headlights. Then, the headlights are matched and classified by the symmetry of the headlights. Finally, the region tracking algorithm is used to locate and track the headlights. The experimental results show that the average accuracy of vehicle detection is more than 97 and the processing speed is more than 15.8% higher than that of the existing methods.
【作者单位】：天津工业大学电子与信息工程学院;
【基金】：天津市科技支撑计划重点项目(14ZCZDGX00033)~~
【分类号】：U495
【正文快照】： 0引言基于计算机视觉技术的车辆检测在智能交通及自动驾驶等领域具有重要的理论意义和实用价值.日间车辆检测常选用边缘特征、形状模板、车辆轮廓等特征检测车辆,而这些特征不适于车辆与环境背景之间差异很小的夜间环境.为检测夜间行驶车辆,文献[1]通过提取和跟踪运动车辆的角

【相似文献】