公路收费站车辆跟踪及抓拍算法研究与系统实现

为保障公路收费站对车辆抓拍和车流统计的抗干扰能力,以静止单孔摄像机获取的检票口车道视频作为研究对象,提出了一种高效的易于扩展的抓拍判断系统框架.在分析常见运动检测方法优劣的基础上,从实时性和鲁棒性考虑,采用基于运动历史图像的改进的帧差法,以提高运动检测的灵敏度;为缓解服务器的计算压力,提出了一种高效的车辆矩形区域快速定位算法,并在此基础上定义了基于时间和空间变化的规则,以排除摄影机前人和杆臂运动对镜头的遮挡,最终构成了抓拍判断系统框架.此外,就多路车道在不同光照下并行地进行了实时抓拍实验,结果显示,在总时长5.5 h的测试样例中,车辆计数平均准确度达87.8%,证明该框架可显著减弱抬杆、落杆的遮挡以及光照变化的影响,提高抓拍的精度. 

Video-Based Vehicle Tracking and Capturing System for Expressway Tollgates

In order to reduce the noise impacts in front of the camera and improve the vehicle capturing precision in expressway tollgate scenes, an efficient and flexible judgment framework for vehicle capture was proposed. Specifically, through the analyses of the common motion detection methods, an improved frame difference approach based on motion history images was applied to the framework to increase the motion detection sensitivity. To relieve the calculation complexity, a fast detection algorithm for searching vehicle rectangular region was given. Furthermore, the spatio-temporal rules for vehicle capture judgment were defined, as a result, the judgment framework was formed. In addition, parallel vehicle capturing experiments were conducted on multiple lanes under varied illumination in real time. The experiment result shows that using the proposed framework, the average precision for a 5.5 h test sequence is up to 87.8%, and it is able to resist vehicle and bar movement noises and luminance variation to improve the vehicle capturing precision.