视频监控中运动车辆检测与跟踪技术研究

发布时间：2018-03-23 11:40

本文选题：运动车辆检测　切入点：车辆跟踪　出处：《安徽大学》2014年硕士论文

【摘要】：智能交通系统(Intelligent Transportation System, ITS)成为如今交通系统发展的方向。其中,利用视频监控技术,高效地对运动车辆进行检测与跟踪成为智能交通系统研究中的一个热点和难点。它是判断车辆行为,解决道路拥堵等问题的前提。本文主要研究路口单一摄像机固定的情况下,监控视频中运动车辆的自动检测、跟踪技术,对原有算法进行了一些改进,完成了相应程序的编写。在实际视频智能监控中,能否实时的进行计算机处理成为衡量算法实用价值的一个重要指标。本文从实际应用的角度,深入研究了运动车辆检测与跟踪算法,主要的工作可概况为以下几点： (1)分析比较运动目标检测算法,最终选用ViBe算法作为本文运动车辆检测算法。针对ViBe易产生鬼影的特点,改进了ViBe算法,实验表明改进的算法能使鬼影更快速地消除。 (2)针对运动车辆阴影对目标检测的影响,通过改进得到了在灰度空间下阈值能自适应更新的阴影去除算法。通过实验,验证了该方法具有较好的去阴影效果。 (3)研究目标跟踪的一般方法,从实时性、准确性上得出所需的车辆跟踪算法。最终改进得到了一种窗口受限的联合Camshift与Kalman滤波的车辆跟踪算法,实验取得了较好的效果。 (4)编程实现了多口标运动车辆检测与跟踪。结合OpenCV编写程序,使用基于检测线触发的机制实现了车辆的计数,多目标的跟踪,完成了车辆的标记以及运动轨迹的自动绘制。
[Abstract]:Intelligent Transportation system (ITSs) has become the developing direction of transportation system nowadays. Efficient detection and tracking of moving vehicles has become a hot and difficult point in the research of intelligent transportation system. It is the premise of judging vehicle behavior and solving road congestion problems. The automatic detection and tracking technology of moving vehicles in the surveillance video, the original algorithm has been improved, and the corresponding program has been written. In the actual video intelligent surveillance, it is an important index to measure the practical value of the algorithm by computer processing in real time. In this paper, the moving vehicle detection and tracking algorithm is deeply studied from the perspective of practical application. The main work can be summarized as follows:. 1) analyzing and comparing the moving target detection algorithms, and finally choosing ViBe algorithm as the moving vehicle detection algorithm in this paper. According to the characteristics of ViBe, the ViBe algorithm is improved. The experiment shows that the improved algorithm can eliminate the ghost image more quickly. 2) aiming at the influence of moving vehicle shadow on target detection, a shadow removal algorithm with adaptive updating threshold in gray space is proposed, and the experimental results show that the method has better shadow removal effect. 3) the general method of target tracking is studied, and the needed vehicle tracking algorithm is obtained from real-time and accuracy. Finally, a window-constrained vehicle tracking algorithm with combined Camshift and Kalman filtering is obtained, and the experimental results are satisfactory. Combined with the OpenCV program, the vehicle count and multi-target tracking are realized by using the mechanism based on the detection line trigger, and the vehicle marking and the automatic drawing of the moving track are completed.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495;TP391.41

【参考文献】