中文摘要：

可靠的车辆跟踪是实现交通事件自动检测的重要前提,车辆跟踪中的车辆相互遮挡则是影响车辆跟踪结果的关键因素.为了解决这一难题,文中提出一种基于ST-MRF模型的自适应车辆跟踪算法.在ST-MRF模型中,把图像分成块,将相邻图像间的块通过它们的矢量联系起来,建立运动序列图像的时空马尔可夫随机场模型并且构造其相应的能量耗费函数,然后利用松弛算法实现目标地图最小化能量计算,从而解决车辆跟踪中的遮挡问题.实验结果表明,跟踪不遮挡的车辆时达到的跟踪成功率为95%,遮挡情况时成功率也可达到91%.通过实验得出以下结论：基于ST-MRF模型的自适应车辆跟踪算法能在交通量比较大,且车辆出现相互遮挡的情况下,能较准确地获得车辆跟踪数据.为以后的交通事件检测提供重要的数据基础.

英文摘要：

Robust vehicle tracking algorithm is an important precondition for traffic event detection,but occlusion is a key influence factor for vehicle tracking.An adaptive vehicle tracking algorithm based on the spatio-temporal Markov random field model is proposed to deal with the problem.The paper defines the ST-MRF to divide an image into blocks as a group of pixels and to optimize labeling of such blocks by connecting blocks between consecutive images referring to their motion vectors.The spatial-temporal Markov random field model based on motion sequence image is formulated and the energy function is determined by the model.Then the minimum energy target image is realized by the metropolis algorithm,and the occlusion problem is solved in vehicle tracking.The Experimental results indicate that the tracking success ratio when vehicles are not occluded with each other is 95%,and when being occluded the tracking success ratio is 91%.The experiment proves that the accurate tracking data can be obtained via the algorithm,even when there is large traffic volume and vehicle mutual occlusion exists.