中文摘要：

针对传统Mean—Shift算法存在无法有效的估计目标方向及变速运动目标这一缺陷，提出了一种改进的Mean—Shift自适应跟踪算法，在传统算法基础上增添了next帧目标预测功能，并应用于智能交通系统。该算法首先提取运动历史信息，并建立数学模型，据预测理论对next帧目标位置进行预测，使其作为Mean—Shift迭代初始，提高迭代效率。实验表明，该跟踪算法具有良好的实时性和鲁棒性，受车辆方向和速度变化影响较小。

英文摘要：

Mean-Shift tracking algorithm is widely used in the field of video surveillance, it is a classical method. But there are lar- ger defects for tracking the changing of direction and speed target under the situation of complex movements, In view of this prob- lem, in this paper, an improved method based on Mean-Shift adaptive tracking is proposed to overcome this disadvantage. The function of prediction next target is added in this method, and the method is also used in intelligent transportation systems The al- gorithm takes the position of next target predicted as the Mean-Shift iteration initial position, in order to reduce their number of it- erations, and rapidly converges to the current position. This method is verified that it has real-time, robustness and less influenced by the changing of direction and speed.