一种改进的基于光流法的运动目标跟踪算法

运动目标跟踪需要从背景中准确地检测出感兴趣目标并实现有效率的跟踪。文章结合Codebook模型和光流法提出了一个新的目标跟踪方法，首先用Codebook模型检测得到感兴趣目标，然后提取感兴趣目标内部的特征点并用光流法进行跟踪，跟踪过程中实时更新用以跟踪的目标内部的特征点。当目标发生遮挡时，采用Kalman滤波器预测目标的位置，遮挡结束后根据Kalman滤波器预测的位置和Codebook检测结果重新初始化感兴趣目标内部的特征点。实验结果表明，该算法具有较好的鲁棒性和较高的准确率，能够满足实时跟踪的要求。

An Improved Moving Object Tracking Algorithm Based on Optical Flow Method

Moving object tracking needs to detect the object-of-interest accurately from the background and then realize tracking efficiently. This paper proposes a novel moving target tracking method combining Codebook model and optical flow method. Firstly, Codebook model is applied to obtain object-of-interest, then extract feature points within object-of-interest to track them using optical flow method. During tracking, the internal feature points of the target are updated real-timely. When occlusion happens, Kalman filter is used to predict the location of object. After recovery from occlusion, the feature points are initialized according to the location predicted by Kalman filter and the detection result of Codebook model. The experimental results show that the algorithm has a good robustness and high detection precision, and can satisfy the requirements of real-time tracking.