基于Markov随机场的区域跟踪算法

Region tracking algorithm based on hidden Markov random field

由于视频序列的对象跟踪相当于把图像帧分割成跟踪与非跟踪两个不重叠区域,为此,引入图像分割算法中的Markov随机场模型,提出了一种多目标模糊规划求取Markov标记场的最优估计来实现区域跟踪的算法。此算法为了克服传统离散Markov随机场运算速度慢的缺点,利用双随机矢量,建立连续的Markov标记场,同时提取区域视觉和运动信息的模糊特征,从而改善了算法的鲁棒性和运算复杂度。最终实验结果表明,此方法不仅跟踪效果好,而且还有运算速度快、抗干扰能力强等特点。

To overcome the shortcoming of traditional Markov Random Field model in region tracking field, a unified multi-objective fuzzy coefficient programming framework was proposed for optimal estimation of the Markov label field, which introduced a doubly stochastic prior model, and extracted the fuzzy feature of region visual and motion characters. Finally, very promising experimental results on some real-world sequences are presented to illustrate the performance of the proposed algorithm.