基于贝叶斯模型与最佳伙伴相似度量的目标跟踪

提出了一种基于产生式与判别式联合模型的视觉目标跟踪算法。首先介绍了一种基于全局颜色特征直方图特征的贝叶斯分类器，检测出若干最有可能属于目标的候选区域，然后利用最佳伙伴相似性度量（Best-Buddies Similarity）得到候选区域与目标模板的相似度，结合概率值与相似度值估计出最优的目标状态。通过划分目标-背景区域模型、目标-干扰区域模型，对可能产生干扰的区域提前进行抑制，降低了长期跟踪可能产生的漂移问题的风险，同时引入了自适应尺度估计机制和在线模型更新策略，以获得更为精准的跟踪结果。在37组具有挑战性的图像序列上与7种优秀的算法对比实验表明，所提出的算法能够有效应对光照变化、遮挡、旋转与尺度变化等多种问题。

Visual Object Tracking using Bayesian Models and Best-Buddies Similarity

This paper presented an object tracking algorithm using both the generative appearance model and the discriminative model. First, the algorithm employed a Bayes classifier based on holistic color histogram to distinguish the object. To avoid the tracker drifting to nearby regions which appear similarly to the object of interest, it identified and suppressed the distracting regions in advance. Second, the algorithm adopted a generative model based on Best-Buddies Similarity (BBS) to measure the similarity between candidate regions with the object template. Finally, the algorithm introduced a multiple scale searching method to enable the tracker to handle scale change, and used an adaptive model update strategy to enhance the tracking performance. Extensive experimental results on 37 challenging sequences show that the proposed algorithm performs favorably against state-of-the-art methods in terms of accuracy and robustness.