Visual tracking based on local patches
Wang Baoyun, Zhou Lei, Ping Deng · 2014
Tracking-by-detection based on online learning has shown superior performance in visual tracking of previously unknown objects. However, most approaches are limited to a fixed-size box representing objects unless applying affine transformation method. They can neither show the object's visible area without shelter nor handle the object complete occlusion and disappearance situation. To overcome the limitations, we propose a novel tracking-by-detection approach bashed on local patches in this article. We extend ferns forest to visual tracking and optimize online learning with the reliability of the predicted object. Moreover, a re-sampling technique is used to obtain a object's scale and visible area without much backgroud and shelter. Besides, in order to optimize online learning method, we establish a novel credibility evaluation standard for the predicted object, which can adapt to complete occlusion and disappearance scene. To show the benefits of our approach, we run our algorithm on various challenging sequences, and compare it with the state-of-the-art methods. The experiment results show that our algorithm enjoys an accurate tracking and a good robustness in tracking rigid and non-rigid objects.