Online Multi-expert Learning For Visual Tracking

Visual tracking is one of the most important topics in computer vision with many applications and has received significant attention of the computer vision community.The task of visual tracking is to estimate the trajectory of a target object in an image sequence,and the only information we know is the initial location of the target object.Although significant progress has been made recently,it still remains a challenging task due to factors like abrupt motion,motion blur,illuminational,occlusion and out of view.The correlation filters based trackers have achieved excellent performance for object tracking in recent years.However,most existing methods use only one filter but ignore the information of the historical filters.In this paper,we propose a novel online multi-expert algorithm for visual tracking.In our proposed scheme,there are former trackers which retain the historical filters,and those trackers will give their predictions in each frame.The current tracker represents the filter of current frame,and both the current tracker and the former trackers constitute our expert ensemble.We use an adaptive Second-order Quantile strategy to learn the weights of each expert,which can take full advantage of all the experts.To simplify our model and remove some bad experts,we prune our models via a minimum entropy criterion.Finally,we propose a new update strategy to avoid the model corruption problem.Extensive experimental results on both OTB2013 and OTB2015 benchmarks demonstrate that our proposed tracker performs favorably against state-of-the-art methods.