Object Tracking Based On Sparse Representation

Object tracking is a very important research direction in the computer vision researcharea, with the goal of locating the target and capturing the change of target’s appearance andmotion. Many researchers have paid attention to the technology in the object tracking. Andthe object tracking has been widely applied to all aspects of life, such as video surveillance,human-computer interaction, intelligent transportation, medical diagnosis, and vehiclenavigation. In recent years, for the field of computer vision, sparse representation has alsobeen widely used, and some excellent achievement have been done, which fully show thehuge potential. So we propose a robust object tracking algorithm based on sparserepresentation in this paper which includes two parts as follows:We propose the effective weighted compressive tracking. The real-time compressivetracking algorithm does not discriminatively consider the sample importance, leading to someunstable results and some drift. Against this shortcoming, a weighted thought is proposed, weintegrate the sample importance into an efficient online learning procedure so that the featuresare much more discriminative, so we can get more stable results.We propose the object tracking method based on local sparse representation. It proposesa local sparse appearance model, which is able to exploit local information of the target viasparse coding on local patches. The tracker with local features is able to copy with targetappearance changes well. In addition, we take into account an occlusion handing scheme inthe paper. In the end, a logistic regression classifier is used to distinguish between backgroundand target object well. We have done some experiments for various challenging videos, theyall demonstrate that we tracker performs favorably against many state-of-the-art algorithms.