A Quantitative Real-Time Analysis Of Object Tracking Algo- rithm For Surveillance Applications

Three major object tracking algorithms are evaluated in this paper. The performances of these algorithms for different video sequences are analysed. Object tracking is an important task in many surveillance applications, some problem and its difficulty depend on several factors such as pose change, various lighting condition, occlusion, dynamic object, scale and object motion, recent tracking algorithm focus on these three things robustness, adaptively and real-time processing. This paper analyses and identifies the major roles of tracking methods to meet their particular challenges and suggest how to design and choose the tracking algorithm for various conditions. Multiple instance learning tracker (MILT), visual tracking decomposition algorithm (VTD) and track-learning-detection method (TLD) algorithms are tested on various challenging sequence to obtain comparative results. Strength and weakness of these algorithms according to the degree of evaluation criteria are also analysed. KeywordsMultiple instance learning tracker; visual tracking decomposition; Track-learning-detection; Object Tracking; Video Surveillance