Multi-object trajectory coupling using online target specific decision making

The color and gradient based sequential state estimation method has proved its applicability in many video based tracking applications. This paper proposes a multi-modal approach applicable to trajectory formation of multiple moving objects with complex random motion structure. The Bayesian framework for tracking is formulated in this paper that incorporate spatio temporal information in selecting significant particles and establishing statistical correlation between prior model of target and its recent observation. It is especially applicable to real time trajectory analysis of situations with miss detection and formation of segmented tracks belonging to same object. The quantitative as well as qualitative performance of the proposed approach is evaluated on various real-world video sequences with challenging environment like random movement between objects and partial occlusion. The proposed approach performs better than other state-of-art method used for multiple moving objects tracking in videos.