An Object-Tracking Algorithm Based on Bayesian-Learning

Real-time object tracking is recently becoming very important in many video processing tasks. Applications as video surveillance, robotics, people tracking, etc., need reliable and affordable video tracking tools. Most of current available solutions are, however, computationally intensive and sometimes require expensive video hardware. In this paper, we propose a new object tracking algorithm for real-time video based on a new probabilistic approach that results in a Bayesian-learning process. This approach infers the trajectory of a moving object by applying a very simple optimization method, which makes the tracking algorithm robust and simple to implement. Experimental results are provided to demonstrate the performance of the proposed tracking algorithm in complex real-time video sequence scenarios.