Computationally Efficient Moving Object Tracking with CAMSHIFT Embedded Kalman Filter Theory
Kabir Hossain, Chi-Min Oh, 이칠우 · International Conference on Human-Computer Interaction · 2012
In this paper we propose a computationally efficient moving object tracking algorithm with CAMSHIFT Embedded Kalman Filter technique. Object tracking using CAMSHIFT still is suffering from fundamental problems such as cluttered background, occlusion and dynamic movement of objects. In order to cope with these problems, the Kalman filter is applied in CAMSHIFT to predict the location of object for efficient object tracking. As Kalman Filter is a set of mathematical equation, so it increases the computational complexity of our propose system. We solve this problem by using random frame selection procedure. We randomly select some frame instead of processing every frame for reducing the computational complexity. The experimental results are shown to prove the robustness, computational efficient and validity of the propose algorithm using various video sequences of moving object.