Real-time human tracking using fusion sensor for home security robot

Abstract

In this paper, we introduce a new human tracking system based on the fusion sensor consisting of the vision sensor and the depth sensor. Conventional tracking systems utilize either the vision or depth data. However, conventional systems are sensitive to illumination changes and often fail to track people in cluttered and crowded environments. To solve these problems, we propose a method of combining the data from both the vision and depth sensors. The proposed system tracks people using the conventional on-line boosting algorithm but the search range of tracking is significantly reduced by exploiting the depth data. Experimental results demonstrate that the proposed technique can robustly track people in real-time.