Accurate indoor tracking using a mobile phone and non-overlapping camera sensor networks

Abstract

Cameras are common in the public building, and become one of the popular sensors for the indoor tracking. However, a camera will produce much data and consume much energy. Consequently, the cameras, deployed in a large area, would have no overlapping field of views (fovs). This leads to we cannot determine the person's position in unobserved area correctly, due to the person's unpredicted motion change, such as taking turns. Smart phone-based dead reckoning (DR) method is recently observed to be an alternative for filling these cameras' gaps to realize seamless tracking. Therefore, in this paper, we propose to accurately track a person in indoor environment by employing particle filtering, which fuses the absolute position results from cameras and the relative position results from the smart phone-based DR method. Experiment results demonstrate the performance of the proposed joint tracking system. As we expected, the dead reckoning's drawbacks such as accumulative errors and the unavailability of the initial position, are solved by the cameras. As well, the positions' estimations between the cameras are filled with the DR method.