Integrating low-resolution surveillance camera and smartphone inertial sensors for indoor positioning

Abstract

Indoor positioning techniques for pedestrians are the keys to the location-based services (LBSs) for human beings. Accuracy, overhead and ease of use are most noteworthy to evaluate an indoor positioning system. In this paper we present a novel indoor positioning system that integrates widely distributed low-resolution surveillance cameras and smartphone inertial sensors, which are perfect devices to build a system with low cost and high ease of use. Our proposed system receives video sequence from the surveillance camera and locates the pedestrians in the video view. The target pedestrian is recognized out of the crowd by matching the features from its smartphone inertial sensors with the pedestrian features extracted from the video sequence. We also apply a CNN-based visual object tracking algorithm to handle the situation that the target pedestrian is partly blocked in the video sequence. Our experimental results show that high positioning accuracy (at centimeter level) can be achieved. We also compare the other aspects of our technique against the state-of-art indoor positioning techniques, and show its superiority in many aspects.