Real-time and fast RGB-D based people detection and tracking for service robots

Abstract

People detection and tracking is one of the most important skills for a mobile service robot sharing its workspace with humans. This paper considers how to detect and track a person in indoor environments for service robots, with applications to human-robot interaction. The people is detected by a RGB-D camera, and then tracked by a mobile robot. This allows the path of the person relative to the service robot to be determined. Simple pose calibration and motion prediction can then be used to estimate the tracking path. The key advance in this paper is a novel people detection and tracking approach for RGB-D data to help determine the person position and moving trajectories in the field of view of the robot. The experimental results are provided to demonstrate the feasibility and effectiveness of our method in real world environments. Our results show that, typically, people can be successfully tracked with an average frame rate of 12 fps. This is sufficient for many applications; greater accuracy could be obtained with a higher resolution sensor.