A Novel Approach of Human Motion Tracking with the Mobile Robotic Platform

Abstract

This paper presents a novel and robust approach to detect and follow a human with a mobile robotic platform. In order to follow a human, both the initial detection of human and the subsequent tracking need to be implemented. As the robot is initially static, initial human detection is done using a background subtraction technique. To remove the outliers objects, filters are formulated based on the aspect ratio and horizontal projection histogram of the human. Human detection in subsequent frames is done by back-projecting the color histograms of the human torso and legs. To make the human detection robust, a shape analysis algorithm is developed to find the “two legs apart pattern” in the vertical projection histogram (VPH) of the detected foreground. For tracking, linear motion controllers are proposed: these require visual information to generate motion commands for the robot. The novelty of our approach includes (1) human tracking using visual information alone, (2) use of simple linear motion controllers to generate the translational and rotational velocities for the robot, and (3) cost effectiveness, as the experimental set up requires only one vision sensor. The current version of our system, runs on a Pioneer P3-DX mobile robot, and can follow human at up to 0.7m/s in an indoor environment.