Person following and mobile camera localization using particle filters

Abstract

This paper investigates visual head tracking and person following with a mobile robot. The position of the mobile robot is determined during person following on the basis of laser readings. The tracking of the head is done using a particle filter built on cues such as color, depth, gradient and shape. The appearance of the head is represented by an ellipse with color histogram in its interior and an intensity gradient along the ellipse boundary. The size of the ellipse is computed from a stereo pair. The localization of the moving camera is achieved using a particle filter framework. A particle filter responsible for localization utilizes laser data obtained from sensor and data obtained from a map representing the environment. Current laser readings are compared with scans coming from a global map in respect of histogram intersection. Experimental results, where a person moves facing the cameras, the robot follows the person and simultaneously determines its position, demonstrate the feasibility of the tracking as well as the localization method.