Simultaneous People Tracking and Recognition with a Service Robot

The research proposed in this book investigates new methods for service robots to detect people using sensor data fusion, and to estimate their position and identity simultaneously. An efficient multisensor solution for human detection is presented, where a mobile robot uses its on-board camera and laser to search for faces and legs in the environment. A new tracking algorithm based on the Unscented Kalman Filter is developed to integrate sensor measurements and estimate the human position. More specifically, after a data association stage that assigns faces and legs to their respective human targets, observations are fused sequentially to track multiple people while these and the robot are either moving or not. Finally, clothes and face recognition algorithms are integrated in a new probabilistic framework for simultaneous tracking and identification of humans. The solution is based on a bank of Bayesian filters that combines laser and visual information to generate a multiple hypothesis estimate of the person being tracked. The proposed system is also applied to a receptionist robot for university open days.

Autorentext

Dr Nicola Bellotto is a research assistant in the Active Vision Lab at the University of Oxford. He holds a PhD in Computer Science from the University of Essex and his research interests include mobile robotics, sensor fusion and Bayesian estimation. Prof Huosheng Hu is with the University of Essex, where he leads the Human Centred Robotics Group.