Multisensor Integration for Human-Robot Interaction

摘要

In this paper we present a solution to human-robot interaction using a combination of visual and laser range information. Human legs are extracted from a laser scan and, at the same time, faces are detected from the camera's image. The information is integrated in a detection procedure that returns direction and distance of the surrounding people. This is eventually used by a mobile robot to approach and start interaction with humans. Unlike other similar applications, our solution works well in real-time even under limited computational resources. Experimental results show good performances of our system.