Distributed visual servoing of a mobile robot for surveillance applications

Abstract

A mobile robot intercepts a target identified in the image plane of a fixed camera without prior knowledge of camera viewpoint or geometry. The target may be identified manually or autonomously. When the robot moves within the camera image plane, the system develops a table of mappings between the image plane and the ground plane in which the robot moves. The robot is identified in the image plane via background subtraction and localized in the ground plane using a particle filter. This adds the potential for robotic intervention to surveillance using fixed camera systems. 1