External cameras and a mobile robot: A collaborative surveillance system

Abstract

This paper presents a system that surveils an indoor environment through the collaborative use of overhead cameras and a mobile robot. The robot is localized both in the image plane of the external camera and a ground plane map of its environment. This simultaneous localization is used to build up a homography between the image plane of each external camera and the ground plane. The robot begins the calibration procedure by moving through a series of pre-determined waypoints in the environment. As it comes into view of each camera, it uses visual servoing to orbit the centre of the camera image and keeping within its eld of view whilst avoiding obstacles. The system builds up a transformation matrix between each camera and the ground plane and proceeds to patrol the environment. When an intruder is detected in any of the external cameras, the system uses the automatically determined homography matrices to calculate the ground plane position of the person and sends out the robot to intercept the intruder. Experiments are conducted with 2 external cameras in this paper but could easily be extended to N cameras.