Safe robot navigation using an omnidirectional camera

The paper presents a safe robot navigation system based on omnidirectional vision. The 360 degree camera images are analyzed for obstacle detection and avoidance and of course for navigating safely in the given indoor environment. This module can process images in real time and extracts the direction and distance information of the obstacles from the camera system mounted on the robot. This two data is the output of the module. Because of the distortions of the omnidirectional vision, it is necessary to calibrate the camera and not only for that but also to get the right direction and distances information. Several image processing methods and technics were used which are investigated in the rest of this paper.

I. Introduction

In recent years visual guided navigation has become very popular in robotics. More recently, omnidirectional cameras have been used for it. These cameras have a benefit that can be used for an effective way to detect the 360 degree environment around the robot. So if the robot is moving backward the system can detect not only the fixed but also the new, moving objects and obstacles behind the robot. In several fields of robotics lots of sensors, lasers, stereo cameras, landmarks are used to help and precise the navigation, but in this work only an omnidirectional camera is equipped on the machine. One drawback is that images are affected by strong distortion and this problem must be solved. The navigation algorithms are designed for indoor environments, so there are three main object classes that can be detected and separated. These are the floor, the walls, and other obstacles.