Autonomous robot navigation based on clustering across images

Abstract

The aim of this paper is to present a novel approach to the autonomous robot navigation based on clustering of image descriptors. The descriptor called Speeded-Up Robust Features (SURF) is a scale- and rotation-invariant detector, which can visually navigate a robot in a large outdoor and indoor environment. By incorporating several clustering methods, which are derived from fuzzy set theory and inspired with biological background, such as Fuzzy C-mean (FCM), Kohonen's Self-Organizing Map (SOM) and Neural Gas algorithm (NG), we detect center positions of natural clusters crosswise the recorded images. Center positions represented by vector prototypes are used as reference points in the decision making of the robot navigation. © 2011 Springer-Verlag.