Matching panoramic image for vision-based robot localization

Abstract

The human ability to localize, build mental maps of environments and navigate in our daily lives has inspired us to develop a system for mobile robots to navigate in unknown environments. During our daily lives, we primarily use vision to recognize landmarks and localize ourselves. In this paper, we therefore introduce a vision based localization, mapbuilding and navigation system. We focus on image matching algorithm for panoramic images and how they are used in localization of a mobile robot. We provide details of our algorithm and present a crtical evaluation based on performance in real-time experiments.