Vision based navigation using Kalman approach for SLAM

Abstract

The aim of this paper is to present a vision based algorithm for simultaneous localisation and map build-ing (SLAM) for indoor environment. The approach is based on the well known predictor-corrector structure of the Extended Kalman Filter and makes use of vi-sion data in the correction phase. To this aim a low cost vision system is presented, with a fast image pro-cessing procedure for features extraction, based on the recognition of artificial sources of light in an office-like environment. Experimental tests, carried out on a robotic wheelchair equipped with the vision system, show the satisfactory performance of the mobile plat-form under realistic operating conditions. 1