A reliable method for vertical edges recognition

This paper describes a natural landmarks recognition method for the purpose of autonomous vehicle´s self localisation in a dynamic structured indoor environment. The perception system used is composed of a 2D laser scanner providing a 180° scan, and of a monocular vision system composed of a camera CCD equipped with a zoom and mounted on a pan-tilt unit. It is assumed that the robot had a global map of the working area but no a priori knowledge about its current location. In order to quickly and reliably establish the correspondence between observed features and known landmarks, a matching process based on geometric hashing is proposed.