Vision-based precision vehicle localization in urban environments

Abstract

Global localization is a challenging problem in which autonomous vehicle has to estimate the self-position with respect to a priori map using perception results. In this paper we present a vision-based localization method for autonomous Vehicles in urban environment. The localization process consists of two stages: coarse localization using topological map and fine localization using metric map. The topological map represented by the holistic image feature provides coarse location, whereas localization from metric map is relatively slow, but more accurate. It is possible to integrate the two stages to make precise and reliable localization. The location system has been tested on autonomous vehicle in outdoor environment. The results show that our method is efficient and reliable.