Image based assistant localization in urban navigation

Abstract

The paper describes such a prototype of a system that can make the urban navigation robust when GPS signal is lost. By making a database of images which usually extracted from texture of building façade and match to the geo-info point in GIS, the system can calculate the pose of a query. Our method is based on SURF (Speed-Up Robust Feature) that can recognize matching facades in two views even if there is change of viewpoint and illumination. In addition Homograph is introduced to estimate the pose of viewpoint. Also, a discussion of this application in urban navigation and a conclusion with the weakness of the system are given in the final.