Image-based Localization in Urban Environments

摘要

This report describes an efficient algorithm to accurately determine the position and orientation of a camera in an outdoor urban environment using camera imagery acquired from a single location on the ground. The requirement to operate using imagery from a single location allows a system using our algorithms to generate instant position estimates and ensures that the approach may be applied to both mobile and immobile ground sensors. Localization is accomplished by registering visible ground images to urban terrain models that are easily generated offline from aerial imagery. Provided there are a sufficient number of buildings in view of the sensor, our approach provides accurate position and orientation estimates, with position estimates that are more accurate than those typically produced by a global positioning system (GPS).