Urban Structure Estimation Using Orthogonal and Parallel Lines

Abstract

In urban areas, buildings will block the line-of-sight of GPS signals while mobile robots are navigating through an environment which will result in localization errors of the robot. Robot localization using camera sensor is another way for exploring the surrounding area. Depth uncertainty in camera images could affect the accuracy of the estimated structure of the surrounding environment. Man-made structures contain many parallel and orthogonal lines. In this study, I use parallel and orthogonal lines in the scene to improve the estimation of the environment. The proposed method improves structure estimation compared with the standard re-projection error method.