Monocular-based pose estimation using vanishing points for indoor image correction

摘要

This paper assesses the estimation of the orientation of a camera rigidly attached to a vehicle body using vanishing points and then uses the estimated rotation to rectify images for navigation in indoor environments. The approach is facilitated by the fact that many indoor environments are rich in vertical and horizontal lines that are used for detecting vanishing points. The performance of the monocular-based navigation is evaluated as well as the image rotation rectification. Results of the proposed methods show that using a single image, we are able to obtain the camera, and thus the platform, 3D orientation as well as accurately correct the image for camera inclination.