Geolocation of a panoramic camera by reference pairing

Abstract

Panoramic cameras are now available to a large audience. They provide good results on photogrammetry application, but they are still limited by their positioning. This project aims to geolocate a commercial 360° camera in an urban environment, by extracting points in fisheye images and match them with reference from a LiDAR (Light Detection and Ranging) dataset. Such reference points are located on the horizon line, visible from the camera point of view. Matching points are then introduced as Ground Control Points to improve the camera positioning accuracy. A fully automatic solution for position refinement, based on LiDAR data is proposed in this paper.