Skyline positioning in urban areas using a low-cost infrared camera

This paper assesses the accuracy with which a user's position can be computed using skyline information, that is, the shape of the sky observed looking upward from the user's location. Whereas this concept has been proposed in the past using wide field-of-view (i.e., fisheye) cameras, this study uses a narrower field-of-view camera that is more commonly available on mobile devices. For testing purposes, an infrared camera was used to more easily separate the sky from surrounding buildings. Images are compared against a database of ideal skylines computed from a 3D building model and the database position whose skyline most closely matches the observed images is selected as the user's final location. Results indicate that median position errors of 10-14 m are possible depending on whether the 3D model contains errors and whether user heading is available.