Skyline-based Positioning in Urban Canyons Using a Narrow FOV Upward-Facing Camera

This paper presents an approach for solving the localization problem using skylines obtained from a 3D city model and from an upward-facing camera. A visual spectrum narrow field-of-view camera similar to that available on most mobile phones is used. The proposed skylines-based positioning method uses vanishing points to determine and correct for the pitch and roll observed when acquiring images from a camera rigidly attached on the top of the vehicle. The corrected images are then segmented into sky and non-sky areas. The obtained binary images are then compared with the ideal images synthesized from the 3D building model. The position solution estimate corresponds to the best match obtained between the observed and synthesized images. The method is evaluated using real camera data collected in downtown Calgary, Alberta, Canada. The results show horizontal position errors ranging from few centimeters to 10 metres in most cases with some outlines caused by external errors such as buildings missing from the 3D city model and segmentation errors.