Vision-based Geo-Localization in Mountainous Regions

Vision-based geo-location is very important for many military applications, especially when the global navigation satellite system (GNSS) is unavailable. However, this is not an easy task and faces many challenges due to the changes of vegetation, illumination and season in mountainous regions. In this work, we regard automatic geo-localization without GNSS recordings as a skyline matching problem, and propose an automated approach for visual localization using digital elevation model (DEM) data and real-captured images. The skyline feature is used as a cue for visual database retrieval. Our system was tested on 50 testing site points within a large-scale area (China, 202.6 km
²
) and an average position error of 43.13 m was obtained within 4.5 s, which demonstrates that this low-cost approach can produce fast and reliable localization results.