AMIGO: accurate mobile image geotagging

With location-based services gaining popularity among mobile users, researchers are exploring the way using the phone-captured image for localization as it contains more context information than the embedded sensory GPS coordinates. We present in this paper a novel mobile image geotagging approach to accurately sense the actual geo-context of a mobile user. The proposed approach, named AMIGO (Accurate Mobile Image GeOtagging), is able to provide a comprehensive set of accurate geo-context based on the current image and its associated scene in the database. The geo-context includes the real locations of a mobile user and the scene, the viewing angle, and the distance between the user and the scene. Specifically, we first perform partial duplicate image retrieval to select crowdsourced images capturing the same scene as the query image. We then employ the structure-from-motion technique to reconstruct a sparse 3D point cloud of the scene. Finally, by projecting the reconstructed scene onto the horizontal plane, we can derive user's location, viewing angle, and distance. The effectiveness of AMIGO has been validated by experimental results.