Geospatial photo retrieval for ground-based disaster monitoring

Recently, a cloud-storage service provides a large amount of geotagged photos. However, conventional point-based photo retrieval is inefficient to find significant photos in some operations, such as a disaster monitoring from a large amount of photos. Therefore, photo retrieval is required to improve the efficiency in the disaster monitoring. We focus on azimuth data taken from digital compasses to improve a geotagged photo browsing in a disaster monitoring. Moreover, we have proposed a context-based photo browsing. Additionally, we have focused on a clustering of filtered geotagged photos based on reverse-geocoding. We conducted an experiment using the GPS cameras in two areas. Then, we have confirmed that our algorithm can perform geospatial retrieval for a street-view photo browsing. Moreover, we have suggested revision methodology of the GPS positional disorder.