作者: Matthew Clements , Avideh Zakhor
DOI: 10.1109/ICIP.2014.7025681
关键词:
摘要: Image geo-localization is an important problem with many applications such as augmented reality and navigation. The most common ways to geo-localize image are use its meta-data GPS or match it against a geotagged database. When neither of those available, still possible apply shadow analysis determine the camera heading for outdoor images. This could be useful pruning search space in applications, example by removing roads incompatible orientations from database Open Street Map. In this paper, we develop novel interactive method deducing global query using shadows it. We start constructing model sun-earth system all at given approximate latitude, compare within under range headings. demonstrate on 54 images known ground truth, show that 52 cases truth lies computed range.