Finding hotspots by location-direction-enabled photographs focusing on intersection of lines of sight

Abstract. We propose a method to extract a set of viewpoint locations and a set of locations of interest from a given set of location-direction-enabled photographs having embedded information regarding the location and direction of the shot. A viewpoint location is a location from which many photographs are taken. A location of interest is a location that that many photographs are taken of. When extracting locations of interest, a location-direction-enabled photograph does not embed the location of its subject; this is a technical difficulty. Therefore, we treat location-direction-enabled photographs as a point-with-directions and propose a novel clustering method for them. The evaluation results show that the method improves precision and recall for extracting locations of interest compared to an existing method. We also show that the proposed method can extract locations of interest that the existing method cannot extract.