Collective Discovery of Geographic Locations of Frequently Photographed Objects Only using the Metadata of Digital Photographs

摘要

In this paper, we propose an algorithm to classify a set of digital photographs by location of frequently photographed objects. The location is estimated by collective intelligence approach based on a collection of intersections of camera vectors of the metadata which a photograph records. In emergency, a rescue team needs information about the road to its disaster area as quick as possible. One of the best solutions for this purpose is an information sharing system between the refugees and the rescue team who share images taken by the refugees. For building this system, the following 2 difficulties are known. Firstly, this system should not inhibit refugee's evacuation behavior while it engages in this sharing system. Secondly, the huge amount of disorganized images received is useless for the rescue team who has not enough time and other resources. The images should be categorized quick and adequately. In this paper, we propose a new classification method of images by geographic location of objects taken by them by collective intelligence. We only use metadata of digital image for this quick classification, namely, time to shoot, the latitude, the longitude and the bearing of its camera. Our method can accept consumer' s digital camera and smartphone which has a low end GPS unit and digital compass. Also, refugees need not to input any more information about objects taken than the photograph while name or location of objects taken have to be inputed in traditional works. In our method， the location of objects taken is estimated by an algorithm based on intersections of camera vectors automatically. The algorithm is confirmed by a series of experiments which use actual photographs taken by a broadly using smartphone.