GeoPhotoHunt: A Mobile Application for Playing Photo Treasure Hunts and Suggesting Tourist Tours

This paper presents a novel mobile application for the implementation on Android mobile devices of a photo treasure hunt. GeoPhotoHunt (GPH) provides services for creating/playing a photographic treasure hunt as a guide for city visiting or for recreational purposes. The application uses computer vision algorithms executed onboard to quantify similarity between the picture captured by the phone camera and the correct image, without requiring a remote server and an internet connection: this is a main requirement for tourists who often pay a lot of money for international data roaming. Both creation and playing phases are performed on the mobile device in order to facilitate the use on site. The image recognition module on the mobile device manages the feature extraction and the calculation of similarity to the stored objects. In this paper we report a rather broad analysis of existing works in the domains of location-based pervasive games applications and image recognition using limited resources. Moreover, in order to select the best image similarity method, we perform a wide comparison of descriptors used to represent images and measures used to evaluate their similarity. Our experiments carried out on 7 image datasets prove that the system is efficient and robust in comparing images with different characteristics