AUTOMATICS HOT SPOT DETERMINATION FROM ONLINE IMAGE COLLECTIONS

摘要

In this paper, an automatics hot spot determination approach based on online image collections is introduced. The basic idea is that each user-uploaded photo with geographical identification metadata is regarded as a vote for hot spot selection. The photos are grouped into several sets by a spatial grouping approach. Based on the voting concept, each grouped photo set represent a hot spot, i.e., food spot or scenic spot. In addition, a representative photo can be selected from each photo set by an image classification algorithm. To demonstrate the feasibility and applicability of the proposed approach, several millions of photos downloaded from online photo-sharing communities are tested and the selected photos representing scenic spots are used in a digital tourist map.