Travelmedia: An intelligent management system for media captured in travel

A media management system exploiting characteristics of travel media is designed to facilitate efficient management and browsing. According to travel schedules, travel media often have implicit thematic structure. Correlation between different modalities also provides implicit cues to media analysis. In this system, we exploit techniques of near-duplicate detection to select representative photos, and determine region-of-interest in photos to enhance browsing experience. For face–name association, a face clustering module based on visual language models is constructed. To systematically segment travel videos of bad visual quality and significant motion, we explore correlation between photos and videos based on approximate visual word histogram matching. Experimental results demonstrate the effectiveness of the proposed approaches and show that they are practical functions.

Research highlights
► We integrate techniques specially designed for analyzing travel media.
► Studies in terms of what, where, and who are conducted.
► A vector space model is proposed for travel media management.
► More flexible methods are used to release constraints in previous works.