Efficient storage and retrieval of geo-referenced video from moving sensors

Abstract

The general use of mobile sensors equipped with video, GPS, and compass is gaining momentum. These sensors continuously stream geo-referenced videos, which provide enormous amounts of data. Efficiently searching through these long-duration video corpuses based on location and time is a challenging task. We propose a novel approach to efficiently store frequent timestamped geographical tags in a spatial database to enable searching the video files within a short response time. Further, we enable retrieving segments of the full video that satisfy certain spatio-temporal conditions in a ranked manner. Our approach achieves short response times as compared to standard methods for saving geographic tags in a spatial database. © 2013 Authors.