GeoSearch: Georeferenced Video Retrieval System ∗ Youngwoo Kim, Jinha Kim, HwanjoYu † Department of Computer Science and Engineering Pohang University of Science and Technology(POSTECH) Pohang, South Korea {ywkim10, goldbar, hwanjoyu}@postech.ac.kr ABSTRACT Conventional video search systems, to find relevant videos, rely on textual data such as video titles, annotations, and text around the video. Nowadays, video recording devices such as cameras, smartphones and car blackboxes are equipped with GPS sensors and able to capture videos with spatiotemporal information such as time, location and camera direction. We call such videos georefer- enced videos. This paper presents a georeferenced video retrieval system, GeoSearch, which efficiently retrieves videos containing a certain point or range in the map. To enable a fast search of georeferenced videos, GeoSearch adopts a novel data structure MBT R (Minimum Bounding Tilted Rectangle) in the leaf nodes of R-Tree. New algorithms are developed to build MBT Rs from geo- referenced videos and to efficiently process point and range queries on MBT Rs. We demonstrate our system on real georeferenced videos, and show that, compared to previous methods, GeoSearch substantially reduces the index size and also improves the search speed for georeferenced video data. Our online demo is available at “http://dm.hwanjoyu.org/geosearch”. Categories and Subject Descriptors H.2.4 [Database Management]: Systems—Query processing; H.2.4 [Database Management]: Systems—Multimedia databases; H.2.8 [Database Management]: Database Applications—Spatial databases and GIS General Terms Algorithms, Performance Keywords Georeferencing, Video search, Spatial Indexing ∗ This work was partially supported by the Brain Korea 21 Project in 2012 and Mid-career Researcher Program through NRF grant funded by the MEST (No. KRF-2011-0016029). This work was also supported by IT Consilience Creative Program of MKE and NIPA (C1515-1121-0003). † corresponding author Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. KDD’12, August 12–16, 2012, Beijing, China. Copyright 2012 ACM 978-1-4503-1462-6 /12/08 ...$15.00. 1. INTRODUCTION With the rapid popularization of video recording devices, not only the media made by professionals but also made by ordinary users called UCC (User-Created Contents) are taking a large por- tion in the web. Since identifying related video requires under- standing of video contents, which is a very heavy and complex process, current video search systems mostly rely on textual data such as video titles, annotations, and text around the video. Nowadays, video capturing devices such as cameras, smartphones and car blackboxes are equipped with GPS sensors and able to cap- ture videos with spatiotemporal information such as time, location and camera direction. We call such videos georeferenced videos. For some applications, such spatiotemporal information plays key roles in querying georeferenced videos. For example, some people may want to find videos of specific event that occurred at specific time and location, e.g., videos of a traffic accident captured by car blackboxes, videos of a goal scene in a soccer play captured by users, or videos of a concert with celebrities captured by users. This paper presents GeoSearch, a georeferenced video retrieval system, which efficiently retrieves videos containing a certain point or range in the map. GeoSearch adopts a novel data structure MBT R (Minimum Bounding Tilted Rectangle) in the leaf nodes of R-Tree, in order to enable a fast search of georeferenced videos, While traditional MBR (Minimum Bounding Rectangle) is popu- larly used to describe an area in spatial indexes such as R-Tree, MBR is not suitable for describing the areas of moving scenes where its location and direction are continously changing. MBT R is designed to describe an area (or scenes) of moving location and direction. New algorithms are developed to build MBT Rs from georeferenced videos and to efficiently process point and range queries on MBT Rs. We demonstrate our system on real georef- erenced videos, and show that, compared to previous methods, by adopting MBT R in the R-Tree, GeoSearch substantially reduces the index size and also improves the search speed for georeferenced video data. Our demo is available at “http://dm.hwanjoyu.org/geosearch”. Due to space limitation, this paper focuses on a high-level overview of techniques and demonstration. Technical details are described in our technical report [10]. 1.1 Related Work Georeferenced Multimedia Search: There are several approaches to the use of location information in image search. Toyama et al [21] used metadata in image search and developed a database for it. They enabled spatial image search by recording images with longitudinal and latitudinal coordinates and timestamps. Several commercial websites, such as Flickr and Woophy [1], also provide georeferenced image search. They focus on showing static images 1540