An efficient index structure for large-scale geo-tagged video databases

Abstract

An unprecedented number of user-generated videos (UGVs) are currently being collected by mobile devices, however, such unstructured data are very hard to index and search. Due to recent development, UGVs can be geo-tagged, e.g., GPS locations and compass directions, at the acquisition time at a very fine spatial granularity. Ideally, each video frame can be tagged by the spatial extent of its coverage area, termed Field-Of-View (FOV). In this paper, we focus on the challenges of spatial indexing and querying of FOVs in a large repository. Since FOVs contain both location and orientation information, and their distribution is non-uniform, conventional spatial indexes (e.g., R-tree, Grid) cannot index them efficiently. We propose a class of new R-tree-based index structures that effectively harness FOVs' camera locations, orientations and view-distances, in tandem, for both filtering and optimization. In addition, we present novel search strategies and algorithms for efficient range and directional queries on FOVs utilizing our indexes. Our experiments with a real-world dataset and a large synthetic video dataset (over 30 years worth of videos) demonstrate the scalability and efficiency of our proposed indexes and search algorithms and their superiority over the competitors.

Feature-centric ranking algorithms for georeferenced video search

“…However, these studies focused on indexing the relative location of objects in the video rather than the geographic placement of the video as such. Lu et al [31] introduced R-tree-based indexes for location, orientation, and distance information of Fields of Views (FOVs). Gilboa-Solomon et al [8] looked into e cient storage of time-stamped geographical tags in a spatial database.…”

Querying geo-tagged videos for vision applications using spatial metadata

“…Kim et al [13] represented the geo-videos as a vector model. Other studies [16,18,21,25,31,34,36] mainly focused on geo-video indexing and query processing. Navarrete et al [25] utilized R-trees [7] and grid files to index the camera locations of videos.…”

“…These two studies [25,34] treated videos as points. These four studies [16,18,21,31] treated videos as FOV objects. Ay et al [31] indexed FOV objects with Rtree.…”

“…Flora et al [6] used polygons to represent the coverage of video segments. Lu et al [18,19] proposed an orientated R-tree (OR-tree) which considers both camera locations and shooting directions during building the index. While these work provide the ability to filter media content based on geospatial information, it does not take into account the specific requirements of vision applications.…”

Video Data Model and Retrieval Service Framework Using Geographic Information

“…Recent works focus on automatically acquiring sensor streams for large‐scale geo‐referenced video organizing, indexing and searching (Kim et al ; Ay et al ; Ma et al ). Lu et al () designed an R‐tree index structure for range and directional queries in the field‐of‐view of videos. The Open Geospatial Consortium (OGC) also defined a view cone model for video frames, and designed a geo‐video web service based on the simple object access protocol (SOAP) (Lewis ).…”