Efficient Indexing and Querying of Geo-tagged Aerial Videos

Driven by the advances in control engineering, material science and sensor technologies, drones are becoming significantly prevalent in daily life (e.g., event coverage, tourism). Consequently, an unprecedented number of drone videos (or aerial videos) are recorded and consumed. In such a large repository, it is difficult to index and search aerial videos in an unstructured form. However, due to the rich sensor instrumentations of drones, aerial videos can be geotagged (e.g., GPS locations, drone rotation angles) at the acquisition time, providing an opportunity for efficient management of aerial videos by exploiting their corresponding spatial structures. Each aerial video frame can thus be represented as its spatial coverage, termed aerial Field-Of-View (aerial-FOV). This effectively converts a challenging aerial video management problem into a spatial database problem on aerial-FOVs. In this paper, we focus on efficient indexing and querying of aerial-FOVs. Unfortunately, aerial-FOVs are shaped in irregular quadrilaterals, and this renders existing spatial indexes inefficient to index aerial-FOVs. Therefore, we propose a new index structure called TetraR-tree that effectively captures the geometric property of aerial-FOVs. Based on the TetraR-tree, we develop two novel search strategies to efficiently process point and range queries on aerial-FOVs. Our experiments using both real-world and large synthetic video datasets (over 30 years' worth of videos) demonstrate the scalability and efficiency of our proposed indexing and querying algorithms.