Efficient video streaming of 360° cameras in unmanned aerial vehicles. An analysis of real video sources

Video streaming data acquired by Unmanned Aerial Vehicles is an innovative service that will be leveraged by several applications ranging from entertainment and surveillance to disaster recovery. 360 cameras provide unprecedented visual information and enable services to a novel level of immersive experience. However, 360 video sources are not still fully characterized, and this holds especially true for drone mounted 360 video sources. This paper presents a thorough analysis of the video traffic associated to several 360 camera sequences, acquired by a pedestrian held camera as well as by a drone mounted camera in various environments and lighting conditions. A fine-grained rate distortion analysis is presented for both video frames and video chunks, thus making this study relevant for HTTP-based video streaming services. The analysis is completed by making publicly available a dataset of 360 video traffic traces that can be used for numerical simulations of Unmanned Aerial Vehicles providing 360 video streaming services.