Drone Video Analytics at the Mobile Edge
Drone video analytics for ocean search and rescue
Distributed Mobile Edge Computing for Real-time, Near Real-time, and Offline Drone Video Analytics
Drones are increasingly being adopted for several aerial video-based applications such as search and rescue, surveillance, and agriculture. But there are two critical shortcomings of the current drone technologies as described below.
Challenge 1: Untethered drones have limited battery life (less than 30 minutes for many commercial, low-cost drones) and computational capability. Therefore, running video analytics on the drone is nearly impossible.
Challenge 2: Another challenge with current drone technology is the wireless connectivity. Drones usually have one wireless interface (e.g., WiFi or LTE). The bandwidth, range, and coverage of WiFi and LTE limit the amount of video that can be transported to remote edge or cloud server for real-time or near real-time video analytics.
Spectronn's distributed multiaccess edge computing technology solves both these challenges. The multi-network SiFi-200 router with an integrated video management system (VMS) connects to two or more wireless links (e.g., LTE from two different mobile operators, LTE and satellite, LTE and WiFi, etc.) concurrently. A policy controller allows the drone user to choose a primary and a back-up link or simultaneous connections to two or more available wireless links.
The SiFi-200 multi-network router dynamically chooses the wireless link to transport the drone video content to a remote edge or cloud server (e.g., AWS, Azure, or Google), depending on the wireless link characteristics such as throughput and latency. The integrated VMS processes the drone video stream depending on the wireless link(s) choice.
A few AI-driven video analytics computations are run locally on Spectronn's GPU-based mobile edge computing device (see below) connected to the SiFi-200 (see figure on the right) router on the drone while other computations are intelligently offloaded to another edge computing device or to the cloud on the ground using optimal wireless link(s). Session persistent offloading ensures that the video stream and computations are uninterrupted even if the wireless links are unreliable or unavailable.
The distributed mobile edge-cloud computing platform is scalable, rapidly deployable, and software-driven. Therefore, this platform presents an unprecendented, plug-and-play capability for drone video analytics. All types of drones are supported by this technology.
APIs make it easy to develop and deploy your own drone video analytics applications easily.
SiFi-200 multi-network router + VMS
Spectronn's drone-based video computing GPU-device