Computer Vision Based Object Detection and Tracking in Micro Aerial Vehicles

Abstract

­­­­The ultimate goal of Computer Vision is to instruct a computer to understand and interpret visual signals and images in real time and to instruct a computer to react to the environment around them. In this work, we describe a system that allows a micro aerial vehicle (MAV), equipped with an onboard camera, to detect and track a moving target object. In an alternative implementation, the MAV instead searches the environment for the target object and flies to it. Due to the limited capability of the drone’s integrated processor, image processing is performed by a ground-based computer that also determines the necessary flight corrections and communicates them to the vehicle. The complete system, comprised of the MAV, off-board computer, and software, operates autonomously, a necessary condition for many of the applications for which such systems may be useful.