Testbed Development and Simulation Study of a Visually Controlled Quadrotor

Abstract

Machine vision and image processing receive more and more attention and are applied to the Unmanned Aerial Vehicles research and industry in the last decade Indeed most of the unmanned quadrotors are equipped with an onboard camera which provides beautiful aerial pictures as well as the information about the drone orientation and positioning This study is the first step towards an Arduino-based quadrotor to perform autonomous flight by vision feedback The work is a prequel to the actual flight of the drone regarding that the image processing for navigation is tested on a XY table The XY table acts as a testbed for 2 dimensional flight of a static attitude A Human-Machine Interface is designed to supervise the data from the sensors and visual image from the wireless camera Two operation modes are studied The first mode allows the user to assign a destination by clicking the target on the image frame received The navigation command then can be determined to guide the drone to that destination point where the destination appears on the image frame center The second mode permits to reach coordinates specified by the user by targeting sequential points The image matching processing necessary to recognize the same point in real time image frames is based on the Speed-Up Robust Features algorithm which gives accurate and rapid results with K-nearest neighbors matcher The efficiency of the image processing algorithm and navigation strategy is demonstrated by the testbed

Date of Award	2014 Aug 20
Original language	English
Supervisor	Chieh-Li Chen (Supervisor)