Open Collections

Autonomous navigation of an unmanned aerial vehicle using infrared computer vision

University of British Columbia

The goal of this thesis is to develop infrared (IR) vision-based navigation methods for small unmanned aerial vehicles (UAVs). As small unmanned aerial systems (sUAS) see increased use, this technology promises to benefit many civilian applications including agriculture, fire suppression, search and rescue, as well as military operations. Small UAVs have proliferated the market, ranging from pocket-sized photography drones controlled using a smartphone, to larger systems used in remote surveying and disaster response applications. UAVs share many commonalities with ubiquitous smartphones, including a powerful on-board processor or Central Processing Unit (CPU), gyroscope, compass, barometer, GPS, and electro-optical colour camera. Incorporating vision-based navigation methods in a UAV autopilot makes use of the on-board camera and processor (often standard features on drones) for precise localization and navigation, in environments where other sensors may fail or prove inaccurate. This work presents two types of high-level altitude and position control systems based on computer vision. Specifically, both navigation systems use IR technology; the first method uses an active IR beacon and computer vision methods to coordinate autonomous localization and landing of a small quadrotor UAV on a stationary platform. The autonomous IR-based landing method was tested using the AR.Drone quadcopter and a computer vision and control script developed in Python. The second method uses passive short-wave infrared (SWIR) and the Normalized Difference Vegetation Index (NDVI), used in remote sensing applications, to coordinate navigation of a small UAV in a GPS-denied indoor greenhouse environment. This method was integrated with a NAVIO2 autopilot to determine the UAV’s position and orientation in a greenhouse row. The real-time image processing pipeline includes machine vision principles such as binarization, thresholding, edge detection, and line detection. Both methods presented in this thesis use an eye-in-hand approach to the image-based visual servoing problem to navigate and land the UAV.

Text

2019-04-20

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/65663

Electrical and Computer Engineering

University of British Columbia

UBCO

http://creativecommons.org/licenses/by-nc-nd/4.0/