Autonomous Inspection of Wind Turbines and Buildings using an UAV
This thesis describes the use of optical flow and Hough transform in local navigation,
particularly the case of inspecting wind turbines and buildings for damages. The goal is
to create an observer capable of estimating the metric velocity and distance to the blade from scaled velocity inputs. Furthermore, a controller capable of inspecting wind turbines and buildings autonomously will be presented and tested.
The first part of the report will address the use of two different computer vision algorithms in local navigation. Firstly, the optical flow algorithm will be presented, with focus on the Horn-Schunck and the Lucas-Kanade method. By using a pyramidal representation of the image, the algorithms were found to provide more accurate optical flow when the motion is large. Secondly, the Hough transform for finding straight lines in an image was investigated. The tests showed that Hough transform can be used on wind turbine blades to estimate the desired velocity vector and the relative angle between the UAV and the blade.
The observer was simulated in MATLAB with velocity vectors provided by an optical flow
algorithm. Two different case studies has been investigated to verify the mathematical
model and observer. The observer was able to successfully estimate the metric velocity
along with the distance.
The guidance law presented in this report was based on the pure pursuit guidance and
PID controllers. The controllers successfully maneuvered the UAV to the desired position
and kept a constant distance to the object. The height controller was tested with both a
stationary and dynamic desired height. The UAV was able to follow the desired height
in both cases.
Place, publisher, year, edition, pages
Institutt for teknisk kybernetikk , 2014. , 105 p.
IdentifiersURN: urn:nbn:no:ntnu:diva-25877Local ID: ntnudaim:10706OAI: oai:DiVA.org:ntnu-25877DiVA: diva2:742048
Johansen, Tor Arne, ProfessorKlausen, Kristian