A Computer Vision Approach for Autonomous Wind Turbine Inspection using a Multicopter
MetadataShow full item record
This thesis studies the mission of autonomous inspection of a wind turbine using a multicopter. Emphasis was placed on recognition and tracking using image processing methods. The Hough line transform was used to extract features of the wind turbine. Hub position was estimated by an algorithm tailored to identify the three-point star resemblance and was tracked by utilizing the Kalman filter. Distance and yaw orientation of the wind turbine were estimated using the pinhole camera model and coordinate transformations. Restricting computational demand was a goal in the program design. Experiments showed accurate position tracking at long range, but with deteriorating performance as range was decreased. Lack of distinctive measurable lengths in the image caused inaccuracy in estimation of distance and yaw orientation. Execution frequency of below 7 Hz was achieved on a single-board computer which was found to be sufficient for reliable control in flight.