| dc.description.abstract | This project has been performed in cooperation with Rolls-Royce Marine. They have
a new electric motor soon hitting the masses, and there is a hope that the magnets
used in this motor can be automatically quality controlled with computer vision. This
thesis takes on that challenge, and investigates possible technologies that can be used.
Multi-view 3D reconstruction, laser stripe scanning and 2D image comparison techniques
has been identified as promising candidates. Requirements for the magnets have been
investigated, and the solutions that showed potential have been implemented. Optical
character recognition has been identified as an important part of an automated quality
control solution, as one need to register the serial number of each magnet. This has
therefore also been implemented.
The different implementations have then been tested and analyzed. It is concluded that
while detecting chips from the corners and edges of the magnet could be possible with
either laser scanning or 2D image comparison, the 2D image comparison is simpler and
less computationally intensive, and therefore better suited for mass production. The tests
where not able to safely detect the damages to the surface, but this might be possible
with a better camera. Further work will be needed to conclude on whether this will be
possible. Cracks where not detected with any of the solutions, and is likely the most
difficult damage to detect. Due to the many difficulties encountered, a proposal has been
made for a half-automated solution, where a human does visual inspection, the tolerance
of the magnet is ensured automatically with a funnel system, and the serial number gets
automatically scanned.
For all implementations there are potentials for improvement. Trying different and better
suited filtering techniques can improve the results of both the laser scanning, the 2D image
techniques, and the optical character recognition. Using a better camera will also improve
the results, as the camera that has been used has low resolution, high levels of noise, and
lacks control of the auto focus, white balance, and aperture. If a 3D solution is chosen,
one will also need a solution for how to identify a particular damage in the generated point
cloud, as this has not been investigated in this report. | en |
