Battery fabrication requires strict quality assurance that can partly be done by image recognition in a production line. In this project, we trained and tested a classifier to recognize different optically visible defects like cracks and holes. The classifiers were trained and tested for different image modalities and resolutions. Defects were quantified and compared for different unit cell electrode production methods.
SINTEF was responsible for machine learning, image analysis and contributed to optical set-up design and image acquisition.