Color-defect classification for printed-matter visual inspection system

An automatic defect classification method that pays attention not only to shape-defects but also to color-defects as well is proposed for a printed-matter visual-inspection device. The required characteristic of the shape-defect classification is to have an automatic select function of the feature variable which removes the unnecessary characteristics-parameter space to use few learning data efficiently. The characteristic of the color-defect classification is to identify an abnormal printing-device roller such as rotary press based on the color difference signal with the background color and the UCR process. We evaluated the proposed method with good result.