Analysis and detection of ceramic-glass surface defects based on computer vision

The whole ceramic-glass manufacturing process is basically operated automatically. The final stage of the process concerned with visual inspection, i.e. product quality inspection, is closely related with computer vision, image processing and pattern recognition. We propose methods to detect surface defects of the ceramic-glass based on digitized images. The thresholds were used to gain binary images. Markov random field models were fitted to binary textures. Finally the experiments carried out on factory samples were used to verify the feasibility of these methods.