FENG Hao,TANG Lei,CAO Ligang
(School of Mechanical and Electronic Engineering, Jingdezhen Ceramic Institute, Jingdezhen 333403, Jiangxi, China)

Abstract: Aiming to detect the kiln dirt on ceramic surface, a process was designed with the segmentation of the parallax figures of black glaze spots and bumps based on the principle of binocular vision, the stereo matching algorithm, combined with the features of kiln dirt.The image pre-processing is used to achieve the stereo matching of the image, and the parallax is segmented to find the initial positioning of the dirt. Finally, in the left view, the precise defect edge calculation is conducted. Verification of the test results shows that this method can effectively identify the kiln dirt on the surface of the ceramics.

Key words: computer vision; kiln dirt; stereo matching; defect detection