LIANG Dong, Gu Jiening, DING Li, ZHANG Chen

(College of Mechanical Engineering, Jiangsu University of Technology, Changzhou 213001, Jiangsu, China)

Abstract: A machine vision-based appearance detection system is proposed for automatically detecting the appearance defects of wet-pressed ferrite magnets. Firstly, the system structure and electrical control part are introduced. Then, according to the characteristics of crack noise on the surface of wet-pressed ferrite magnets, multi-scale gray transform is used to enhance the contrast of feature domain, and fast discrete Fourier transform is used to locate the defect accurately. Finally, the image is segmented by hard threshold, and the accuracies of gray morphological filtering and soft morphological hybrid filtering are compared. Experiments show that the soft morphological hybrid filter is more suitable for multi-texture defect recognition of wet-pressed ferrite magnets.
Key words: wet-pressed ferrite magnet; multiscale gray level change; discrete Fourier transform; soft morphological hybrid filter