ZHOU Jiang，REN Kun，SHUAI Ying-qi，CHEN Ying-hao

（School of Mechanical Engineering and Automation，Zhejiang Sci-Tech University，Hangzhou 310018，China）

Abstract：Aiming at low manual detection efficiency，high rate of false detection and high labor intensity，defect types and characteristics of magnetic steel sheet were researched，and the process of detection was analyzed，an algorithm based on machine vision was proposed to achieve automated defect detection of magnetic steel sheet. The area of the target region，obtained by means of the closing operation and binarization，was compared with the template to select missing corner and adhesion defects. The mean filter was used to smooth the initial

image，the gray value difference operation was utilized between the resulting image and the template image，and then the region with gray value more than a predetermined value was selected to mark the defect of crack or crystallization. The results indicate that the algorithm can determine the defects of missing corner，adhesion，crystallization and crack with 10%higher in accuracy than manual detection， improved the detection efficiency and reduces labor。

Key words：machine vision；binarization；closing operation；mean filter；defect detection；magnetic steel