Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 02, 2018 - June 05, 2018
This paper describes a new method to recognize the product's defect by using image processing. By using image processing in visual inspection, it is possible to automate the visual inspection which has been done by human eye. This system has three traits. First, this system can get the defect's region by comparing the product's picture with its superior one instead of defect detection which using threshold value. By using this method, it can be strong against external disturbance such as light sources and product color. Second, it uses image processing to judge the defects multifacetedly from the area, strength of the defect, and difference in luminance. Third, this system uses Support Vector Machine to evaluate defect. So, it is possible to set the difference between clear small defects, and indistinct large defects.
