A Visual Inspection System for Surface Mounted Devices on Printed Circuit Board

The object of this study is to develop a more reliable and faster visual inspection system for printed circuit board inspection. In order to reach this goal, the inspection process was divided into two stages, namely, screening stage and classification stage. In the first stage, only one image feature is abstracted from the examined image and is used as a screening index to quickly screen out most normal components fast. In the second stage, neural networks are used to integrate all image feature information available to more precisely inspect those left after the screening test. Since there are numerous image features available, the way to select proper image features also worth of discussion. In this study, parting coefficient is used as an index for selecting proper image features. The proposed system is trained by a set of revised image data first. Image collected from production line were then used to test the trained system. Experimental results show the feasibility of the proposed system