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

Tvio approaches to detect defects on printed circuit board had been evaluated. One is the direct comparison of the tested image with a template image. Before the comparison, we adopted an interpolation method to reconstruct the test image such that the orientation and position of components shown on the test image are the same as those on the template image. The second approach is using image features to detect and classify defects. We proposed a two steps inspection scheme. The inspection system is divided into the screening stage and the classification stage. The object of the screen stage is to quickly screen out most normal components to reduce overall processing time. Only one image feature is used as the screen index. At the classification stage, the neural networks were adopted to integrate all image feature information available to more precisely classify those fail to pass the screening test.