PCB Micro-Soldering Status Inspection System Research based on Deep Learning

Abstract

In the semiconductor process, the soldering state is one of the important processes. This error can be one of the main causes of fatal effects on other electronic components. Until now, all soldering status have been inspected by humans. This causes many false positive errors. This study experimented with micro-soldering status inspection using artificial intelligence which has 1-stage and 2-stage compound scaling. As a result of the experiment, PCB-soldering condition inspection showed low false positives in 1-stage detector (YOLOv5) unlike other objects.