Automated PCB inspection in small series production based on SIFT algorithm

There has long been a concern to improve the techniques of image processing for PCB's inspection. However, it is noticed the focus is mainly on the inspection of PCB in mass production. Even with a low volume of production, the small series production has become increasingly relevant. This work presents an image processing pipeline for mounted PCB inspection in small series production, strongly based on SIFT (Scale Invariant Feature Transform) algorithm. The defect types covered are: absent, rotated/inverted, misplaced and wrong. Experiments were performed using a software developed with the proposed approach. That software runs in an automatic optical inspection machine with an industrial camera and controlled illumination. The overall accuracy of the inspections is around 80%.