Faster R-CNN assessment for air bubbles detection in the conformal coating application

Abstract

The detection of defects in Printed Circuit Boards (PCB) during the assembly process is an important quality requirement for the electronic manufacturing.The present paper emphasizes the assessment of computer vision implying Faster-RCNN object detection architecture, with the aim of air bubbles defects detection in PCB following the conformal coating process.Several image configurations have been used to increase artificially the training model’s performance, such as air bubbles size, location (on a flat PCB’s surface, between components leads) and illumination, etc.Toward reaching this cap, we used a random pattern of choice regarding the images with different configurations to properly evaluate its performance, especially the accuracy, precision and sensitivity.Our results showed that Faster-RCNN delivers the lowest detection performance for the air bubbles located between components leads compared to the ones located on flat PCB surfaces.This paper shows the accuracy, precision and sensitivity performance results by applying the Faster-RCNN object detection architecture for air bubbles defect detection in the conformal coating application, with aim of providing reference for future research in the field of air bubbles detection.