Developing a surface mount technology defect detection system for mounted devices on printed circuit boards using a MobileNetV2 with Feature Pyramid Network.

This study proposes a method to develop an innovative robust fully automatic surface mount technology (SMT) defect detection system using MobileNetV2 with Feature Pyramid Network (FPN). The aim of the system is to detect mounted devices on Printed Circuit Boards (PCB) in real-time with good precision and relatively fast detection speed to improve quality control in the production industry. The design of the proposed system consists of data acquisition, data preprocessing, augmentation, labeling, and the detection model. Data acquisition presents the capturing the data, equipment involved and the setup, while data preprocessing explains the techniques employed to improve the quality of the dataset. To create robustness, the data was diversified and multiplied through the process of augmentation, followed by labeling to mark and tag regions of interest with their respective labels. The MobileNetV2 was utilized lastly, concatenated with FPN and a Single Shot MultiBox Detector (SSD). The proposed system displays a strong performance with a precision of 97.9%, recall of 96.3%, and F 1 score of 97.1%. The detection speed is relatively fast at 33.5FPS with an inference time of 30 ms per image. The proposed detection system demonstrates good performance at a competitive speed, and can detect mounted devices on PCBs in real-time with high confidence. [ABSTRACT FROM AUTHOR]