Publication: Study on detection and classification of printed circuit board (PCB) defects by using deep learning
Loading...
Date
2024-07
Authors
Mohamad Eiman bin Abd Hamid
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The Printed Circuit Board (PCB) is a component that is extensively utilised in electronic industry products and is very significant to our daily lives. Due to technological constraints, defective PCBs will always occur during the manufacturing process. The detection and classification of defects in PCBs are crucial for ensuring the reliability and functionality of electronic devices. Traditional methods of PCB inspection such as human visual inspection, Continuity multimeter testing and rule-based algorithms are labour-intensive, subjective and may not be effective in detecting subtle or complex defects. Therefore, there is a growing interest in the application of advanced techniques such as deep learning and image processing that are able to automate the detection and classification of PCB defects. This proposal presents an automatic visual inspection technique for detecting and classifying defects in PCBs. The proposed technique utilises You Only Look Once version 5 (YOLOv5). The proposed automatic visual inspection technique provides a promising solution for detecting and classifying defects in PCBs. The proposed technique can be further improved and optimised to meet the specific needs of different industries and applications. These studies demonstrate the potential of deep learning in addressing the limitations of traditional methods and developing efficient systems for detecting and classifying various types of PCB defects. This project was created using YOLOv5 and was used to detect and analyse pictures related to PCB defects. Finally, able for future opportunities to detect defects of PCB with better efficientand easy implementation. The introduced automatic visual inspection technique aims not only to enhance the efficiency of PCB production processes but also to reduce costs and the time required for quality testing. With the integration of this technology, it is expected that the electronic industry can accelerate product innovation, enhance global competitiveness and provide more reliable solutions for an increasingly demanding market. The results of this study demonstrated the effectiveness of the YOLOv5 model in accurately detecting and classifying PCB defects. The model was trained on a dataset containing 4995 annotated images, covering six types of defects: missing hole, mouse bite, open circuit, short, spur, and spurious copper. The model's performance was evaluated using metrics such as mean Average Precision (mAP), precision, recall, and F1 score. The YOLOv5 model achieved high detection accuracy, with mAP values exceeding 0.8 for most defect categories. The results indicate that the model can reliably identify and classify PCB defects and make it a viable solution for automated PCB inspection.