Asian Journal of Mathematics and Computer Research

With the rapid advancement of the electronics manufacturing industry, the quality requirements for circuit board production continue to increase. To effectively reduce product defect rates, PCB appearance defect detection has become critical in industrial production. Initially, PCB defect inspection relied on manual visual checks. However, with the increasing integration of electronic devices, this manual approach can no longer meet the practical demands of industrial production. To address these challenges, this study proposes a computer vision-based algorithm for detecting product appearance defects and develops a user-friendly PCB defect detection system using Python and PyQt5. First, we examined the characteristics and detection requirements of PCB board appearance defect detection. We analyzed the features of defects that may arise during PCB board production, collected existing datasets, and analyzed the current challenges in PCB board appearance defect detection. We designed the overall structural framework of the system solution and the detection workflow for the defect detection algorithm. Second, we selected the YOLOv8 deep learning model, which excels in the field of object detection due to its efficient detection speed and high detection accuracy. We trained the YOLOv8 model using the dataset, adjusting model parameters and optimizing algorithms to achieve optimal detection performance. Following training, we evaluated the model's performance metrics, including detection accuracy and recall rate. Based on the evaluation results, we implement necessary optimizations to enhance the model's generalization capability and detection performance. Finally, we integrate the trained YOLOv8 model into the PCB board defect detection system and conduct comprehensive testing to ensure the system's stability and reliability in actual production environments. The improved YOLOv8_SE achieved 89.1% mAP and 52.3 FPS, outperforming prior models.