The research on surface defect detection method of circuit board based on convolutional neural network

To address the issues of low efficiency and strong subjectivity in manual inspection for surface defects in industrial products, this study proposes an intelligent detection method based on an improved YOLOv8 algorithm. By introducing adaptive image scaling and Mosaic data augmentation techniques to optimize the preprocessing pipeline, combined with an enhanced CSPDarknet53 backbone network, multi-scale feature fusion mechanism, and decoupled detection head design, the model's capability for detecting minute defects is significantly improved. Experiments on the NEU-DET steel defect dataset demonstrate that this method achieves an average detection accuracy ([email protected]) of 96.37%, with particularly outstanding performance in detecting prominent defects such as crazing (97.15%) and rolled-in scale (97.82%). Additionally, an interactive detection system developed using PyQt5 enables real-time detection and visual analysis, providing an efficient and reliable intelligent solution for industrial quality inspection.