Abstract: Aiming at the deficiencies of traditional methods in the defect detection of Printed Circuit boards (PCB), an intelligent detection scheme based on the improved YOLOv5 model is proposed.Firstly, the YOLOv5 model is adopted as the benchmark to conduct a quantitative comparison of the mainstream single-stage detectors; Secondly, the bidirectional Feature Pyramid Network (Bi-FPN) structure is adopted for model design improvement to enhance the detection accuracy of the model for minor defects and reduce the false detection rate.Finally, the PCB-1386 industrial dataset was constructed, the multimodal data augbling strategy was implemented, a three-stage progressive transfer learning scheme was designed to adapt to the small sample characteristics of industrial quality inspection, and a three-layer evaluation system was established.The experimental results show that the detection performance of the improved YOLOv5 model is significantly better than that of the traditional detection model.It has the advantages of high detection accuracy, fast reasoning speed and strong index stability, which can significantly reduce the defect detection cost of enterprises.It has good economic and practical value and huge market application potential.