基于RetinaNet的PCB焊点数字射线缺陷图像检测

Automatic Detection of Weld Defects in X-Ray Image Based on RetinaNet

为了避免焊接缺陷引起的故障和质量问题,PCB焊点检测已经成为电子产品生产制造中的重要环节.使用基于深度学习的数字射线无损检测来检查PCB电路板内部焊点缺陷,可以在提高生产效率的同时减轻工人的劳动压力.本文建立了3种常见的数字射线下PCB焊点缺陷图像数据集,基于RetinaNet搭建了自动检测网络模型.经过训练测试,该模型对于缺陷图片的平均检测准确率达到了92.7%,能够有效地提高X射线下PCB焊点缺陷检测的性能和效率.

To prevent failures and quality issues caused by welding defects,PCB solder joint inspection has become a critical step in the manufacturing of electronic products.The application of deep learning-based digital X-ray non-destructive testing to examine internal solder joint defects within PCB circuit boards can increase production efficiency while reducing the labor pressure on workers.Three common digital X-ray datasets of PCB solder joint defects were established and an automated detection network model based on RetinaNet was constructed.After training and testing,the model achieved an average detection accuracy of 92.7%for defect images.Experimental results demonstrate that the proposed model effectively enhanced the performance and efficiency of PCB solder joint defect detection under X-ray inspection.