Defect Sample Generation System Based on DCGAN for Glass Package Electrical Connectors

摘要

Precision glass packaged electrical connectors are widely used to connect precision modular electrical appliances. Manual detection is currently used due to its small size. In order to improve the detection efficiency and accuracy, deep learning is applied to the automatic detection of the defect. In this paper, in order to solve the problem of lack of defect samples, a defect sample generation system for micro-precision glass package electrical connectors based on deep convolutional generation confrontation network (DCGAN) is constructed. A network model of defect samples based on DCGAN is constructed, and algorithm design, parameter settings, network training and experiments are performed. More than 1,500 samples were generated. Finally, the generated samples were tested for defects (including missing blocks and bubbles). The results show that the samples generated by this method can be used as real samples in defect detection training.