基于Transformer的FPC缺陷检测模型研究

With the wide application of FPC in various electronic products, the demand for industrial inspection of FPC has gradually increased. In response to the low efficiency and high missed detection rate caused by the main reliance on manual visual inspection and electrical testing for FPC defect detection, as well as the difficulty of traditional computer vision models in extracting effective information from complex FPC defects, this paper designs an FPC defect detection model based on Transformer. FPC defects have the characteristics of complex and diverse defect and background conditions, high similarity between images, and large changes in defect image scale. Therefore, this paper integrates the Transformer model with the Faster R-CNN model, uses the multi-head non-local Transformer module (MNT) to enhance the ability to extract defect features, and designs the channel selection fusion feature pyramid module (CSF-PAFPN) to enhance the model\'s detection performance for multi-scale defect features. Compared with other mainstream models, the MAP improvement is more than 0.5%, and compared with the Faster R-CNN model, the MAP improvement is 1.77%. The effectiveness of each module is verified through ablation experiments. The experimental results show that the model can meet the actual production requirements in terms of accuracy, recall rate, and generalization ability.