Machine Vision Based Paper Defect Detection

Aiming at improving the accuracy and speed of paper disease detection in the production process of modern high-speed paper machines, an automatic paper defect detection scheme based on the frame difference method and MobileNetV3-SSD model is proposed, which can efficiently realize the defects of paper for detection. In this study, we utilize the frame difference method to quickly identify the dynamic changes on the paper surface, and combine it with the lightweight deep learning architecture MobileNetV3-SSD to achieve efficient detection of paper defects. By introducing a feature pyramid network (FPN) and an attention mechanism (SE module), the model is able to adapt to different types and sizes of defects, significantly improving the detection accuracy. Experimental results show that compared with the original SSD algorithm, our method improves the average precision (AP) by 5.11% to 83.5%, and at the same time reduces the average processing time from 56.3 ms/frame to 34.8 ms/frame, which is of significant value for industrial applications.