Automatic Detection and Classification of Steel Surface Defect Using Deep Convolutional Neural Networks

Automatic detection of steel surface defects is very important for product quality control in the industry. However, traditional method cannot be well applied production line, because its low accuracy and slow running speed. The current, popular algorithm (based on deep learning) also has problem accuracy, there still a lot room improvement. This paper proposes combining improved ResNet50 enhanced faster region convolutional neural networks (faster R-CNN) to reduce average time improve accuracy. Firstly, image input into model, which add deformable revolution network (DCN) cutout classify sample with without defects. If probability having defect less than 0.3, directly outputs Otherwise, samples are further R-CNN, adds spatial pyramid pooling (SPP), feature (FPN), matrix NMS. final output location classification or sample. By analyzing data set obtained real factory environment, this can reach 98.2%. At same time, other models.