A deep-learning-based approach for fast and robust steel surface defects classification. (October 2019)

• Record Type: Journal Article
• Authors: Fu, Guizhong; Sun, Peize; Zhu, Wenbin; Yang, Jiangxin; Cao, Yanlong; Yang, Michael Ying; Cao, Yanpeng
• Is Part Of: Optics and lasers in engineering. Volume 121(2019)
• Page Start: 397
• Page End: 405
• Publication Date: 2019-10
• Subjects: Surface inspection -- Defect classification -- Convolutional neural network -- Feature extraction -- Multi-receptive field; Lasers in engineering -- Periodicals; Optical measurements -- Periodicals; Optics -- Periodicals; Lasers en ingénierie -- Périodiques; Mesures optiques -- Périodiques; Optique -- Périodiques; 621.36605
• DOI: 10.1016/j.optlaseng.2019.05.005
• Languages: English
• ISSNs: 0143-8166

Abstract

Highlights: A novel end-to-end SqueezeNet-based model is proposed to achieve accurate recognition of steel surface defects. Two effective techniques are presented to improve defect recognition accuracy of our proposed CNN model. A diversity-enhanced testing dataset of steel surface defects is constructed to evaluate the robustness of classification models. Our method runs in real-time and achieves significantly higher classification accuracy compared with the state-of-the-art defect classifiers. Abstract: Automatic visual recognition of steel surface defects provides critical functionality to facilitate quality control of steel strip production. In this paper, we present a compact yet effective convolutional neural network (CNN) model, which emphasizes the training of low-level features and incorporates multiple receptive fields, to achieve fast and accurate steel surface defect classification. Our proposed method adopts the pre-trained SqueezeNet as the backbone architecture. It only requires a small amount of defect-specific training samples to achieve high-accuracy recognition on a diversity-enhanced testing dataset of steel surface defects which contains severe non-uniform illumination, camera noise, and motion blur. Moreover, our proposed light-weight CNN model can meet the requirement of real-time online inspection, running over 100 fps on a computer equipped with a single NVIDIA TITAN X Graphics Processing Unit (12G memory). Codes and a diversity-enhanced testing dataset will be made publicly available.