Systematic Review of Steel Surface Defect Detection Methods on the Open Access Datasets of Severstal and the Northeastern University (NEU)

Steel is the essential component in various arias such as construction and infrastructure, mechanical and automotive items, metal goods, transportation, electrical devices, home appliances, etc. The surface quality of the steel affects the quality of the work done in these areas. In this study, the articles in the field of steel surface defect detection are reviewed. Since reproducible science is very necessary goal, only the articles that use the open access datasets are included. These open datasets are the Northeastern University (NEU) surface defect dataset and SEVERSTAL: Steel surface defect dataset. Within the scope of the study, traditional image processing methods and deep learning methods used in 100 articles published between 2013–2022 years were categorized. In addition, the software packages, frameworks, and comparison metrics used are summarized. The published articles are summarized using 11 figures and 14 tables, giving information about the properties of datasets, the methods for data augmentation and generation, image preprocessing, feature extraction, segmentation, detection, and classification, the performance metrics used to evaluate these methods, and the software packages and the frameworks used to develop these methods.