Steel plates, as the cornerstone of modern manufacturing, play a decisive role in the performance and reliability of products. Therefore, the detection of steel plate surface defects is of great research significance and application value. The main challenges faced by research on steel plate surface defect detection include the diversity of defect types, the indistinct features, and the high requirement for real-time detection. By using deep learning technology to study steel plate surface defect detection, this research systematically summarizes the commonly used datasets and the application of target detection algorithms in surface defects. It also comprehensively organizes and analyzes the advantages and disadvantages of these algorithms. The application of supervised learning, unsupervised learning, and semi-supervised learning methods in steel plate surface defect detection is discussed, key indicators for evaluating the current algorithm performance are evaluated, challenges facing future steel plate surface defect detection are discussed, and future research trends are prospected.