Surface defect detection of steel strips based on improved YOLOv4. (September 2022)

• Record Type: Journal Article
• Authors: Li, Mengjiao; Wang, Hao; Wan, Zhibo
• Is Part Of: Computers & electrical engineering. Volume 102(2022)
• Publication Date: 2022-09
• Subjects: Strip surface defect detection -- YOLOv4 -- Convolutional block attention module -- Spatial pyramid pooling -- Receptive field block; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2022.108208
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: The modified YOLOv4 algorithm for defect detection of industrial strip steel. Add a Convolutional Block Attention Module to the YOLOv4's backbone network. Receptive Field Block replaced the path aggregation network. The outperforms most of the mainstream target detection networks. Abstract: During the production and processing of steel strips, the production process and external factors lead to surface defects that negatively impact the strips' integrity and functionality. However, traditional manual defect detection algorithms cannot meet modern accuracy requirements. Therefore, we propose a steel strip surface defect detection method based on the improved you-only-look-once version 4 (YOLOv4) algorithm. The attention mechanism is embedded in the backbone network structure, and the path aggregation network is modified into a customised receptive field block structure, which strengthens the feature extraction functionality of the network model. From the final experimental results, relative to the original YOLOv4 algorithm, the proposed algorithm's mean average precision values in the detection of four types of steel strip defects is improved by 3.87%, reaching 85.41%, thereby providing a new detection method for daily steel strip surface defects. Graphical abstract: Image, graphical abstract