An efficient and real-time steel surface defect detection method based on single-stage detection algorithm
Hongkai Zhang,Qiqi Miao,Suqiang Li,Chunping Wang,Sixian Chan,Jie Hu,Cong Bai
DOI: https://doi.org/10.1007/s11042-024-19565-2
IF: 2.577
2024-06-16
Multimedia Tools and Applications
Abstract:Within the realm of computer vision, the detection of defects holds significant importance for industrial production and quality assurance. However, current object detection methods encounter notable challenges when tasked with defect detection, including high inter-class similarity and substantial intra-class variability. These challenges restrict the performance of conventional approaches, leading to imprecise defect detection and localization. In response to these limitations, this study proposes a novel real-time, end-to-end defect detection approach based on a single-stage detection algorithm. Our method introduces several innovations to address these challenges and significantly enhance defect detection capabilities. Firstly, we develop a pioneering cross-layer feature aggregation method, enabling the network to effectively capture intricate defect patterns and improve its adaptability to defect detection tasks. Additionally, we introduce an attention-enhanced decoupled head, which prioritizes critical features while optimizing computational efficiency. Through comprehensive testing on the NEU-DET defect detection dataset, our approach demonstrates remarkable performance gains over conventional object detection methods. Notably, our method achieves an impressive detection speed of 68 FPS while surpassing existing methods in terms of detection accuracy. These results highlight the practical significance of our approach, meeting the stringent requirements of industrial production for both accuracy and speed. Furthermore, the efficacy and robustness of our method are further validated using the GC10-DET dataset, underscoring its potential for real-world applications across diverse industrial settings.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering