EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

FSNB-YOLOV8: Improvement of Object Detection Model for Surface Defects Inspection in Online Industrial Systems

Jun Li, Jinglei Wu, Yanhua Shao

Applied Sciences · 05.09.2024 · 14 (17) · 7913

DOI: 10.3390/app14177913 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Applied Sciences, online erschienen am 05.09.2024. Verfasst von Jun Li und 2 weiteren Autor:innen. Die Arbeit wurde bisher 14-mal zitiert (Crossref) und nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Basic Research Project of the Translational Application Project of the “Wise Eyes Action”.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The current object detection algorithm based on CNN makes it difficult to effectively capture the characteristics of subtle defects in online industrial product packaging bags. These defects are often visually similar to the texture or background of normal product packaging bags, and the model cannot effectively distinguish them. In order to deal with these challenges, this paper optimizes and improves the network structure based on YOLOv8 to achieve accurate identification of defects. First, in order to solve the long-tail distribution problem of data, a fuzzy search data enhancement algorithm is introduced to effectively increase the number of samples. Secondly, a joint network of FasterNet and SPD-Conv is proposed to replace the original backbone network of YOLOv8, which effectively reduces the computing load and improves the accuracy of defect identification. In addition, in order to further improve the performance of multiscale feature fusion, a weighted bidirectional feature pyramid network (BiFPN) is introduced, which effectively enhances the model’s ability to detect defects at different scales through the fusion of deep information and shallow information. Finally, in order to reduce the sensitivity of the defect position deviation, the NWD loss function is used to optimize the positioning performance of the model better and reduce detection errors caused by position errors. Experimental results show that the FSNB_YOLOv8 model proposed in this paper can reach 98.8% mAP50 accuracy. This success not only verifies the effectiveness of the optimization and improvement of this article’s model but also provides an efficient and accurate solution for surface defect detection of industrial product packaging bags on artificial assembly systems.

Förderung

  • Basic Research Project of the Translational Application Project of the “Wise Eyes Action”

Zitieren

APA

Li, J., Wu, J., & Shao, Y. (2024). FSNB-YOLOV8: Improvement of Object Detection Model for Surface Defects Inspection in Online Industrial Systems. Applied Sciences, 14(17), 7913. https://doi.org/10.3390/app14177913

Vancouver

Li J, Wu J, Shao Y. FSNB-YOLOV8: Improvement of Object Detection Model for Surface Defects Inspection in Online Industrial Systems. Applied Sciences. 2024;14(17):7913. doi:10.3390/app14177913

BibTeX
@article{Li2024,
  author = {Li, Jun and Wu, Jinglei and Shao, Yanhua},
  title = {FSNB-YOLOV8: Improvement of Object Detection Model for Surface Defects Inspection in Online Industrial Systems},
  journal = {Applied Sciences},
  year = {2024},
  volume = {14},
  number = {17},
  pages = {7913},
  doi = {10.3390/app14177913}
}
Kostenlos starten →