EUVIMEDEUVIMED entdecken
Crossref

MSFF-YOLOv11: a multi-scale feature fusion enhanced YOLOv11 for ampoule head defect detection

Xin Chen, Zhizhong Liu, Ximing Yang, Zhendong Cui, Quan Z Sheng

Measurement Science and Technology · 16.10.2026 · 37 (41) · 415402

DOI: 10.1088/1361-6501/aea6fc ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 13.10.2026. Verfasst von Xin Chen und 4 weiteren Autor:innen. Die Arbeit nennt 37 Literaturangaben. Gefördert von: Key Research and Development Project of Shandong Province, Australian Research Council (ARC) Future Fellowship, Discovery Project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Ampoule head defect detection is essential in pharmaceutical quality control but remains challenging because of complex structures and diverse defect morphologies. To address the limitations of existing methods in feature extraction and feature loss, we propose a YOLOv11-based multi-scale feature fusion model for ampoule head defect detection, MSFF-YOLOv11. In the feature extraction stage, we introduce a Poly Kernel Inception block that leverages parallel multi-scale convolutional kernels to enhance the representation of defects with varying scales and shapes. In the feature fusion stage, we design a multi-scale feature fusion module and adjust the network structure to effectively mitigate feature loss caused by downsampling operations. Furthermore, in the detection head, the bounding box regression loss is replaced with Powerful-IoU version 2, thus enhancing the localization precision and robustness for multi-scale defects in the ampoule head. Experimental results on a self-built dataset show that MSFF-YOLOv11 achieves a precision of 98.1% (+6.3%), a recall of 94.1% (+1.0%), and an mAP50 of 97.0% (+2.9%) compared with the baseline model, YOLOv11n. In addition, MSFF-YOLOv11 outperforms the second-best model, YOLOv13n, by 0.7% in mAP50. These results demonstrate the effectiveness of the proposed model and its strong potential for ampoule head defect detection.

Förderung

  • Key Research and Development Project of Shandong Province
  • Australian Research Council (ARC) Future Fellowship
  • Discovery Project
  • National Natural Science Foundation of China

Zitieren

APA

Chen, X., Liu, Z., Yang, X., Cui, Z., & Sheng, Q. Z. (2026). MSFF-YOLOv11: a multi-scale feature fusion enhanced YOLOv11 for ampoule head defect detection. Measurement Science and Technology, 37(41), 415402. https://doi.org/10.1088/1361-6501/aea6fc

Vancouver

Chen X, Liu Z, Yang X, Cui Z, Sheng QZ. MSFF-YOLOv11: a multi-scale feature fusion enhanced YOLOv11 for ampoule head defect detection. Measurement Science and Technology. 2026;37(41):415402. doi:10.1088/1361-6501/aea6fc

BibTeX
@article{Chen2026,
  author = {Chen, Xin and Liu, Zhizhong and Yang, Ximing and Cui, Zhendong and Sheng, Quan Z},
  title = {MSFF-YOLOv11: a multi-scale feature fusion enhanced YOLOv11 for ampoule head defect detection},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {41},
  pages = {415402},
  doi = {10.1088/1361-6501/aea6fc}
}
Kostenlos starten →