En
期刊文章
文章精选
当期目录
过刊浏览
高被引
高下载
特邀文章
专刊合集
待刊文章
Volume 114
您当前的位置：首页 > 期刊文章 > 过刊浏览 > Volumes 108-119 (2025) > Volume 114
YOLOv8l-Voronoi ensemble learning for enhanced detection and segmentation of nanoparticle agglomerates in micro fluidized beds
Juhui Chen a b *, Buyang Peng a b, Dan Li a b, Yongxin Zhu a b, Xifeng Cao a b, Michael Zhuravkov c d, Siarhei Lapatsin c d, Wenrui Jiang d
a School of Mechanical and Power Engineering, Harbin University of Science and Technology, Harbin, 150080, China
b Key Laboratory of Advanced Manufacturing and Intelligent Technology, Ministry of Education, Harbin University of Science and Technology, Harbin, 150080, China
c Heilongjiang Key Laboratory for International Cooperation on Gear Transmission of Maritime and Air Equipment, Harbin, 150001, China
d School of Mechatronics Engineering, Harbin Institute of Technology, Harbin, 150001, China
10.1016/j.partic.2026.05.002
Volume 114, July 2026, Pages 394-407
Received 18 March 2026, Revised 23 April 2026, Accepted 6 May 2026, Available online 13 May 2026, Version of Record 19 May 2026.
E-mail: chenjuhui@hrbust.edu.cn
Highlights

• YOLOv8l-Voronoi ensemble for nanoparticle agglomerate detection and segmentation.

• An adaptive overlapping tiling strategy reduces the agglomerate miss rate from 19.4% to 13.1%.

• A Voronoi-SVM module suppresses false alarms, increasing the mAP@0.5 to 90.1%.

• A cascaded U-Net model refines boundaries, achieving an MIoU of 0.876.

• Quantitative analysis reveals ZnO content reduces agglomerate size and improves sphericity.


Abstract

This study proposes an integrated YOLOv8l-Voronoi ensemble framework for accurate detection and segmentation of nanoparticle agglomerates in micro-fluidized beds. To address challenges posed by high density, small size, and complex morphologies of agglomerates, an adaptive overlapping tiling strategy is introduced, reducing the agglomerate miss rate from 19.4% to 13.1% compared to the baseline YOLOv8l model. A Voronoi polygon partitioning combined with a support vector machine (SVM) is further incorporated to suppress false alarms caused by illumination variations, and a cascaded U-Net model is employed for pixel-level boundary segmentation. Experimental results show that the proposed framework achieves an mAP@0.5 of 90.1% on a self-built dataset, representing an 8.4 percentage point improvement over the baseline, while maintaining a high segmentation accuracy (MIoU>0.87). The method provides a robust and precise solution for online monitoring and quantitative analysis of nanoparticle aggregation behavior.

Graphical abstract
Keywords
YOLOv8l; Nanoparticle agglomerates; Voronoi diagram; Support Vector Machine (SVM); Micro fluidized bed
文章导读