Wyniki wyszukiwania - BazTech

Ograniczanie wyników

1 Metrology and Measurement Systems

1 2025

Powiadomienia systemowe

Sesja wygasła!

Znaleziono wyników: 1

Liczba wyników na stronie

Wyniki wyszukiwania

Sortuj według:

Ogranicz wyniki do:

Non-destructive detection of ceramic ball surface defects based on improved YOLOv8

Gu Guo-Qing, Wang Han-Lei, Miao Xu-Yi, Yan Jing-Yi, Du Jian-Zhou

Metrology and Measurement Systems

2025

Vol. 32, nr 1

1--18

To address issues such as severe specular reflection, low detection accuracy, and large model parameters in ceramic ball detection, an improved YOLOv8 model, named YOLOv8-AP, is proposed for ceramic ball surface defects detection. Firstly, the coaxial light source is employed to reduce the specular reflection effect and an efficient image acquisition platform is established to obtain defect samples. Additionally, various data augmentation techniques are utilized to expand the dataset, and both the ADown module and an improved Powerful-IoU loss function are introduced to optimize the YOLOv8 network, significantly enhancing the detection efficiency for small target defects. Experimental results show that the proposed improved YOLOv8-AP model can achieve a mean average precision of 96.1% for the detection of the ceramic ball surface defects, which greatly enhances the defect detection accuracy compared to the traditional models and can hope to meet the intelligent and automatic detection requirements of ceramic ball detection online applications.