EMG-YOLO: road crack detection algorithm for edge computing devices

Frontiers in Neurorobotics, Год журнала: 2024, Номер 18

Опубликована: Июль 2, 2024

Introduction Road cracks significantly shorten the service life of roads. Manual detection methods are inefficient and costly. The YOLOv5 model has made some progress in road crack detection. However, issues arise when deployed on edge computing devices. main problem is that devices directly connected to sensors. This results collection noisy, poor-quality data. adds computational burden model, potentially impacting its accuracy. To address these issues, this paper proposes a novel algorithm named EMG-YOLO. Methods First, an Efficient Decoupled Header introduced optimize head structure. approach separates classification task from localization task. Each can then focus learning most relevant features. reduces model’s resources time. It also achieves faster convergence rates. Second, IOU loss function upgraded MPDIOU function. works by minimizing top-left bottom-right point distances between predicted bounding box actual labeled box. addresses complex computation high current model. Finally, GCC3 module replaces traditional convolution. performs global context modeling with input feature map obtain information. enhances capabilities Results Experimental show improved better performance all parameter indicators compared mainstream algorithms. EMG-YOLO improves accuracy 2.7%. mAP (0.5) (0.9) 2.9% 0.9%, respectively. new outperforms environments Discussion proposed effectively poor data quality achieved through optimizing structure, upgrading function, introducing modeling. demonstrate significant improvements both efficiency, especially environments. Future research further explore more lightweight efficient object models for

Язык: Английский

Survey of automated crack detection methods for asphalt and concrete structures

Innovative Infrastructure Solutions, Год журнала: 2024, Номер 9(11)

Опубликована: Окт. 27, 2024

Язык: Английский