Enhanced YOLOv8-based pavement crack detection: A high-precision approach

At present, the repair of cracks is still implemented manually, which has the problems of low identification efficiency and high labor cost. Crack detection is the key to realize the mechanical and intelligent crack repair. To solve these problems, an improved automatic recognition algorithm based o...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:PloS one Jg. 20; H. 5; S. e0324512
Hauptverfasser: Zhang, ZuXuan, Zhang, HongLi, Zhang, TongJia
Format: Journal Article
Sprache:Englisch
Veröffentlicht: United States Public Library of Science 22.05.2025
Public Library of Science (PLoS)
Schlagworte:
Accuracy
Algorithms
Asphalt pavements
Biology and Life Sciences
Computer and Information Sciences
Computer vision
Construction Materials
Cracking
Cracks
Datasets
Deep learning
Engineering and Technology
Neural networks
Pavements
Physical Sciences
Real time
Recognition
Repair
Research and Analysis Methods
Sewing machines
Social Sciences
China
ISSN:1932-6203, 1932-6203
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:At present, the repair of cracks is still implemented manually, which has the problems of low identification efficiency and high labor cost. Crack detection is the key to realize the mechanical and intelligent crack repair. To solve these problems, an improved automatic recognition algorithm based on YOLOv8 model, YOLOV8-DGS is proposed in this study. Firstly, this paper introduces deep separable Convolution (DWConv) into YOLOv8 backbone network to capture crack information more flexibly and improve the recognition accuracy of the model. Secondly, GSConv is used in the neck part to reduce computation and enhance feature representation, especially in the processing of multi-scale fracture features. Through these improvements, YOLOv8-DGS not only improves the accuracy of small cracks, but also ensures the real-time and high efficiency of intelligent joint filling equipment in practical applications. Experimental results show that the Precision, Recall, F1-score, mAP50 and FPS of the YOLOv8-DGS algorithm in pavement crack detection are 91.6%, 90%, 90.8%, 92.4% and 85 frames, respectively. At the same time, the recognition rate of different types of cracks in the model reached more than 86%, which increased by 20.5% compared with the YOLO11 model. This method can provide theoretical basis for automatic crack identification and technical support for automatic seam filling machine.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
Competing Interests: The authors have declared that no competing interests exist.
ISSN:1932-6203
1932-6203
DOI:10.1371/journal.pone.0324512