A Rapid Concrete Crack Detection Method Based on Improved YOLOv8

An improved YOLOv8 (You Only Look Once version 8) model is proposed to tackle the challenges of low detection accuracy and slow speed resulting from the complex background and shape diversity of concrete cracks. Firstly, a lightweight feature fusion module, GE_Conv, is designed by integrating the Gh...

Full description

Saved in:
Bibliographic Details
Main Authors: Yongzhen Wang, Jiacong He
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10945319/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850214258118230016
author Yongzhen Wang
Jiacong He
author_facet Yongzhen Wang
Jiacong He
author_sort Yongzhen Wang
collection DOAJ
description An improved YOLOv8 (You Only Look Once version 8) model is proposed to tackle the challenges of low detection accuracy and slow speed resulting from the complex background and shape diversity of concrete cracks. Firstly, a lightweight feature fusion module, GE_Conv, is designed by integrating the Ghost Module and Efficient Channel Attention in series. This module is embedded into the neck network to preserve more feature information during downsampling and accelerate the model’s inference speed. Secondly, the DBB_Bottleneck is introduced into the C2f module, combining the lightweight GE_Conv with the structurally re-parameterized Diverse Branch Block, enhancing the model’s multi-scale feature extraction capability. Furthermore, the introduction of the GF_Detect detection head significantly reduces the number of model parameters while improving detection performance. Finally, the WIoUv3 loss function is employed to dynamically assign anchor boxes of varying qualities, thereby enhancing the accuracy of anchor box positioning. Experimental results demonstrate that the proposed model achieves a detection precision of 92.9% and a mAP@50 (mean average precision at an Intersection over Union threshold of 0.5) of 77.8%. Compared to state-of-the-art algorithms such as Faster R-CNN, SSD, RetinaNet, YOLOv5s, YOLOv7-tiny, and the original YOLOv8s, the proposed model exhibits superior performance in both detection accuracy and generalization capability. Additionally, the model achieves an average detection time of 10.20 ms per image, demonstrating its practical feasibility. This study not only improves the accuracy and speed of crack detection but also significantly reduces the computational complexity of the model, advancing the development of lightweight and practical crack detection algorithms. The proposed model can be widely applied to crack inspection tasks for roads, bridges, coal mine shafts, and other scenarios, providing efficient and reliable technical support for crack maintenance and management.
format Article
id doaj-art-224460076f034baebf39de760ce03ff4
institution OA Journals
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-224460076f034baebf39de760ce03ff42025-08-20T02:08:57ZengIEEEIEEE Access2169-35362025-01-0113592275924310.1109/ACCESS.2025.355582510945319A Rapid Concrete Crack Detection Method Based on Improved YOLOv8Yongzhen Wang0https://orcid.org/0009-0008-0215-8473Jiacong He1School of Mechanical and Electrical Engineering, China University of Mining and Technology, Xuzhou, ChinaSchool of Mechanical and Electrical Engineering, China University of Mining and Technology, Xuzhou, ChinaAn improved YOLOv8 (You Only Look Once version 8) model is proposed to tackle the challenges of low detection accuracy and slow speed resulting from the complex background and shape diversity of concrete cracks. Firstly, a lightweight feature fusion module, GE_Conv, is designed by integrating the Ghost Module and Efficient Channel Attention in series. This module is embedded into the neck network to preserve more feature information during downsampling and accelerate the model’s inference speed. Secondly, the DBB_Bottleneck is introduced into the C2f module, combining the lightweight GE_Conv with the structurally re-parameterized Diverse Branch Block, enhancing the model’s multi-scale feature extraction capability. Furthermore, the introduction of the GF_Detect detection head significantly reduces the number of model parameters while improving detection performance. Finally, the WIoUv3 loss function is employed to dynamically assign anchor boxes of varying qualities, thereby enhancing the accuracy of anchor box positioning. Experimental results demonstrate that the proposed model achieves a detection precision of 92.9% and a mAP@50 (mean average precision at an Intersection over Union threshold of 0.5) of 77.8%. Compared to state-of-the-art algorithms such as Faster R-CNN, SSD, RetinaNet, YOLOv5s, YOLOv7-tiny, and the original YOLOv8s, the proposed model exhibits superior performance in both detection accuracy and generalization capability. Additionally, the model achieves an average detection time of 10.20 ms per image, demonstrating its practical feasibility. This study not only improves the accuracy and speed of crack detection but also significantly reduces the computational complexity of the model, advancing the development of lightweight and practical crack detection algorithms. The proposed model can be widely applied to crack inspection tasks for roads, bridges, coal mine shafts, and other scenarios, providing efficient and reliable technical support for crack maintenance and management.https://ieeexplore.ieee.org/document/10945319/Concrete cracksobject detectionlightweight modulewise intersection over unionyou only look once version 8
spellingShingle Yongzhen Wang
Jiacong He
A Rapid Concrete Crack Detection Method Based on Improved YOLOv8
IEEE Access
Concrete cracks
object detection
lightweight module
wise intersection over union
you only look once version 8
title A Rapid Concrete Crack Detection Method Based on Improved YOLOv8
title_full A Rapid Concrete Crack Detection Method Based on Improved YOLOv8
title_fullStr A Rapid Concrete Crack Detection Method Based on Improved YOLOv8
title_full_unstemmed A Rapid Concrete Crack Detection Method Based on Improved YOLOv8
title_short A Rapid Concrete Crack Detection Method Based on Improved YOLOv8
title_sort rapid concrete crack detection method based on improved yolov8
topic Concrete cracks
object detection
lightweight module
wise intersection over union
you only look once version 8
url https://ieeexplore.ieee.org/document/10945319/
work_keys_str_mv AT yongzhenwang arapidconcretecrackdetectionmethodbasedonimprovedyolov8
AT jiaconghe arapidconcretecrackdetectionmethodbasedonimprovedyolov8
AT yongzhenwang rapidconcretecrackdetectionmethodbasedonimprovedyolov8
AT jiaconghe rapidconcretecrackdetectionmethodbasedonimprovedyolov8