Journal of Computer Applications ›› 2024, Vol. 44 ›› Issue (5): 1613-1618.DOI: 10.11772/j.issn.1001-9081.2023050635

Special Issue: 多媒体计算与计算机仿真

• Multimedia computing and computer simulation • Previous Articles     Next Articles

Embedded road crack detection algorithm based on improved YOLOv8

Huantong GENG1,2, Zhenyu LIU1(), Jun JIANG1, Zichen FAN3, Jiaxing LI1   

  1. 1.School of Computer Science,Nanjing University of Information Science and Technology,Nanjing Jiangsu 210044,China
    2.School of Information Technology,Jiangsu Open University,Nanjing Jiangsu 210036,China
    3.School of Software,Nanjing University of Information Science and Technology,Nanjing Jiangsu 210044,China
  • Received:2023-05-22 Revised:2023-08-16 Accepted:2023-08-18 Online:2023-09-01 Published:2024-05-10
  • Contact: Zhenyu LIU
  • About author:GENG Huantong, born in 1973, Ph. D., professor. His research interests include multi-objective optimization, deep learning.
    JIANG Jun, born in 1998, M. S. candidate. His research interests include transfer learning, target detection.
    FAN Zichen, born in 2001, M. S. candidate. His research interests include artificial intelligence, target detection.
    LI Jiaxing, born in 2000, M. S. candidate. His research interests include model compression, semantic segmentation.
  • Supported by:
    National Natural Science Foundation of China(51977100)

基于改进YOLOv8的嵌入式道路裂缝检测算法

耿焕同1,2, 刘振宇1(), 蒋骏1, 范子辰3, 李嘉兴1   

  1. 1.南京信息工程大学 计算机学院, 南京 210044
    2.江苏开放大学 信息工程学院, 南京 210036
    3.南京信息工程大学 软件学院, 南京 210044
  • 通讯作者: 刘振宇
  • 作者简介:耿焕同(1973—),男,安徽宣城人,教授,博士生导师,博士,CCF高级会员,主要研究方向:多目标优化、深度学习
    蒋骏(1998—),男,安徽马鞍山人,硕士研究生,主要研究方向:迁移学习、目标检测
    范子辰(2001—),男,江苏宿迁人,硕士研究生,主要研究方向:人工智能、目标检测
    李嘉兴(2000—),男,山西长治人,硕士研究生,主要研究方向:模型压缩、语义分割。
    第一联系人:刘振宇(1999—),男,安徽淮北人,硕士研究生,主要研究方向:人工智能、目标检测
  • 基金资助:
    国家自然科学基金资助项目(51977100)

Abstract:

Deploying the YOLOv8L model on edge devices for road crack detection can achieve high accuracy, but it is difficult to guarantee real-time detection. To solve this problem, a target detection algorithm based on the improved YOLOv8 model that can be deployed on the edge computing device Jetson AGX Xavier was proposed. First, the Faster Block structure was designed using partial convolution to replace the Bottleneck structure in the YOLOv8 C2f module, and the improved C2f module was recorded as C2f-Faster; second, an SE (Squeeze-and-Excitation) channel attention layer was connected after each C2f-Faster module in the YOLOv8 backbone network to further improve the detection accuracy. Experimental results on the open source road damage dataset RDD20 (Road Damage Detection 20) show that the average F1 score of the proposed method is 0.573, the number of detection Frames Per Second (FPS) is 47, and the model size is 55.5 MB. Compared with the SOTA (State-Of-The-Art) model of GRDDC2020 (Global Road Damage Detection Challenge 2020), the F1 score is increased by 0.8 percentage points, the FPS is increased by 291.7%, and the model size is reduced by 41.8%, which realizes the real-time and accurate detection of road cracks on edge devices.

Key words: YOLOv8 (You Only Look Once version 8), object detection, lightweight, attention mechanism, road crack

摘要:

在边缘端设备部署YOLOv8L模型进行道路裂缝检测可以实现较高的精度,但难以保证实时检测。针对此问题,提出一种可部署到边缘计算设备Jetson AGX Xavier上的基于改进YOLOv8模型的目标检测算法。首先,利用部分卷积设计Faster Block结构以替换YOLOv8 C2f模块中的Bottleneck结构,并将改进后的C2f模块记为C2f-Faster;其次,在YOLOv8主干网络中的每个C2f-Faster模块之后接一个SE(Squeeze-and-Excitation)通道注意力层,进一步提高检测的精度。在开源道路损害数据集RDD20(Road Damage Detection 20)上的实验结果表明:所提方法的平均F1得分为0.573,每秒检测帧数(FPS)为47,模型大小为55.5 MB,相较于GRDDC2020 (Global Road Damage Detection Challenge 2020)的SOTA(State-Of-The-Art)模型,F1得分提高了0.8个百分点,FPS提高了291.7%,模型大小减小了41.8%,实现了在边缘设备上对道路裂缝实时且准确的检测。

关键词: YOLOv8, 目标检测, 轻量化, 注意力机制, 道路裂缝

CLC Number: