主办：陕西省汽车工程学会
ISSN 1671-7988  CN 61-1394/TH
创刊：1976年
汽车实用技术

汽车实用技术 ›› 2022, Vol. 47 ›› Issue (17): 25-30.DOI: 10.16638/j.cnki.1671-7988.2022.017.005

• 智能网联汽车 • 上一篇    

基于 YOLOv5 网络的轮胎面缺陷检测分析

王鹏辉，王旭飞*，刘怡帆，周 鹏，惠继强   

  1. 陕西理工大学 机械工程学院
  • 出版日期:2022-09-15 发布日期:2022-09-15
  • 通讯作者: 王旭飞
  • 作者简介:王鹏辉（1998—），男，硕士研究生，研究方向为自动驾驶，E-mail:1594368403@qq.com。 通讯作者：王旭飞（1975—），男，副教授，研究方向为车辆动力学及控制，自动驾驶，E-mail: wxf@snut.edu.cn。
  • 基金资助:
    陕西省重点实验室项目（18JS020）。

Analysis on Tire Tread Defect Detection Based on YOLOv5 Network

WANG Penghui, WANG Xufei*, LIU Yifan, ZHOU Peng, HUI Jiqiang   

  1. School of Mechanical Engineering, Shaanxi University of Science and Technology
  • Online:2022-09-15 Published:2022-09-15
  • Contact: WANG Xufei

PDF

摘要： 当前时有发生因轮胎面缺陷导致汽车在行驶中发生严重的交通事故，轮胎面缺陷智能 检测对避免这类交通事故的发生具有重要意义。深度学习技术被越来越多地用于目标检测领 域，文章基于卷积神经网络模型 YOLOv5 提出一种轮胎面缺陷智能检测方法。首先建立具有 4 种轮胎面缺陷特征的数据集，然后通过 YOLOv5 网络训练数据集，最后用训练好的网络模 型在测试集上检测。实验结果显示，在检测轮胎面缺陷任务中，YOLOv5 网络模型的平均检 测精度（mAP）达到 65.4%，检测速度可达到 38FPS，相较于 YOLOv4 网络模型与 Faster-RCNN 网络模型分别提高约 4.1%与 31.6%。对进一步研究更有效的轮胎面缺陷智能检测方法提供了 参考。

关键词: 轮胎面缺陷；YOLOv5；深度学习；缺陷检测

Abstract: At present, serious traffic accidents occur due to tire tread defects. Intelligent detection of tire tread defect is of great significance to avoid such traffic accidents. Deep learning technology has been used more and more in the field of target detection. This paper proposes an intelligent detection method of tire tread defect based on the convolutional neural network model YOLOv5. Firstly, the data set with four tire tread defect characteristics was established.Secondly,the data set was trained by YOLOv5 network. Finally, the trained network model was used to detect the tire tread defect on the test set. The experimental results show that the YOLOv5 network model has a mean average precision (mAP) of 65.4% and a detection speed of 38FPS in tire tread defect detection task, which is about 4.1% and 31.6% higher than YOLOv4 network model and Faster-RCNN network model, respectively. It provides a reference for further research on more effective intelligent detection methods for tire tread defect.

Key words: Tire tread defect; YOLOv5; Deep learning; Defect detection