|
二维码（扫一下试试看！） |
| 基于YOLOv5算法的轮胎表面损伤检测算法的改进研究 |
| Improvement Research on Tire Surface Damage Detection Algorithm Based on YOLOv5 Algorithm |
| Received:December 20, 2023 Revised:December 20, 2023 |
| DOI： |
| 中文关键词: 轮胎；损伤检测；YOLOv5算法；注意力机制；高效解耦头 |
| 英文关键词: tire；damage detection；YOLOv5 algorithm；attention mechanism；efficient decoupling head |
| 基金项目: |
| 中图分类号: |
| Author Name | Affiliation | E-mail | | CAO Jinfeng* | Qingdao University of Technology | caojinfeng@qut.edu.cn | | XUE Maolin | Qingdao University of Technology | | | CAO Yingjie | Qingdao University of Technology | | | ZHENG Jianfeng | Qingdao University of Technology | | | PENG Bo | Qingdao University of Technology | |
|
| Hits: 406 |
| Download times: 178 |
| 中文摘要: |
| 提出基于YOLOv5算法的一种轮胎表面损伤检测改进算法（简称DCS-YOLOv5算法）。该算法构建了包含多种轮胎损伤图像的数据集，使用高效解耦头分离分类和回归任务，采用CARAFE轻量级上采样算子生成包含细节信息的特征图，并引入SK注意力机制，增强了对多尺度轮胎损伤目标的特征提取和自适应能力。研究结果表明：DCS-YOLOv5算法的精确率、召回率和平均精度均值分别为84.7%，78.0%和85.9%，比YOLOv5n算法分别提高了7.2%，3.4%和6.5%；DCS-YOLOv5算法的检测速度为135帧·s-1，能够快速完成轮胎的损伤检测，可用于加油站、收费站等低速通行场所。 |
| 英文摘要: |
| An improved tire surface damage detection algorithm based on YOLOv5 algorithm （referred to as DCS-YOLOv5 algorithm） was proposed.It constructed a dataset containing multiple tire damage images，used efficient decoupling heads to separate classification and regression tasks，used CARAFE lightweight upsampling operator to generate feature maps containing detailed information，and introduced SK attention mechanism to enhance feature extraction and adaptive ability for multi-scale tire damage targets. The research results showed that the accuracy，recall，and average precision of the DCS-YOLOv5 algorithm were 84.7%，78.0%，and 85.9%，respectively，which were 7.2%，3.4%，and 6.5% higher than those of the YOLOv5n algorithm.The detection speed of DCS-YOLOv5 algorithm was 135 frames per second，which could quickly complete tire damage detection and could be applied to low-speed traffic places such as gas stations and toll stations. |
|
View Full Text
View/Add Comment Download reader |
| Close |
