一种BiFPN-YOLOv8铁轨表面缺陷检测网络模型
A BiFPN-YOLOv8 for Rail Surface Defect Detection Network Model
冯庆贺 1江铭凯 1郝巧红 1赵晓蕾 2赵强 1杨富超1
作者信息
- 1. 河南工学院 智能工程学院,河南 新乡 453003
- 2. 中国电波传播研究所,山东 青岛 266107
- 折叠
摘要
针对铁轨表面存在的剥落、孔洞、瘢痕、划痕、裂缝、磨损缺陷,构建了一种BiFPN-YOLOv8 铁轨表面缺陷检测网络模型.在数据准备中,通过Gauss噪声变换和水平变换进行铁轨表面图像数据增强;在网络构建中,通过在原YOLOv8 网络模型中引入加权双向金字塔结构BiFPN构建BiFPN-YOLOv8 网络模型用于提升铁轨表面缺陷检测性能;在实验仿真中,通过在RSDDs数据集上定量的对比和定性的检测结果评价,验证了BiFPN-YOLOv8 网络模型在铁轨表面缺陷检测任务上的精准性和适用性.
Abstract
Aiming to defect such as peeling,holes,scars,scratches,cracks and wear on the rail surface,a BiFPN-YOLOv8 rail surface defect detection network model is constructed.In the data preparation phase,the rail surface image data is augmented through Gauss noise transformation and rotation transformation;In the network construction phase,in order to improve the performance of rail surface defect detection,the BiFPN-YOLOv8 is constructed by introducing the weighted bidirectional pyramid structure BiFPN into the original YOLOv8 network model;In the experimental simulation phase,the quantitative comparison and qualitative detection result evaluation on the RSDDs data set illustrates the accuracy and applicability of the BiFPN-YOLOv8 network model in the rail surface defect detection task.
关键词
铁轨表面/缺陷检测/BiFPN-YOLOv8Key words
rail surface/defect detection/BiFPN-YOLOv8引用本文复制引用
基金项目
河南省重点研发与推广专项(科技攻关)项目(232102210094)
河南省重点研发与推广专项(科技攻关)项目(222102210291)
河南省国际科技合作项目(242102520036)
出版年
2024
NETL
NSTL
万方数据