摘要
摩擦片的裂纹数目和长度是衡量高铁制动性能的核心评定标准之一,有效的裂纹检测对高铁的安全运行具有重要意义。提出基于CSPDarkNet53主干网络架构的改进算法,实现摩擦片裂纹的在线自动检测。一方面融合双路特征提取网络以增强对于裂纹特征检测的敏感度,有效提高摩擦片裂纹检测的准确率;另一方面在YOLO检测模块预测框的去冗余计算环节中,采用目标框加权融合算法(weighted fusion algorithm of target box,WBF)降低误检率。实验结果表明,相较于当前最具有代表性几类目标检测算法,本文采用的方法准确率显著提高,平均精度提升7.64%。
The number and length of cracks in the friction lining is one of the core evaluation criteria to measure the braking performance of high-speed railways.Effective crack detection is of great significance to the safe operation of high-speed railways.This paper proposes an improved algorithm based on the backbone network architecture of CSPDarkNet53 to realize online automatic detection of friction plate cracks.Firstly,the dual-path feature extraction network is fused to enhance the sensitivity to crack feature detection and effectively improve the accuracy of friction plate crack detection;Secondly,in the de-redundancy calculation of prediction box of YOLO detection module,the weighted fusion algorithm of target box(WBF)is used to reduce the false detection rate.The experimental results show that compared with the current most representative types of target detection algorithms,the accuracy of the method used in this paper is significantly improved,and the average accuracy is increased by 7.64%.
作者
张景博
汪日伟
刘凤连
温显斌
ZHANG Jingbo;WANG Riwei;LIU Fenglian;WEN Xianbin(Key Laboratory on Computer Vision and Systems,Ministry of Education of China,Key Laboratory on Intelligence Computing and Novel Software Technology of the City of Tianjin,Tianjin University of Technology,Tianjin 300384,China;WenZhou University OuJiang College,Zhejiang 325035,China)
出处
《光电子.激光》
CAS
CSCD
北大核心
2021年第9期962-969,共8页
Journal of Optoelectronics·Laser
关键词
裂纹检测
目标检测
深度学习
摩擦片
crack detection
target detection
deep learning
friction flakes