多尺度卷积残差网络钢轨伤损检测研究  

Rail Damage Detection Based on Multiscale Convolution Residual Network

作　　者：曾树华 黄银秀[3] 黄昌兵[1,2] Zeng Shuhua

机构地区：[1]湖南省高铁运行安全保障工程技术研究中心,湖南株洲412006 [2]湖南铁路科技职业技术学院,湖南株洲412006 [3]湖南化工职业技术学院,湖南株洲412006

出　　处：《工业控制计算机》2023年第9期41-43,共3页Industrial Control Computer

基　　金：湖南省自然科学基金(2020JJ7054)。

摘　　要：针对传统钢轨探伤大多依赖人工获取特征信息、工作繁琐且效率低的问题,提出了融合改进残差模块的多尺度卷积网络钢轨表面伤损图像识别模型。该网络模型采用深度可分离卷积代替传统卷积减少模型参数量,加快模型训练速度;运用残差网络模块,避免网络堆叠过深引起的过拟合现象;并利用多尺度卷积取代了传统的单尺度卷积,使得网络加宽以获取更多的特征信息,增强了模型特征信息的提取能力,从而提高了模型的识别精度。通过对三种典型钢轨伤损进行对比试验,本方法检测精度和运算速度优于传统方法。In view of the problem that traditional rail examination mainly rely on manual extraction of features,which are complex and inefficient,a multiscale convolutional network model for rail damage surface damage image recognition using improved residual modules is proposed in this paper.The network model uses deep separable convolution instead of traditional convolution to reduce the amount of model parameters and speed up model training;Using residual network module to avoid over fitting caused by too deep network stacking,and using multiscale convolution instead of traditional single scale convolution widens the network to obtain more feature information,enhances the ability to extract model feature information,and improves the recognition accuracy of the model.Through comparative tests on three typical rail damage types,the detection accuracy and calculation speed of this method are superior to traditional methods.

关 键 词：图像识别 多尺度卷积 深度可分离卷积 钢轨表面伤损 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程] U216.3[交通运输工程—道路与铁道工程]