基于卷积神经网络的接触网吊弦端点检测算法

Convolution Neural Network based Algorithm for Inspection of End Points of OCL Droppers

吊弦作为高速铁路接触网接触悬挂装置中的重要组成部分,对接触网系统的安全稳定具有至关重要的作用。由于吊弦自身结构细小难以定位,使得利用计算机进行故障自动检测成为一个难题。针对上述问题及吊弦结构特点,本文提出一种基于卷积神经网络的吊弦端点检测算法,将残差网络(ResNet)作为特征提取网络,利用反卷积层将输出的低分辨率特征图转变为高分辨率特征图,并对吊弦端点位置进行预测。实验测试结果表明,该算法对吊弦端点坐标预测具有较好的效果,具有良好的可行性。

As an important part of OCL suspension device of high speed railway,droppers are playing an important role in the safety and stability of OCL.Because of its small structure and difficult to locate,it is difficult to inspect the fault automatically by computer.In view of the above problems and the structural characteristics of droppers,this paper proposes an algorithm of dropper end point inspection based on convolutional neural network.The residual network(ResNet)is used as the feature extraction network,and the output low-resolution feature map is transformed into high-resolution feature map by using deconvolution layer,and the position of dropper end point is predicted.The experimental results show that the algorithm has good effect and good feasibility for the prediction of the end point coordinates of droppers.