一种基于异构串联卷积神经网络的铁路异物检测模型

Railway Foreign Body Detection Model Based on Heterogeneous Series Convolutional Neural Network

针对目前铁路异物侵限报警系统中存在的误报率较高,物体分类能力有限的问题,本文设计了一套非接触式异物侵限报警系统.提出一种采用异构串联卷积核和加入归一池化层的卷积神经网络模型用于系统图像识别,通过对比试验和系统测试,该模型在小样本训练情况下具有较好的泛化能力,能有效保障系统的稳定性与精确度.

A set of non-contact foreign body intrusion alarm system is presented to solve high false positive rate and limited object classification capabilities in current railway foreign body intrusion alarm system.This paper proposed a convolutional neural network model with heterogeneous series convolution kernel and normalized pool layer for image recognition.Through comparison experiments and system tests,the model has better generalization ability in the case of small sample training,which can effectively guarantee the stability and accuracy of the system.