列车走行部远程监测及图像AI识别技术研究

Research on remote monitoring and image AI recognition technology of train running gear

随着我国对轨道交通建设的大力投入,铁路车辆的状态监测与故障修复的相关要求也日益增长。尤其在当前提速、重载背景下,列车走行部故障的发生率显著增加。针对这一问题,文中基于MIMX RT1064高效图像微处理器,采用OpenART mini高速摄像头获取行进中的列车走行部图像,再通过ESP8266 Wi-Fi模块高频远程无线传输到PC端,在图像进行滤波降噪处理后利用InceptionNet-V3卷积神经网络模型进行训练,最后反馈给OpenART mini摄像头,从而能更准确实时监测走行部状态。结果表明,该系统在能完成列车走行部图像高频稳定传输的同时,也能准确可靠地识别列车转向架故障。

With the strong investment in the construction of rail transportation,the relevant requirements for the condition monitoring and fault repair of railway vehicles also rise.Particularly under the background of speed rise and heavy load,the incidence of train running gear failure increases substantially.To fix this problem,based on MIMX RT1064 efficient image microprocessor,this paper uses OpenART mini highspeed camera to obtain the moving train running gear image,and then transmits it to PC through ESP8266 Wi-Fi module.After the image is filtered and de-noised,the InceptionNet-V3 convolutional neural network model is used for training,and the final feedback is provided to OpenART mini so that the running state can be checked more precisely in real time.The results indicate that the system can not only complete the high frequency and steady transmission of the train running part image,but also accurately and reliably identify the train bogie faults.