摘要
基于接触网安全巡检装置(2C)采集的海量图像数据,提出高铁接触网异物自动化智能检测方法,以实现稳健、可靠、精准的高铁接触网安全异常检测。该方法面向2C图像的特点以及接触网安全运行需求,首先对图像进行预处理,然后设计基于深度神经网络的异物检测方法,利用已标定样本训练异物检测模型,并通过预训练和重训练步骤进行深度学习模型的优化,最后将训练好的模型应用于真实场景中对特定异物进行自动检测。对采集的2C图像进行相关试验,结果表明,该方法可以快速有效地检测出接触网异物,准确率达到96.5%以上,具有较高的应用价值。
An intelligent foreign object detection method for OCS of high speed railways is developed based on a large number of images acquired by the OCS safety patrol device(2C), to achieve stable, reliable and accurate detection for OCS system. This method makes full use of features of 2C images and targets safety requirements of OCS system, and the 2C images are preprocessed before a foreign object detection method is designed based on deep neural network, where a calibrated sample is used to train the foreign object detection model and optimize it through pre-training and re-training. The model is then used in real operation scenarios for automatic detection. Tests on acquired 2C images show that this method can detect OCS foreign objects quickly and efficiently with accuracy over 96.5%, being of high application value.
作者
徐伟
吴泽彬
刘建新
丁道华
詹天明
徐洋
XU Wei;WU Zebin;LIU Jianxin;DING Daohua;ZHAN Tianming;XU Yang(Nanjing Power Supply Section,China Railway Shanghai Group Co Ltd,Nanjing Jiangsu 210011,China;School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing Jiangsu 210094,China;Nanjing Zhiliansen Information Technology Company,Nanjing Jiangsu 210012,China)
出处
《中国铁路》
2019年第10期39-44,共6页
China Railway
关键词
高速铁路
接触网
异物
智能检测
深度学习
自动化
high speed railway
OCS
foreign objects
intelligent detection
deep learning
automation
