摘要
针对目前主机厂新造、高级修的高速动车组与城轨车辆的电气调试工序中供应商设备与诊断维护软件缺少标准对外通信接口,数字化调试平台无法获取调试数据,从而形成“数据孤岛”的问题。文章提出了一种基于PaddleOCR框架与python语言的数据采集系统,该系统可以实现对用户选定的目标区域进行图像捕捉与处理、版面分析与恢复、字符切割与识别,然后将截图中的文本信息输出为可编辑的文本格式,最终通过http通信协议上传到调试专用的数字化平台。本系统首次将基于深度学习框架的新一代OCR技术应用于我国轨道车辆的制造生产中,并辅以良好的人机交互界面降低车间工人操作难度。同时通过模型训练与算法应用识别结果准确度得到大幅提升,试验结果表明对现场使用的界面进行识别后字符识别准确率在97%以上,可以满足调试工序的生产需要。
The current electrical commissioning process for newly built and overhauled high-speed EMUs and urban rail vehicles at OEMs lacks a standard external communication interface between supplier equipment and diagnostic maintenance software,and the digital commissioning platform is unable to obtain commissioning data,thus forming a“data island”problem.This paper proposes a data acquisition system based on the PaddleOCR framework and Python language.The system is capable of performing image capture and processing,layout analysis and reconstruction,character segmentation and recognition for user-selected target regions,subsequently converting the extracted text information from screenshots into editable text formats.The processed data is ultimately transmitted to a dedicated digital debugging platform via the HTTP communication protocol.This system represents the first application of a new-generation deep learning-based OCR technology in China's rail vehicle manufacturing process,complemented by an intuitive human-machine interface(HMI)to significantly reduce operational complexity for workshop personnel.Through model training and algorithmic optimization,the recognition accuracy has been significantly improved.Test results demonstrate that the character recognition accuracy rate exceeds 97%for on-site interface captures,fully meeting the production requirements of debugging processes.
作者
宋清波
张春光
SONG Qingbo;ZHANG Chunguang(CRRC Tangshan Locomotive&Rolling Stock Co.,Ltd.,Tangshan 063000,China;School of Automation and Electrical Engineering,Dalian Jiaotong University,Dalian 116028,China)
出处
《智慧轨道交通》
2026年第1期23-27,共5页
INTELLIGENT RAIL TRANSIT
