基于图像处理的核设施仪表识别研究及应用  

Research and Application of Instrument Recognition of Nuclear Facility Based on Image Processing

作　　者：张明星[1] 干阳琳 谢超 ZHANG Mingxing;GAN Yanglin;XIE Chao(Nuclear Power Institute of China,Chengdu,Sichuan Province,610213 China)

机构地区：[1]中国核动力研究设计院,四川成都610213

出　　处：《科技资讯》2025年第6期49-52,共4页Science & Technology Information

摘　　要：为了降低核设施仪表的改造成本和核安全风险,需要采用非接触式的方式对仪表数据进行采集,应用图像处理技术对数字仪表自动识别进行研究和应用。首先,利用开源计算机视觉库(Open Source Computer Vision Library,OpenCV)实现图像裁剪、变换、色阶调整、灰度处理、阈值处理等预处理。在阈值处理方面,针对最大类间方差(OTSU)、自适应和基于区域生长3种阈值算法进行对比实验,确认了OTSU阈值算法为最优算法。其次,通过Tesseract方法和穿线法分别对不同类型的仪表图像进行字符识别。最后,在岷江试验堆上选取不同工艺设备的数字仪表进行应用。结果表明,本文提出的方法对数码管式字体的识别正确率达92%以上、对印刷体数字的识别正确率达96%以上,能够满足核设施仪表参数的自动采集需求。In order to reduce the cost of instrument renovation and nuclear safety risks in nuclear facilities,non-contact methods need to be used to collect instrument data,and image processing technology should be applied to study and apply digital instrument automatic recognition.Firstly,utilizing the Open Source Computer Vision Library(OpenCV)technology to implement pre-processing such as image cropping,transformation,color grading adjustment,grayscale processing,threshold processing,etc;In terms of threshold processing,comparative experiments were conducted on three algorithms:OTSU,adaptive threshold algorithm,and region growing threshold algorithm,and it was confirmed that OTSU threshold algorithm is the optimal algorithm.Secondly,character recognition is performed on different types of instrument images using Tesseract method and threading method respectively;Finally,digital instruments with different process equipment were selected for application on the Minjiang experimental reactor.The results show that the recognition accuracy of digital tube style fonts is over 92%,and the recognition accuracy of printed numbers is over 96%,so the method in this paper can meet the automatic acquisition requirements of nuclear facility instrument parameters.

关 键 词：核设施仪表 图像处理技术 OPENCV OTSU 阈值算法 穿线法 

分 类 号：TM623[电气工程—电力系统及自动化]