工厂仪表检测与识别研究现状  

Research Status of Factory Instrument Detection and Identification

作　　者：王德帅 王新佩 武涛 刘泽皓 邵海燕[1,4] WANG De-shuai;WANG Xin-pei;WU Tao;LIU Ze-hao;SHAO Hai-yan(School of Mechanical Engineering,University of Jinan,Jinan 250022,China;Shandong Hardware&Weighing Instrument Association,Jinan 250102,China;Jinan Huibang Automatic Control Instrument Co.,Ltd,Jinan 250014,China;Institute of Control Valve Technology,Jinan 250022,China)

机构地区：[1]济南大学机械工程学院,山东济南250022 [2]山东省五金与衡器行业协会,山东济南250102 [3]济南汇邦自控仪表有限公司,山东济南250024 [4]控制阀技术研究所,山东济南250022

出　　处：《山东工业技术》2024年第6期20-28,共9页Journal of Shandong Industrial Technology

基　　金：企业委托开发项目(W2023356);济南市自主创新团队项目(2019GXRC013)。

摘　　要：本文分析了当前仪表检测和识别的关键方法和技术,主要包括仪表表盘图像的获取方式及设备、表盘图像处理方法和表盘数据提取算法。仪表表盘图像的获取方式除了人工巡检外,还有足式、轮式、轨道式巡检机器人搭载摄像头检测等传感器进行自主巡检。表盘图像处理是为了修正各种干扰,如表盘倾斜、环境光线过暗、曝光或背光、图像模糊和图像遮挡等,以便获得更完整、更准确的仪表图像。表盘数据提取目前有常规的图像处理法和基于神经网络的深度学习法。常规图像处理方法通过特征提取和模式匹配,获取仪表指针数值和刻度盘读数。深度学习方法如卷积神经网络和循环神经网络有助于更准确和高效地对仪表关键特征进行识别。未来发展方向更倾向于基于深度学习模型和优化算法,进一步提升仪表检测的准确性和效率、提高仪表数据处理和分析能力,实现对仪表数据的快速读取从而实现更高精准的检测与识别。This manuscript analyzes the key methods and technologies of current instrument detection and identification,mainly including the acquisition method and equipment of instrument dial images,dial image processing method and dial data extraction algorithm.In addition to manual inspection,the acquisition method of instrument dial images also includes foot-type,wheel-type,and track-type inspection robots equipped with sensors such as camera detection for autonomous inspection.The purpose of dial image processing is to correct various interferences,such as dial tilt,too dark ambient light,exposure or backlight,image blur and image occlusion,so as to obtain a more complete and more accurate instrument image.There are currently conventional image processing methods and deep learning methods based on neural networks for dial data extraction.Conventional image processing methods obtain instrument pointer values and dial readings through feature extraction and pattern matching.Deep learning methods such as convolutional neural networks and recurrent neural networks help to identify key features of instruments more accurately and efficiently.The future development direction is more inclined to be based on deep learning models and optimization algorithms to further improve the accuracy and efficiency of instrument detection,improve instrument data processing and analysis capabilities,and achieve rapid reading of instrument data to achieve more accurate detection and identification.

关 键 词：仪表检测与识别 深度学习 图像处理 数值提取 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]