基于视觉技术的X、γ剂量率仪数字识别系统  

X、γ doserate meter digital recognition system based on computer vision technology

作　　者：王雨青 黄政林 刘新昊 李英帼 韦应靖[1,4] WANG Yuqing;HUANG Zhenglin;LIU Xinhao;LI Yingguo;WEI Yingjing(China Institute for Radiation Proctection,Taiyuan 030006,China;Shanxi Key Laboratory for Radiation Safety and Protection,Taiyuan 030006,China;CNNC Key Laboratory for Radiation Protection Technology,Taiyuan 030006,China;Institute of Nuclear and New Energy Technology,Tsinghua University,Beijing 100084,China)

机构地区：[1]中国辐射防护研究院,山西太原030006 [2]辐射安全与防护山西省重点实验室,山西太原030006 [3]中核集团辐射防护技术重点实验室,山西太原030006 [4]清华大学核能与新能源技术研究院,北京100084

出　　处：《现代电子技术》2025年第6期118-126,共9页Modern Electronics Technique

摘　　要：为提高X、γ剂量率仪检定、校准的自动化程度,提出一种基于计算机视觉的X、γ剂量率仪数字识别系统。该系统可实现摄像头视频采集、采集图像预处理、仪表图像的文本检测与识别以及识别结果的后处理。分别采用DBNet与CRNN-CTC作为文本检测与文本识别的模型,比较不同骨干网络结构对模型的影响。在平衡准确性和速度指标后,选择MoblieNetV3作为文本检测和识别模型的骨干网络,对32种常见的X、γ剂量率仪进行识别实验。结果表明:经算法过滤后,仪器识别准确率可达到100%;对于绝大部分仪器,使用数字识别系统效率可以提高20%以上,但是对于显示界面刷新面积较大的仪器,摄像机难以识别,仍需要人工检定。In order to improve the automation degree of X、γ doserate meters verification and calibration,a X、γ doserate meter digital recognition system based on computer vision technology is proposed.The system can realize camera video acquisition,collected image preprocessing,text detection and recognition of instrument images and post-processing of recognition results.DBNet and CRNN-CTC(convolutional recurrent neural network-connectionist temporal classification)are used as text detection and text recognition model,respectively,and the impact of different backbone networks on models is compared.After balancing accuracy and speed indicators,MoblieNetV3 is selected as the backbone network for text detection and recognition model.The recognition tests were conducted on 32 common X、γ doserate meters.The results show that after the filtering algorithm,the instrument recognition accuracy can reach 100%.For most of the instruments,the use of digital recognition system can improve the efficiency of more than 20%,but for the display interface refresh area of large instruments,the camera is difficult to identify,still need manual verification.

关 键 词：数字识别系统 机器视觉技术 深度学习 图像处理 文本检测 文本识别 计量检定 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP274[电子电信—信息与通信工程]