摘要
针对气体检测仪型式评价试验中检测任务繁重,劳动强度大等问题,提出开发一套3工位气体检测仪图像示值自动检测系统。该系统在原有型式评价试验的硬件基础上,利用机器视觉技术和数字图像处理方法,自动地完成对仪器示值图像的预处理、特征数字分割,并利用神经网络技术,准确识别字符数值并保存供后续处理。为适应不同类型数字显示仪器示值识别的需要,该系统选用自适应较强、可预训练的Back Propagation(BP)神经网络算法作为数字识别算法。通过对5组不同样式的规范示值样本进行测试,测试结果表明,示值识别准确率可达100%,验证该系统对多样式数字示值均具有良好的识别效果。
An automatic detection system for the image indication of 3-station gas monitors is proposed in this paper,which can solve the heavy inspection tasks and labor intensity in the type evaluation test for the gas monitors.Based on the original hardware system of the type evaluation test,machine vision technology and digital image processing technology were used to the preprocessing of indication image and the segmentation of the characteristic numbers.The neural network technique was used to recognize character values,which would be saved for the subsequent treatment.Back Propagation(BP)neural network with strong adaptive and pre-training advantages was used to recognize character values,which can adapt to different types of instruments with variable indications.Five groups of samples with different standard indications were tested.The results show that the indication recognition accuracy can achieve 100%,which proved the system has a good recognition effect on multiform digital indication.
作者
蒋晓慧
王岩岩
暴海霞
沈皆磊
JIANG Xiaohui;WANG Yanyan;BAO Haixia;SHEN Jielei(Suzhou Institute of Metrology,Suzhou 215000,China;Wenzheng College of Soochow University,Suzhou 215000,China)
出处
《中国测试》
CAS
北大核心
2020年第S02期89-93,共5页
China Measurement & Test
基金
江苏省质量技术监督局科技项目(KJ18ZB03)
关键词
机器视觉
气体检测仪型式评价试验
BP神经网络
数字识别
machine vision
gas monitors type evaluation test
BP neural network
character values recognition
