基于卷积神经网络模型的仪表智能识别算法  

Intelligent Instrument Recognition Algorithm Based on Convolutional Neural Network Model

作　　者：石玮玮 孙辉 李晓峰 程远方 王涛 SHI Weiwei;SUN Hui;LI Xiaofeng

机构地区：[1]国营洛阳丹城无线电厂,河南洛阳471000

出　　处：《科技创新与应用》2022年第32期20-23,共4页Technology Innovation and Application

摘　　要：针对数显游标卡尺字符识别场景,该文提出一种基于卷积神经网络(CNN)检测模型的仪表智能识别系统。首先,从数字式游标卡尺测试现场采集图像样本,并对其分辨率和大小进行归一化;其次建立CNN模型来训练图像样本并提取特征,根据图像特征提取图像样本中的数字显示区域,并提取出游标卡尺中的数字;最后,构建数字式游标卡尺的数据集,并利用浅层神经网络模型对其进行识别。实验测试结果表明,所提出的CNN模型对仪表字符的整体识别率达到95%以上,单个字符识别率为98.86%,远高于其他算法,该模型具有良好的鲁棒性和泛化能力。Aiming at the character recognition scene of digital vernier caliper,this paper proposes an instrument intelligent recognition system based on convolutional neural network(CNN)detection model.First,image samples are collected from the digital vernier caliper test site,and their resolution and size are normalized;then a CNN model is established to train the image samples and extract features,extract the digital display area in the image samples according to the image features,and extract the numbers in the vernier caliper;finally,a dataset of digital vernier calipers is constructed and recognized using a shallow neural network model.The experimental test results show that the proposed CNN model has an overall recognition rate of over 95%for instrument characters,and a single character recognition rate of 98.86%,which is much higher than other algorithms,and the model has good robustness and generalization ability.

关 键 词：目标检测 模式识别 卷积神经网络 数字式仪表 特征提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]