摘要
针对突跳式温控器产品编码金属表面反光、凹凸不平以及采用钢印字符导致产品编码识别率下降等问题,提出将字符图像的灰度特征、字符分块占空比以及字符8个特征点之间线段总长度作为神经网络输入的编码识别方法。首先对图像中产品编码区域进行提取、滤波去噪、二值化和字符分割等处理工作;再把分割后的字符归一化为16×16维的图像;最后通过MATLAB编程仿真和工业现场实验得出的结果表明,该方法实现了对温控器编码字符的准确识别,提高了温控器编码相似字符的识别率。
About the problem that the low recognition rate of the product coding is due to the metal surface reflective and uneven and seal character,put forward to use grayscale features and extracted character block duty cycle,and the total length of the character segment between eight points as a neural network input. Firstly,the product coding region in the image is extracted,filtered and denoised,binarized and pre-processed by character segmentation,and then the segmented characters are normalized to 16 ×16 images. Finally,using MATLAB programming to complete the recognition process simulation and industrial field experiment results show that this method realizes the accurate character recognition and improves the recognition rate of similar characters.
引文
[1]陈文华,朱旭洋,陈晓英,等.突跳式温控器的失效分析[J].工程设计学报,2013,20(4):287-292.
[2]刘芳.基于改进型BP神经网络的字符识别算法研究[J].价值工程,2014(10):206-207.
[3]张西,唐恬,李良容.基于神经网络的隧道行车车牌识别技术研究[J].公路,2017(7):205-210.
[4]张国云,向灿群,吴健辉,等.基于改进BP网络的车牌字符识别方法研究[J].计算机应用与软件,2017,34(4):243-248
[5]周喜寿,陈天星.基于小波包分析和BP神经网络的字符识别方法[J].计算机应用技术,2010,3(37):43-45.
[6]朱正礼.基于三层BP神经网络的字符识别系统的实现[J].现代计算机,2006(245):93-95
[7]沈清波,常龙昆.基于字符特征叠加提取与BP神经网络的字符识别[J].辽宁石油化工大学学报,2014,34(6):74-78.
[8]鹿琛,王珊珊.基于BP神经网络的车牌字符识别方法[J].山东农业大学学报(自然科学版),2017,48(1):113-116.
[9]张可,张高燕,吴苏,等.基于BP神经网络的字符识别系统[J].计算机与现代化,2009,161(1):63-65.
[10]罗帅.基于深度特征学习的电子电路故障诊断[D].合肥:合肥工业大学,2017.
[11]李然.基于深度学习的短文本情感倾向性研究[D].北京:北京理工大学,2015.
[12]余胜威,吴婷,罗建桥. MATLAB GUI设计入门与实战[M].北京:清华大学出版社,2016.