基于图像处理的数字仪表字符自动识别研究  

Automatic character recognition and simulation of digital instrument based on image processing

作　　者：刘晶 LIU Jing(School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094,China)

机构地区：[1]南京理工大学计算机科学与工程学院,南京210094

出　　处：《信息技术》2020年第4期84-87,91,共5页Information Technology

摘　　要：为提高数字仪表图像识别率,结合机器视觉技术,在图像预处理的基础上,提出一种PSO的LSSVM参数优化方案。在图像预处理阶段,采用Hough变换校正,并通过形态学处理和二值化,以提高图像采集的效率;然后采用连通域的方式实现对图像区域的定位分割,并通过七段码实现图像特征提取;针对数字仪表识别是多分类的问题,采用OAO分类法将LSSVM算法拓展到多分类,并引入PSO算法对LSSVM参数进行优化;最后以OpenCV开源库、MATLAB作为工具,对上述算法进行验证。结果表明,0~9数字字符识别率为99.2%,整体图像样本测试识别率为99.6392%,高于其他算法,体现了较好的识别效果。In order to improve the recognition rate of digital instrument image,combining with machine vision technology and based on image preprocessing,a PSO LSSVM parameter optimization scheme is proposed to verify the efficiency of image recognition.In order to improve the efficiency of image acquisition,Hough transform correction is used in image preprocessing stage,and morphological processing and binarization are used to improve the efficiency of image acquisition.Then,connected area method is used to locate and segment image regions,and seven-segment code is used to extract image features.LSSVM is aimed at binary classification,while digital instrument recognition is used.The LSSVM algorithm is extended to multi-classification by using OAO classification method,and the parameters of LSSVM are optimized by introducing PSO algorithm.Finally,the above algorithms are validated by using OpenCV open source library and MATLAB.The results show that the image processing scheme constructed in this paper has a recognition rate of 99.2%for 0-9 digital characters.The recognition rate of the whole image sample test is 99.6392%,which is higher than other algorithms,reflecting better superiority.

关 键 词：机器视觉 仪表图像 LSSVM算法 OAO分类 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]