多类指针式仪表识别读数算法研究  

Research on pointer identifying and number reading algorithm of multi-class pointer instruments

作　　者：张雪飞 黄山[1] Zhang Xuefei;Huang Shan(School of Electrical Engineering and Information,Sichuan University,Chengdu 610065,China)

机构地区：[1]四川大学电气信息学院,成都610065

出　　处：《电测与仪表》2020年第16期147-152,共6页Electrical Measurement & Instrumentation

摘　　要：针对多类指针式仪表,文中研究可以进行统一自动识别读数的算法框架。基于指针式仪表的结构特性,结合特征点的不可替代性提出,根据尺度不变特征点(Scale-Invariant Feature Transform,SIFT)匹配程度自动检测仪表类别。采用快速直线检测(Line Segment Detector,LSD)提取待定直线区域,创新地利用直线边缘的方向性约束得到指针两侧边缘,完成指针识别。研究中设计快速生成模板工具,保存仪表刻度角度-读数映射表,促成多类仪表自动读数。在此基础上,添加万用表评估算法优劣综合评估检测识别率、提取指针、读数精确度,满足设计要求,最终算法运行在Android系统的手持仪设备上。Considering existing multi-class pointer instruments in factories,this paper provided an algorithm framework focusing on identifying the pointer and reading the number automatically.Based on thestructural property of pointer instrument,combining with the non-substitutability,it automatically recognizes the category of instruction being tested according to the matched feature of scale-invariant feature transform(SIFT).After that,it extracts set of line regions with line segment detector(LSD)algorithm,and utilizes the specific line with direction limit to obtain bilateral margin of pointer to complete the pointer identification.Otherwise,a novel model generation tool is designed to record a mapping table between angle and number to enable automatic reading of multi-class meters.On this basis,the multimeter evaluation algorithm is added to comprehensively evaluate the detection and recognition rate,increase the accuracy of pointer extracting and tnumber reading,so as to meet the design requirements.Finally,the algorithm has worked on Android system with stable result.

关 键 词：多类仪表 SIFT LSD 自动读数 万用表 

分 类 号：TM93[电气工程—电力电子与电力传动]