摘要
针对圆形表盘指针仪表需经人工判读、耗时较长读数不精确等问题,设计了一套指针智能识别系统。该系统采用Faster-RCNN法对仪表盘进行定位,提取仪表盘,并对目标图像进行灰度化、k-Means二值化处理;采用旋转虚拟直线拟合方法确定指针,采用角度法实现仪表读数识别。实验结果显示,该系统的仪表读数识别准确率不低于97.87%,可满足工业应用的精确度要求。
Aiming at the problems of manual interpretation,time-consuming and inaccurate readings of circular dial pointer instruments in industry,an intelligent identification system is designed.First the Faster-RCNN method is used to locate the dashboard and extract the dashboard;then the target image is grayed and processed by k-means binary processing.Finally,the rotating virtual straight line fitting method is adopted to determine the pointer,and the angle method is used to achieve meter reading recognition.The experimental results prove that the accuracy of the meter reading recognition of the system is greater than or equal to 97.87%,which meets certain accuracy requirements of industrial applications.
作者
唐霞
苏盈盈
罗妤
王艳玲
TANG Xia;SU Yingying;LUO Yu;WANG Yanling(College of Electrical Engineering,Chongqing University of Science and Technology,Chongqing 401331,China)
出处
《重庆科技学院学报（自然科学版）》
CAS
2021年第2期87-90,共4页
Journal of Chongqing University of Science and Technology：Natural Sciences Edition
基金
重庆市技术创新与应用示范项目“深度学习框架下面向氮氧化物减排的垃圾焚烧状态在线监控系统设计”(CSTC2018JSCX-MSYBX0023)
重庆市自然科学基金项目“非平行平面下融合表面微分几何复杂场景三维立体智能视觉研究及应用”(CSTC2018JCYJAX0239)
重庆市基础研究与前沿探索专项“面向无害化垃圾焚烧发电的二噁英异常排放复合成因诊断方法”(CSTC2019JCYJ-MSXM0220)
重庆科技学院硕士研究生创新计划项目“面向工业的通用指针式多仪表智能识别系统设计”(YKJCX1920410)。
