摘要
针对基于机器视觉的仪表自动校准系统,提出了一种在HSV颜色空间中进行指针仪表读数识别的方案。首先通过对比指针位置不同的两幅表盘图像获得不包含指针的参考表盘图像,然后在其明度分量中通过Hough圆检测进行表盘中心定位,在饱和度分量中进行刻度区域定位,并采用连通域标记法定位刻度线质心,用这些质心和表盘中心建立读数识别参考系统。针对待识别的表盘图像,基于帧差分法进行指针区域定位并确定指针的最小外接矩形,将其质心视为指针质心以克服质心偏移问题,在检测到刻度线质心和指针质心后,采用角度法进行读数识别。实验结果表明该方法能精确有效地确定指针仪表读数。
Aiming at automatic calibration systems using machine vision, an HSV color space based method is presented for determining the readings of a pointer gauge. The reference image of the dial plate without pointers is first obtained by comparing two gauge images with pointers in different positions. The center of the dial region is then determined using Hough circle detection in the luminance value component, and the scale region is detected in the saturation component. The centroids of scale marks are finally located, and a reference system for reading recognition is built up with these centroids and the center of the dial region. Given a gauge image corresponding to an indicating value, it is subtracted from the reference image to detect the pointer region. The minimal bounding rectangle of the pointer region is determined and its centroid is then taken as the centroid of the pointer region to overcome the problem of centroid deviation. Finally, the indicating value is calculated using an angle method by putting the centroid of the pointer in the reference system for reading recognition. Experimental results reveal that this method is able to determine the readings of the pointer gauge accurately and efficiently.
作者
张志锋
王凤琴
田二林
薛冰
李祖贺
ZHANG Zhi-feng;WANG Feng-qin;TIAN Er-lin;XUE Bing;LI Zu-he(School of Software Engineering,Zhengzhou University of Light Industry,Zhengzhou 450002,China;School of Computer and Communication Engineering,Zhengzhou University of Light Industry,Zhengzhou 450002,China;Henan Relations Co.,Ltd.,Zhengzhou 450001,China)
出处
《控制工程》
CSCD
北大核心
2020年第3期581-586,共6页
Control Engineering of China
基金
国家自然科学基金项目(61975187,61873246,61702462,61702464)
河南省科技攻关计划项目(182102210607,192102210108,192102210243)
河南省高等学校重点科研项目计划(15A510041,16A520028)
郑州轻工业大学博士科研基金项目(2014BSJJ077,2018BSJJ040)
郑州轻工业大学星空众创空间孵化项目(2018ZCKJ318)。
关键词
机器视觉
指针仪表
读数识别
帧差分
最小外接矩形
Machine vision
pointer gauge
reading recognition
frame difference
minimal bounding rectangle
作者简介
张志锋(1978-),男,河南郸城人,硕士,副教授,主要从事软件工程和人工智能等方面的教学与科研工作;通讯作者:李祖贺(1983-),男,河南南召人,博士,讲师,主要从事计算机视觉和机器学习等方面的教学与科研工作。
