基于机器视觉的机械式表盘自动读表技术的实现-《郑州大学学报(工学版)》

文章信息/Info

Title:: Implementation of Automatic Reading MechanicalGauge Technique Based on Machine Vision

Author(s):: CHEN Ji-hua; Ll Yong; TIAN Zeng-guo; LI Lei; School of Physical Engineering,Zhengzhou University 450001,China

Keywords:: mechanical gauge; canny operator; hough transform; image recognition; automatic-meter reading

摘要:: 针对当前机械式表盘自动读表技术存在计算量大、需采集多幅图像以及需要标定等问题,提出了基于改进的Canny边缘检测算子的指针旋转中心求解算法,该方法可基于单帧表盘图像快速解算出其回转中心及半径,并定位出表盘的有效显示区域,避免了复杂背景的干扰;基于表盘指针所在直线经过圆心这一前提,设计了过定点的Hough直线检测算法,显著降低了表盘指针检测时间.详细论述了算法的设计原理,并给出时间复杂度分析.实验表明:该方案算法稳定且计算量小,检测速度提高4~5倍，并可减少存储空间,具有良好的工程实用价值.

Abstract:: In order to overcome the shortcomings of current automatic recognition schemes applied to mechanicalgauges including huge computing quantity,calibration based on multi dial’s images,this paper proposes a rotationcenter acquisition method based on improved Camny edge detection operator,which can acquire rotation center aswell as radius quickly based on single dial image. and designs hough transform algorithm passing by a specifiedpoint based on precondition that dial’s pointer centerline and rotation center locate on same line,which decreasesdetecting time significantly. Design principle and time complexity analysis of algorithms are introduced in detail.Experimental results demonstrate the algorithms stable with low computational complexity,detecting speed is 4 ~5times faster with memory space reduced. All of which verifies algorithms’ possibility in engineering application.