计算机与数字工程2025，Vol.53Issue(3)：857-864,8.DOI:10.3969/j.issn.1672-9722.2025.03.041
基于指针方向和刻度定位的仪表读数识别
Reading Recognition of Meter Based on Pointer Direction and Scale Positioning
摘要
Abstract
The existing recognition methods of pointer meter reading network mainly depend on the accuracy of neural network model and have high requirements for image quality.Therefore,combined with the characteristics of pointer and dial,a method of reading recognition of meter based on pointer direction and scale positioning is proposed.This method uses deep convolution neural network to segment the pointer region and the dial region.The rotation center of pointer and the direction point of pointer are deter-mined by the pointer region,then the direction vector of pointer is calculated.The rotation center of pointer and the dial region deter-mine the region of interest including dial.After the polar coordinate transformation,graying and adaptive threshold binarization,the pixel value distribution of scale is counted by the projection method,and the position of scale is determined.Finally,the meter read-ing is calculated.The experimental results show that,compared with the conventional reading recognition method,the average an-gle error of pointer is significantly reduced by 45%,and the average error of reading of meter is as low as 0.071 7.关键词
指针式仪表/读数识别/图像分割/方向向量/刻度定位/极坐标变换Key words
pointer meter/reading recognition/image segmentation/direction vector/scale positioning/polar coordinate transformation分类
信息技术与安全科学引用本文复制引用
邱慧敏,陈黎..基于指针方向和刻度定位的仪表读数识别[J].计算机与数字工程,2025,53(3):857-864,8.基金项目
国家自然科学基金项目(编号:61773297)资助. （编号:61773297）
