基于改进YOLOv5轻量化的指针式仪表识别与读数方法研究  

Research on the Identification and Reading Method of Pointer Instrument Based on Improved YOLOv5 Light Weight Model

作　　者：陈群伟 连晖 刘宇轩 CHEN Qunwei;LIAN Hui;LIU Yuxuan(State Grid Fujian Electric Power Co.,Ltd.Ningde Power Supply Company,Ningde 352000,China)

机构地区：[1]国网福建省电力有限公司宁德供电公司,宁德352000

出　　处：《机电技术》2024年第4期43-47,120,共6页Mechanical & Electrical Technology

摘　　要：文章针对指针式仪表自动读数存在的读数精度低、鲁棒性差及实时性不足等问题,提出了一种基于改进YOLOv5轻量化模型的指针式仪表识别与读数的方法。首先,采用改进的YOLOv5识别并定位复杂场景中待检测仪表的位置,提取仪表区域;然后,通过特征点匹配与透视变换对仪表图像进行校正;最后,利用M-LSD算法提取指针并通过直线拟合计算出指针所指向的刻度读数。试验结果表明:文章提出的方法最大误差为0.05mA,在最小刻度范围内,该方法具备良好的鲁棒性、精确性和实时性,可满足不同复杂场景下的实际应用。To solve problems such as low reading precision,poor robustness,and lack of real-time automatic reading of pointer meters,this paper puts forward a method of identification and reading of pointer meters based on a lightweight model.First,the improved YOLOv5 is used to identify and locate the position of the instrument to be detected in the complex scene,and the instrument area is extracted.Then,the instrument image is corrected by feature point matching and perspective transformation.Finally,the M-LSD algorithm is used to extract the pointer and calculate the scale reading of the pointer by straight line fitting.The experimental results show that the maximum error of the proposed method is 0.05mA,and the proposed method has good robustness,accuracy,and real-time performance within the minimum scale range,and can meet the practical application in different complex scenarios.

关 键 词：指针式仪表 YOLOv5 识别定位 图像校正 直线拟合 仪表读数 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程] TH86[机械工程—仪器科学与技术]