基于深度学习的智能电表缺陷检测系统设计  

Design of an Intelligent Electric Meter Defect Detection System Based on Deep Learning

作　　者：任瑾 龚圣高 谢华标 邓鉴坚 龙伟杰 REN Jin;GONG Shenggao;XIE Huabiao;DENG Jianjian;LONGWeijie(School of Electronic and Electrical Engineering,Zhaoqing University,Zhaoqing,Guangdong 526061;Zhaoqing Gaoyao Power Supply Bureau of Guangdong Power Grid,Zhaoqing,Guangdong 526061;Zhaoqing Branch of China Mobile Communications Group Guangdong Co.Ltd.,Zhaoqing,Guangdong 526061,China)

机构地区：[1]肇庆学院电子与电气工程学院,广东肇庆526061 [2]广东电网肇庆高要供电局,广东肇庆526061 [3]中国移动通信集团广东有限公司肇庆分公司,广东肇庆526061

出　　处：《肇庆学院学报》2025年第2期63-68,共6页Journal of Zhaoqing University

基　　金：肇庆市科技创新指导类项目(2023040303006);肇庆学院校级科研基金项目(QN202339)。

摘　　要：智能电表在生产、运输过程中常因多种因素导致显示缺陷或外观信息磨损丢失从而影响电表的正常使用.为此,设计基于深度学习的智能电表缺陷检测系统,通过采集电表图像进行预处理,在处理后的图像中提取出不同的检测区域,将深度学习和机器视觉检测技术融合,分别对LCD显示屏、外观标牌和拉闸灯3个区域进行分类定位和缺陷检测.测试证明该系统检测准确率高,漏检率低,可适用于大规模检测,满足实际应用需求.During the production and transportation processes,smart meters often have display defects or the wear and loss of appearance information due to various factors,thus affecting the normal use of the meters.For this reason,a defect detection system for smart meters based on deep learning is designed.By collecting meter images for preprocessing,different detection areas are extracted from the processed images.The deep learning and machine vision detection technologies are integrated to conduct classification,positioning and defect detection on the three areas of the LCD display screen,the appearance signboard and the switch-off light respectively.Tests have proved that the system has high detection accuracy and a low missed detection rate,and it can be applied to large-scale detection and meet the requirements of practical applications.

关 键 词：深度学习 机器视觉 智能电表 缺陷检测 

分 类 号：TM714[电气工程—电力系统及自动化]