中国电力 ›› 2021, Vol. 54 ›› Issue (2): 147-155.doi: 10.11930/j.issn.1004-9649.202004116

• 电网 • 上一篇    

电力设备红外图像缺陷检测

黄锐勇1, 戴美胜1, 郑跃斌1, 黄勤琴2, 康立烨2, 苟先太2, 周维超3   

  1. 1. 广东电网有限责任公司潮州供电局,广东 潮州 521000;
    2. 西南交通大学 电气工程学院,四川 成都 611756;
    3. 四川赛康智能科技股份有限公司,四川 成都 610041
  • 收稿日期:2020-04-15 修回日期:2020-07-20 发布日期:2021-02-06
  • 作者简介:黄锐勇(1983-),男,高级工程师,从事电力安全方面工作,E-mail:179430690@qq.com;苟先太(1971-),男,通信作者,博士,副教授,从事电网智能化、人工智能技术研究,E-mail:491098063@qq.com
  • 基金资助:
    四川省人工智能重大专项资助项目(电力网络智能化关键技术研究及应用示范,2018GZDZX0043);中国南方电网有限责任公司科技项目(基于巡检机器人红外成像测温智能诊断系统的技术研究,035100KK52190003)

Defect Detection of Power Equipment by Infrared Image

HUANG Ruiyong1, DAI Meisheng1, ZHENG Yuebin1, HUANG Qinqin2, KANG Liye2, GOU Xiantai2, ZHOU Weichao3   

  1. 1. State Grid Chaozhou Electric Power Co., Ltd., Chaozhou 521000, China;
    2. School of Electrical Engineering, Southwest Jiaotong University, Chengdu 611756, China;
    3. Sichuan Scom Intelligent Technology Co., Ltd., Chengdu 610041, China
  • Received:2020-04-15 Revised:2020-07-20 Published:2021-02-06
  • Supported by:
    This work is supported by Major Special Projects of Artificial Intelligence in Sichuan Province(Research and Application Demonstration of Key Technologies for Intelligent Power Network, No.2018GZDZX0043) and Science & Technology Project of CSG (Research on the Technology of Intelligent Diagnosis System Based on Infrared Imaging Temperature Measurement of Patrol Robot, No.035100KK52190003)

摘要: 机器人在巡检过程中采集到的红外图像很难反映设备目标的纹理信息。人工方法或传统机器学习方法不能精准识别和分类电力设备缺陷,同时其他环境因素容易导致误判。采用CenterNet结合结构化定位的算法模型,通过对现场红外图像数据样本收集、训练及验证算法模型的计算,实现从复杂的红外图像中以较高的准确率将不同变电站设备及其部件识别定位出来。根据设备部件表面温度范围值和识别定位出的变电站设备类型,结合相关温度规范实现电力设备红外图像缺陷检测。实验结果表明,该方法提高了电力设备红外图像缺陷检测的检测精度,为电力设备红外图像智能检测提供了新的思路。

关键词: 红外图像, 电力设备, CenterNet, 结构化定位, 缺陷检测

Abstract: The infrared image collected by the robot during inspection is hard to reflect the texture information of the equipment target. The artificial methods or traditional machine learning methods cannot accurately identify and classify the defects of power equipment, and other environmental factors may easily lead to false judgment. In this paper, the algorithm model of CenterNet combined with structured positioning is adopted. Through collecting field infrared image data samples, the algorithm model is trained and verified to identify and position different substation equipment and its components with high accuracy from complex infrared images. According to the surface temperature range of equipment components and the type of substation equipment, the infrared image is combined with relevant temperature specifications to realize the defect detection of power equipment. The experimental results show that this method improves the accuracy of infrared image for detecting the defects of power equipment, and provides a new idea for infrared image used for intelligent detection of power equipment.

Key words: infrared image, power equipment, CenterNet, structured positioning, defect detection