摘要
针对城市综合管廊电力舱地形复杂、人工巡检强度大的问题,提出一种基于四足机器人的综合管廊电力舱巡检方法,开发基于自行走四足机器人的电缆综合巡检平台,利用无监督学习缺陷检测模型,优化模型损失函数,提高了模型识别准确率。通过在雄安新区容东综合管廊实际训练,证明了基于四足机器人的自主巡检体系的适用性。
In order to solve the problems of difficult terrain and high-intensity inspection of cable cabin in urban utility tunnel,a new inspection mothed based on quadruped robot is presented.An inspection system based on quadruped robot is also developed.Using a detection model based on positive sample training,this system optimized the mode loss function and improved the model recognition accuracy.After training for a period of time in Rongdong Urban Utility Tunnel in Xiong’an New Arca,the applicability of the autonomy inspection system based on quadruped robot is proved.
作者
刘海峰
池威威
李志雷
贾志辉
张继超
LIU Haifeng;CHI Weiwei;LI Zhilei;JIA Zhihui;ZHANG Jichao(State Grid Hebei Electric Power Co.,Ltd.Xiong'an New Area Power Supply Company,Xiongan New Area 071000,China)
出处
《河北电力技术》
2022年第4期25-28,59,共5页
Hebei Electric Power
基金
国网河北省电力有限公司科技项目(kj2020-082)。
关键词
综合管廊
缺陷识别
正样本学习
comprehensive pipe gallery
defect detect
positive sample learning
