摘要
针对移动定位和目标感知两大问题,研究软件与硬件协同的巡检机器人系统设计方案。在移动定位方面,设计一种结合激光雷达、单目摄像头、惯性测量单元、GPS等多种传感器的即时定位与地图构建的技术方案,并采用在巡检环境中设置二维码路标辅助视觉定位的方法;在目标感知方面,以表计信息提取任务为例,设计一种基于深度学习的表计检测、定位与文本识别方法。结果表明,通过引入基于5G移动通信及Wi-Fi的网络通信功能实现了具有网络管理平台的巡检机器人系统,系统达到实际应用要求,有效提升了巡检的自动化和智能化水平。
Aiming to two major problems of mobile positioning and object perception,the hardware-software co-design schemes of inspection robot system were explored.For the mobile positioning,a simultaneous localization and mapping strategy by combining LiDAR,monocular camera,inertial measurement unit,GPS and other sensors was designed,and an improved visual positioning scheme was introduced by using the ArUco marker road sign detection.For the object perception,taking the task of meter information extraction as an example,the deep learning based meter detection,localization and text recognition methods were adopted.The results show that,by incorporating the 5G mobile communication and Wi-Fi network communication functions,an intelligent inspection robot system with network management platform is implemented.The developed inspection robot system has met the requirements in practical applications,which effectively enhances the automation and intelligence level of inspection.
作者
毕逢东
周淦
张晨光
姬少英
彭良瑞
闫睿劼
BI Fengdong;ZHOU Gan;ZHANG Chenguang;JI Shaoying;PENG Liangrui;YAN Ruijie(PetroChina Natural Gas Marketing Company,Beijing 100101,China;Beijing Elitenect Technologies Company,Beijing 100085,China;Department of Electronic Engineering,Tsinghua University,Beijing 100084,China;Beijing National Research Center for Information Science and Technology,Beijing 100084,China)
出处
《中国石油大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2024年第3期180-187,共8页
Journal of China University of Petroleum(Edition of Natural Science)
基金
国家自然科学基金项目(U1636124)。
关键词
巡检机器人
移动定位
目标感知
深度学习
文本识别
inspection robot
mobile positioning
object perception
deep learning
text recognition
