摘要
基于永磁体的定位技术为运动跟踪、机器人定位导航和医疗器械跟踪领域提供了一种无线、高精度、低成本的解决方案。为解决基于磁偶极子模型和LM(Levenberg-Marquardt)算法的定位方法过于依赖初始值、计算耗时受限的问题,利用基于磁偶极子模型先验知识的约束条件构造惩罚函数,提出一种融合密集卷积网络(DenseNet)和注意力机制(SE Block)的永磁定位方法。实验结果表明:在48~118 mm的高度范围内,本文方法定位精度可达(1.79±1.05)mm和1.12°±0.53°,平均计算耗时降低至1.6 ms,提升了永磁定位系统计算的速率和稳定性。
Tracking technology based on permanent magnetic provide a wireless,high-precision,low-cost solution for the fields of motion tracking,robot positioning and navigation,and medical device tracking.Aiming at the problems that the tracking approach based on the magnetic dipole model and Levenberg-Marquardt(LM)algorithm depends too much on initial values and limited computing time,the penalty function is constructed using constraint condition of priori knowledge based on magnetic dipole model,a permanent magnetic tracking approach fuses DenseNet and SE Block is proposed.The experimental result shows that the tracking precision is(1.79±1.05)mm and 1.12°±0.53°in height range of 48~118 mm,and the average computing time is reduced to 1.6 ms.This approach improves the computing speed and computational stability of the permanent magnetic tracking system.
作者
郭鹏飞
戴厚德
杨千慧
姚瀚晨
黄巧园
GUO Pengfei;DAI Houde;YANG Qianhui;YAO Hanchen;HUANG Qiaoyuan(Fujian Institute of Research on the Structure of Matter,Chinese Academy of Sciences,Fuzhou 350002,China;University of Chinese Academy of Sciences,Beijing 100049,China;School of Informatics,Xiamen University,Xiamen 361005,China)
出处
《传感器与微系统》
CSCD
北大核心
2024年第2期37-40,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61973293)
中国科学院国际伙伴计划对外合作重点项目(121835KYSB20190069)
中央引导地方科技发展专项资金项目(20d0581bfa,2021L3047)。
关键词
磁定位
深度学习
密集卷积网络
注意力机制
magnetic tracking
deep learning
dense convolutional network(DenseNet)
attention mechanism