摘要
为了提高可见光室内三维定位的准确率与稳定性,提出一种基于深度学习与接收信号强度滤波的可见光室内定位方法。首先,采集每个参考点接收的所有接收信号强度样本,使用聚类算法对接收信号强度样本进行滤波,剔除低值簇与高值簇中的高偏差接收信号强度样本;然后,在高精度无线电地图上训练深度神经网络回归模型,学习接收器位置与接收信号强度样本之间的统计关系。仿真结果表明,方法在6m×6m×3m室内环境下的平均定位误差小于4cm,且支持LED数量的扩展。
To improve the accuracy and stability of the visible light indoors three dimensional positioning,a visible light indoors positioning method based on deep learning and received signal strength indication filtering is proposed.Firstly,all received signal strength indication samples acquired on each reference point are collected,a clustering algorithm is used to filter the received signal strength indication samples,the received signal strength indication samples with significant deviation in the low value cluster and high value cluster are eliminated;then,the deep neural networks regression model is trained on the high accuracy radio map,so as to learn the statistical relationship between the receiver position and the receive signal strength.Simulation results show that the average positioning error of the proposed method is lower than 4cm in a 6m X 6m X 3m room,it also has the ability of LED scalability.
作者
曹海锋
王晓亮
CAO Haifeng;WANG Xiaoliang(Department of information Engineering,Shanxi Engineering Vocational College,Taiyuan 030000,China;School of Computer Science,Northeast Electric Power University,Jilin 132012,China)
出处
《光学技术》
CAS
CSCD
北大核心
2024年第3期361-367,共7页
Optical Technique
基金
山西省教育科学规划院项目(GH-220858)。
关键词
深度学习
数据聚类
无线电地图
接收信号强度
数据滤波
deep learning
data clustering
radio map
received signal strength indication
data filtering
