基于BP神经网络的RFID室内定位算法研究  

Research on RFID Indoor Location Algorithm Based on BP Neural Network

作　　者：邓昀 朱彦[1,2] 杨逸夫 程小辉 李朝庆 DENG Yun;ZHU Yan;YANG Yi-fu;CHENG Xiao-hui;LI Chao-qing(Guangxi Key Laboratory of Embedded Technology and Intelligent System,Guilin University of Technology,Guilin 541004,China;Department of Information Science and Engineering,Guilin University of Technology,Guilin 541004,China;Key Laboratory of Hunan Province for New Retail Virtual Reality Technology,Hunan University of Commerce,Changsha 410205 .China)

机构地区：[1]广西嵌入式技术与智能系统重点实验室,广西桂林541004 [2]桂林理工大学信息科学与工程学院,广西桂林541004 [3]湖南商学院新零售虚拟现实技术湖南省重点实验室,长沙410205

出　　处：《小型微型计算机系统》2019年第8期1707-1712,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61262075,61662017)资助;广西科技计划项目(桂科AD16380059)资助;广西高等学校资助项目(KY2015YB120)资助

摘　　要：室内定位技术的研究一直都是近年来物联网研究的热点.为了验证一种廉价的RFID(Radio Frequency Identification)设备也有着良好的室内定位效果,提出一种基于廉价的n RF24l01芯片的主动RFID标签与K-means,SVM(Support Vector Machine)和BP(Back Propagation Neural Network)神经网络三种算法相融合的定位算法.首先借助参考标签来建立指纹数据库,通过K-means的聚类算法,把收集到的指纹数据聚成K类,以此将定位区域划分为K个宏区域,再对每个宏区域建立SVM分类模型以及BP神经网络模型.最后采用具体实例对于室内定位性能进行测试.结果表明,当对区域划分为2类,3类,4类的时候,算法的均方根误差分别为1. 0863 M,0. 9265M,0. 9567M,可见当划分3类宏区域时,误差最小,该误差范围满足了室内定位研究的需求.The research of indoor positioning technology has always been a hot topic in the Internet of things in recent years. To verify cheap RFID (Radio Frequency Identification) equipment also has a good effect on indoor positioning,proposing a localization algorithm based on cheap n RF24 l01 chip which combines active RFID tags with K-means,SVM (Support Vector Machine) and BP (Back Propagation Neural Network) algorithms. Firstly,the fingerprint database is built with the help of reference labels,and the collected fingerprint data is clustered into K classes by K-means clustering algorithm. Then the localization region is divided into K macro regions.Then SVM classification model and BP neural network model are established for each macro region. Finally,a specific example is used to test indoor positioning performance. The results showthat the root mean square error of the algorithm is 1. 0863 M,0. 9265 M,0.9567 Mrespectively when the region is divided into 2 categories,3 categories and 4 categories. It can be seen that the error is the smallest when the region is divided into three categories,and the error range meets the needs of indoor positioning research.

关 键 词：室内定位 神经网络 聚类 宏区域 均方根误差 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]