基于RSS指纹的室内定位方法

发布时间：2017-12-28 20:16

  本文关键词：基于RSS指纹的室内定位方法　出处：《湘潭大学》2016年硕士论文　论文类型：学位论文

  更多相关文章： 位置指纹 室内定位 解耦 定位精度 RSS 无线局域网

【摘要】：近年来,基于无线接入点(Access Point,AP)接收信号强度(Received Signal Strength,RSS)的位置指纹(Location Fingerprinting,LF)室内定位技术已成为国内外位置感知研究的热点。基于RSS指纹的室内定位技术通过利用在感兴趣区域(Range of Interest,ROI)采集到的来自各个AP的RSS值来推断观察者或场景内物体的坐标,且事先不需知道AP的位置。此外基于位置指纹的定位技术不需要添加额外的硬件支持,可以方便的应用到移动设备中,展现出明显的优势。论文深入研究了基于RSS的室内定位技术,并分析了提高定位精度的方法。本文提出基于位置指纹的解耦室内定位,通过对X轴和Y轴独立地进行定位决策,以期在减少决策代价的同时提高定位精度。论文主要工作如下:首先,给出了基于RSS的无线局域网(Wireless Local Area Network,WLAN)室内定位系统的研究依据。并分析了几种基于WLAN的室内定位系统的优缺点,对比得出基于LF的定位技术是目前应用最多、发展前景最好的方法。其次,分析概括了定位技术的理论基础。本文先从WLAN网络的组成、拓扑结构及工作模式简要描述了无线局域网。在此基础上,给出当前应用较为广泛的几种在WLAN环境下的室内定位技术,并重点阐述了LF定位原理。然后介绍了三种应用最多的定位算法,即K近邻(K-nearest Neighbor,K-NN)法、朴素贝叶斯法和支持向量机(Support Vector Machine,SVM)法。接下来,实现了基于最小二乘支持向量机(Least Squares Support Vector Machine,LS-SVM)的LF室内定位。本文首先提出基于LS-SVM的指纹定位模型,再给出了LS-SVM指纹样本训练的具体实现过程。接下来详细阐述了如何利用一对一和一对多方法将多分类问题转化为多个二值分类问题。仿真结果表明,较传统SVM、K-NN方法的分类准确率高且计算代价小。最后,实现了基于RSS指纹的解耦室内定位。文章首先提出了轴向解耦位置指纹定位框架,然后给出了解耦定位的训练与预测的具体实现过程;最后将LS-SVM、SVM、K-NN等分类器应用到解耦定位框架中,实验结果验证了解耦定位方法的有效性。
[Abstract]:In recent years, location Fingerprinting (Location Fingerprinting, LF) indoor location technology based on Access Point (AP) Received Signal Strength has become a hot research topic in location awareness at home and abroad. The indoor location technology based on RSS fingerprint is used to infer the coordinates of objects in observers or scenes by using RSS values collected from Range of Interest (ROI) collected in different regions, and do not need to know the location of AP in advance. In addition, location technology based on location fingerprint does not require additional hardware support, which can be easily applied to mobile devices, showing obvious advantages. This paper studies the indoor positioning technology based on RSS, and analyzes the methods to improve the positioning accuracy. In this paper, a location based fingerprint decoupling indoor location is proposed. Location decision is made independently by X axis and Y axis, in order to reduce the cost of decision-making and improve positioning accuracy. The main work of this paper is as follows: first, the research basis of the Wireless Local Area Network (WLAN) indoor location system based on RSS is given. The advantages and disadvantages of several indoor location systems based on WLAN are analyzed. It is concluded that the location technology based on LF is the most widely applied and the best way of development. Secondly, the theoretical basis of positioning technology is analyzed and summarized. This paper briefly describes the wireless LAN from the composition, topology and working mode of the WLAN network. On this basis, several kinds of indoor positioning technology which are widely used in WLAN environment are given, and the principle of LF positioning is emphasized. Then we introduce three most widely applied location algorithms, namely K K-nearest (Neighbor), K-NN, naive Bayes and support vector machine (Support Vector Machine). Next, the LF indoor location based on the Least Squares Support Vector Machine (LS-SVM) is implemented. This paper first proposes a LS-SVM based fingerprint location model, and then gives the specific implementation process of LS-SVM fingerprint sample training. Then we elaborate on how to transform the multi classification problem into multiple two value classification problems by one to one and one to many methods. The simulation results show that the classification accuracy of the traditional SVM and K-NN methods is high and the calculation cost is small. Finally, the decoupling indoor location based on RSS fingerprint is realized. This paper first puts forward the axial decoupling of fingerprint positioning frame, and then gives the concrete realization of the process of training and prediction of the decoupling location; LS-SVM, SVM, K-NN classifier is applied to decouple positioning frame, the experiment results validate the effectiveness of decoupling positioning method.
【学位授予单位】：湘潭大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN925.93
，

本文编号：1347261