基于数据挖掘分类聚类理论的指纹法室内定位优化

发布时间：2018-01-25 20:43

本文关键词： 室内定位数据挖掘指纹法 WKNN K均值聚类嵌入式系统　出处：《北京交通大学》2014年硕士论文　论文类型：学位论文

【摘要】：位置信息是重要的信息资源,位置信息的应用关系到国防、医药、工业和我们生活的方方面面。定位系统的研究曾经是国防科技的重要技术研究方向,自从GPS等卫星定位系统普及,定位芯片逐渐植入各式各样的电子产品,其中包括人们目前大量使用智能手机,这样,定位功能开始走入寻常百姓的生活之中。智能手持设备的普及催生了新的生活需求,基于位置信息的服务(Location Based Services, LBS)应运而生,LBS逐渐成为学术和工业界的研究热点。卫星定位系统使得室外定位技术已经非常成熟,但是微波信号不能穿透墙壁进入室内,不能满足更大的室内定位需求。本文讨论的主要问题是定位技术的这一个新的方向——室内定位技术。本文对室内环境下的无线局域网(Wireless Local Area Network, WLAN)信号传播特点进行了调研,对采集到的信号强度进行统计,将信号强度的分布特点和时变特性进行了简单总结。并且详细阐述了基于WLAN环境的接收信号强度(Received Signal Strength, RSS)的指纹法定位理论,包括定位系统的主要逻辑模型和主要的定位算法。将室内定位系统与信息检索系统进行类比,同样可以分为在线阶段和离线阶段。同时引入了数据挖掘方面的一些观点,根据数据挖掘应用于知识发现的功能特点,将数据挖掘理论体系融入指纹法室内定位研究当中,指导定位系统改进。本文重点研究了室内定位系统在线阶段以加权的K邻近(Weighed K-Nearest Neighbors, WKNN)算法作为定位算法的定位性能,根据训练数据的统计分析给出定位算法参数的选定数值,并且分析了各个参数对定位性能的影响情况。在离线阶段优化了指纹数据的组织形式,将指纹数据进行聚类管理,以求减小在线定位时查找信息的计算量,通过改进后的适用于锚点指纹数据结构的K均值聚类算法对离线指纹库聚类分析,得到聚类结果。文章给出了特定环境下的最佳聚类参数选定办法,分析了室内定位中和离线指纹库聚类相关的性能分析公式,以求更加深刻地认识室内定位问题。文章最后描述了室内定位系统的开发成果,本系统是搭载在嵌入式平台的手持硬件系统,利用USB无线网卡采集数据,定位算法通过纯软件实现,定位效果和理论值比较接近,从而佐证了指纹法室内定位系统在工程上的可实现性。
[Abstract]:Location information is an important information resource, the application of location information is related to national defense, medicine, industry and all aspects of our lives. Since the popularity of satellite positioning systems such as GPS, positioning chips have been implanted in a wide variety of electronic products, including the current mass use of smartphones. Positioning function began to enter the lives of ordinary people. The popularity of smart handheld devices gave birth to new needs of life. Location Based Services (LBSs) emerge as the times require. LBS has gradually become a research hotspot in academia and industry. The outdoor positioning technology has been very mature due to the satellite positioning system, but microwave signal can not penetrate the wall into the room. The main problem discussed in this paper is this new orientation of localization technology-indoor positioning technology. In this paper, the wireless local area network (WLAN) in indoor environment (. Wireless Local Area Network. The characteristics of WLAN) signal propagation are investigated and the signal intensity collected is analyzed. The distribution and time-varying characteristics of signal intensity are briefly summarized, and the received signal intensity based on WLAN environment is described in detail. Received Signal Strength. RSS-based fingerprint location theory, including the main logical model of the location system and the main location algorithm. The indoor positioning system and information retrieval system are compared. It can also be divided into online stage and off-line stage. At the same time, some viewpoints of data mining are introduced, according to the functional characteristics of data mining application in knowledge discovery. The theoretical system of data mining is integrated into the research of fingerprint indoor location. This paper focuses on the study of weighted K-nearest Neighbors in the on-line phase of the indoor positioning system. The WKNN) algorithm is used as the location performance of the localization algorithm. According to the statistical analysis of the training data, the selected values of the location algorithm parameters are given. At the off-line stage, the organizational form of fingerprint data is optimized, and the fingerprint data is clustered and managed in order to reduce the computation of searching information in online location. Through the improved K-means clustering algorithm suitable for anchor fingerprint data structure, the clustering results of off-line fingerprint database are obtained. The method of selecting the best clustering parameters in a specific environment is given in this paper. The performance analysis formulas related to off-line fingerprint database clustering in indoor positioning are analyzed in order to better understand the indoor positioning problem. Finally, the paper describes the development results of indoor positioning system. This system is a handheld hardware system based on embedded platform, using USB wireless network card to collect data, the localization algorithm is realized by pure software, and the positioning effect is close to the theoretical value. Therefore, it proves the engineering realizability of the fingerprint indoor positioning system.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP311.13;TN925.93

【共引文献】