基于Wi-Fi的KNN-PIT室内自适应指纹定位技术研究

发布时间：2018-02-24 14:41

本文关键词： Wi-Fi指纹定位 KNN PIT 虚拟参考点自适应　出处：《江西师范大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着移动通信和互联网技术的高速发展,基于位置服务(Location Based Services,LBS)的应用需求日趋强烈。由于全球导航卫星系统(Global Navigation Satellite System,GNSS)可以提供连续、高精度的室外位置信息,实现了诸如车辆跟踪、车辆及行人导航等室外位置服务。但是在室内复杂多变的环境中,GNSS因信号减弱或衰竭无法导航定位,因而高精度的室内定位技术成为研究热点。因此,基于短距离无线通信的的室内定位方案应运而生。由于Wi-Fi是无线通信标准,具有传输距离远、信号保真度高、移动性强、组网便捷等特点,并且在大型公共场所等室内环境已经广泛部署,基于Wi-Fi指纹的定位技术成为室内LBS应用中定位技术的首选。但是当前Wi-Fi指纹定位方案仍存在一些问题以待解决,如信号在室内传播中受多径效应、非视距等因素导致的时变性影响定位的可靠性;参考点划分的密集程度决定算法的复杂度,影响定位的实时性;传统的KNN定位算法只能粗略估计定位点的位置范围,不能对定位点范围进一步约束。针对以上问题,本文提出一种改进的KNN—三角形内点(KNN—PIT)室内定位算法。主要工作及创新点如下:(1)根据室内空间结构特征,建立具有类标号的位置指纹库。传统的指纹库仅包含位置和对应的接收信号强度指示(Received Signal Strength Indication,RSSI)向量。在指纹库中增加类标号位置属性,有助于缩减定位的匹配区域,降低算法复杂度。(2)引入虚拟参考点,利用最佳三角形内点(point in triangulation,PIT)原理进一步约束目标点的定位区域,自适应地使用定位算法进行定位。虚拟参考点并不在指纹库中真实存在,它是在KNN算法定位时假定出来,不仅有助于提高定位精度,也有助于降低指纹库容量,降低计算复杂度。(3)综合运用高斯滤波、均值滤波技术,降低离线和在线阶段的信号随机误差带来的定位影响。离线阶段对采集到的大样本Wi-Fi信号数据进行高斯滤波处理,去除误差较大的干扰值。在线阶段采用均值滤波降低信号的单次随机误差影响。最后,通过实验结果表明:改进后的KNN-PIT定位算法与传统KNN定位算法相比可以更好地估计用户的实际位置,降低定位误差,提高定位实时性。
[Abstract]:With the rapid development of mobile communications and Internet technology, the demand for location-based Based services (LBSs) applications is increasing. Because Global Navigation Satellite system (GNSS) can provide continuous and high precision outdoor location information, Outdoor location services such as vehicle tracking, vehicle and pedestrian navigation are realized. However, in the complex and changeable indoor environment, GNSS is unable to locate because of signal weakening or failure, so high-precision indoor positioning technology has become a research hotspot. The indoor positioning scheme based on short range wireless communication emerges as the times require. Because Wi-Fi is the wireless communication standard, it has the characteristics of long transmission distance, high signal fidelity, strong mobility, convenient networking, etc. The localization technology based on Wi-Fi fingerprint has become the first choice in indoor LBS application. However, there are still some problems in the current Wi-Fi fingerprint location scheme to be solved. For example, the reliability of localization is affected by multipath effect and non-line-of-sight effect, the complexity of the algorithm is determined by the density of reference points, and the real time of location is influenced by the density of reference points. The traditional KNN localization algorithm can only roughly estimate the location range of the location point, but can not further constrain the location point range. In this paper, an improved KNN- triangle interior point KNN-PIT-based indoor location algorithm is proposed. The main work and innovation are as follows: 1) according to the characteristics of indoor spatial structure, The traditional fingerprint database contains only the location and the corresponding received signal strength indication received Signal Strength indication RSSI vector. Adding the class label position attribute to the fingerprint database can help to reduce the matching area of the location. In order to reduce the complexity of the algorithm, the virtual reference point is introduced, and the location area of the target point is further constrained by the principle of the optimal triangle interior point in triangulation site, and the location algorithm is used adaptively. The virtual reference point does not exist in the fingerprint database. It is assumed in the localization of KNN algorithm, which not only helps to improve the accuracy of location, but also helps to reduce the capacity of fingerprint database and reduce the computational complexity. To reduce the impact of random errors in off-line and on-line signal positioning. Gao Si filter is used to process the large sample Wi-Fi signal data in off-line phase. In the online stage, the mean filter is used to reduce the single random error of the signal. Finally, The experimental results show that the improved KNN-PIT localization algorithm can better estimate the actual location of the user, reduce the positioning error and improve the real-time location than the traditional KNN location algorithm.
【学位授予单位】：江西师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN92

【相似文献】