基于位置指纹识别的WiFi室内定位算法研究与实现

发布时间：2019-06-22 09:43

【摘要】：传统的GPS等定位技术在室外已经能够实现米级的精确定位,但在室内环境下难以精确定位。随着移动智能终端的普及和室内广泛覆盖的WiFi为低成本、高精度的室内定位技术提供了可能。其中基于位置指纹识别的WiFi室内定位以其实现简单、成本低、定位精度高等优势成为室内定位技术的研究热点。本文深入的研究了基于位置指纹识别的WiFi室内定位算法,为了进一步提高该算法的定位精度和定位速度,从离线阶段和在线阶段提出算法的改进。离线阶段,随着室内定位环境的增大,采集的指纹点也相应增多,在线阶段的匹配计算量就会大大增加。考虑到定位的实时性,将数据挖掘中的聚类算法应用到离线阶段数据库的处理。本文提出对二分Kmeans聚类算法进行改进,改进算法将聚类相似度定义为信号强度与坐标两者欧氏距离的乘积。通过该算法聚类后,有效改善了聚类后指纹对应坐标个别离散的情况,改善了聚类的效果,提高了定位精度。在线阶段,本文提出了基于加权欧氏距离的自适应K值的WKNN算法。该算法对检测到待定位点的AP的信号强度加权,然后计算待定位点的指纹与数据库中指纹的加权欧氏距离,筛选出K个邻近点后去除其中的离散点,对剩下的点进行加权求平均。本文还针对聚类边界定位点提出改进算法,有效改善了聚类边界定位点的定位精度。实验结果表明,通过这两个阶段对算法的改进,有效的改善了定位精度,减小了定位时的计算量。改进后算法平均定位误差为1.24米,相比于传统的WKNN,定位误差减小了0.47米,可获得27.2%以上的定位误差的改善。匹配指纹数据量减小了(Q-1)/Q,定位时间相应的缩短(Q-1)/Q,Q是聚类的数量。最后本文将提出的算法应用到车库导航系统中,其中定位系统模块采用C/S架构,包括客户端,服务器,数据库三个模块的设计与实现。本文的研究可以为基于位置指纹识别的WiFi定位算法的进一步研究提供理论支持,同时也为各大定位及导航系统中确定当前位置提供相应的算法支持。
[Abstract]:Traditional GPS and other positioning technologies have been able to achieve accurate positioning of meters outside, but it is difficult to locate accurately in indoor environment. With the popularity of mobile intelligent terminals and the extensive indoor coverage of WiFi, it is possible for low cost and high precision indoor positioning technology. Among them, WiFi indoor location based on position fingerprint recognition has become the research focus of indoor positioning technology because of its simple implementation, low cost and high positioning accuracy. In this paper, the WiFi indoor location algorithm based on position fingerprint recognition is deeply studied. in order to further improve the positioning accuracy and speed of the algorithm, the improvement of the algorithm is proposed from the offline stage and the online stage. In the off-line stage, with the increase of indoor positioning environment, the number of fingerprint points collected increases correspondingly, and the amount of matching computation in the online stage will be greatly increased. Considering the real-time performance of location, the clustering algorithm in data mining is applied to the processing of offline database. In this paper, an improved binary Kmeans clustering algorithm is proposed. The clustering similarity is defined as the product of signal strength and Euclidean distance between coordinates. After clustering, the clustering algorithm effectively improves the individual discretization of the corresponding coordinates of fingerprint after clustering, improves the effect of clustering, and improves the positioning accuracy. In the online phase, an adaptive K-value WKNN algorithm based on weighted Euclidean distance is proposed. In this algorithm, the signal strength of AP detected is weighted, and then the weighted Euclidean distance between the fingerprint of the unlocated point and the fingerprint in the database is calculated. The K adjacent points are selected and the discrete points are removed, and the remaining points are weighted and averaged. This paper also proposes an improved algorithm for clustering boundary location, which effectively improves the positioning accuracy of clustering boundary location. The experimental results show that through the improvement of the algorithm in these two stages, the positioning accuracy is effectively improved and the computational complexity is reduced. The average positioning error of the improved algorithm is 1.24 meters, which is reduced by 0.47 meters compared with the traditional WKNN, positioning error, and more than 27.2% of the positioning error can be improved. The matching fingerprint data volume decreases (Q 鈮，

本文编号：2504465