摘要
室内定位算法是基于位置服务领域研究的难点之一。针对室内定位应用场合和精度问题,提出了利用K-means和KNN融合算法对覆盖率广的WiFi信号进行指纹定位。WiFi指纹定位主要问题是前期指纹库数据的精确以及后期数据的匹配效果。首先,对WiFi信号的概率分布进行研究,弥补了一直以来由于K-means是无监督学习带来的k值的难以选取的缺陷,提高了指纹库的精确性,同时确保数据实时性。后期在线数据处理利用KNN分类算法进行后期在线定位过程的准确性。经多个实验场景测试结果表明,该算法在室内定位精度上3 m定位精度概率保持在78.4%,4 m精度保持在93.6%,基本上保证了室内定位精度的要求。
Indoor positioning algorithm is one of difficulties in location-based service field research. In allusion to the appli- cation scene and precision problems of indoor positioning, the view that using the K-means and KNN tusion algorithm to perform fingerprint positioning for wide-coverage WiFi signals is proposed. The main problems of WiFi fingerprint positioning lie in the data accuracy of the front-stage fingerprint database and the matching effect of late-stage data. The probability distribution of WiFi signals is studied to make up the det）et that it is diffieuh to select the k value as K-means is an unsupervised learning for a long time, which can improve the accuracy of the fingerprint database, and meanwhile ensure the real-fime performanee of data. During the process of late-stage online data processing, the KNN classification algorithm is adopted to obtain the accuracy of the late-stage online positioning. The algorithm was tested in muhiple experimental scenes. The results show that the 3 m posi- tioning accuracy probability of the algorithm remains 78.4% in indoor positioning accuracy, and the 4 m positioning accuracy probability remains 93.6%, which can basically ensure the accuracy requirement of indoor positioning.
作者
丁承君
宇中强
朱志辉
DING Chengjun;YU Zhongqiang;ZttU Zhihui(School of mechanical engineering, Hebei University of Technology, Tianjin 300130, China)
出处
《现代电子技术》
北大核心
2018年第10期39-42,46,共5页
Modern Electronics Technique
基金
天津市科技计划项目(14ZXCXGX00123)
天津市科技计划项目(14ZCDZGX00811)~~
