摘要
针对目前室内定位依靠Wi-Fi电磁指纹库方法实现室内人员定位进行判别存在误差大以及时效性低的问题,本文提出一种融合K近邻(K-NN)的改进密度峰值聚类(K-IDPC)算法。引入关联系数和K-NN思想,解决了普通密度峰值聚类(DPC)算法对定位数据密度不均衡,聚类中心区分度不高的问题,进而提高了对定位环境的鲁棒性。并结合数据切分算法,对离线电磁数据进行切割,使得大数据集分为若干小数据集,降低了计算复杂度。实验结果表明:提出的室内定位方法,同传统的K均值(K-means)、具有噪声应用的基于密度空间聚类(DBSCAN)、DPC聚类算法相比,能够有效地提高室内定位的效果。
Aiming at the problem of large error and low timeliness occur while current indoor positioning,relying on Wi-Fi electromagnetic fingerprint library method to realize indoor personnel positioning is proposed.An improved density peak clustering (K-IDPC) algorithm fuses K-NN is proposed.The correlation coefficient and K-NN idea are introduced to solve the problem that the ordinary density peak clustering (DPC) algorithm has unbalanced positioning data density and the distinction degree of clustering center is not high,which improves the robustness to the positioning environment.Combined with the data segmentation algorithm,the offline electromagnetic data is cut,so that the big data set is divided into several small data sets,which reduces the computational complexity.The experimental results show that the proposed indoor positioning method can effectively improve the indoor positioning effect compared with the traditional K-means,DBSCAN and DPC clustering algorithm.
作者
何洋
吴飞
贺成成
朱海
毛万葵
HE Yang;WU Fei;HE Chengcheng;ZHU Hai;MAO Wankui(School of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China;Shanghai Huace Navigation Technology Ltd,Shanghai 201702,China)
出处
《传感器与微系统》
CSCD
2019年第11期46-49,53,共5页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61272097)
上海市科技学术委员会重点项目(18511101600)
关键词
Wi-Fi定位
密度峰值聚类
关联系数
K近邻
数据切割
Wi-Fi positioning
density peak clustering(DPC)
correlation coefficient
K-nearest neighbor (K-NN)
data cutting