基于有序聚类和MSKPCA的室内定位算法

Indoor positioning algorithm based on orderly cluster andmulti-scale kernel principal component analysis

针对指纹室内定位算法中环境动态变化对Wi-Fi信号的干扰和定位实时性较差的问题,提出一种基于有序聚类和多尺度核主成分分析的Wi-Fi指纹室内定位算法。离线阶段采用参考点可检测接入点序列的最长公共子序列衡量相似度,通过有序聚类划分子区域。在线阶段先进行粗定位,选择最优尺度的核主成分分析模型处理子区域指纹数据,使用朴素贝叶斯加权K近邻算法预测目标节点位置。实验结果表明,该算法可有效提升定位精度,86.7%的定位误差在1.2 m以内。

To solve the problem that the interference of Wi-Fi signal caused by dynamic changes of environment in fingerprint indoor localization algorithm and poor real-time of localization,a Wi-Fi fingerprint indoor positioning algorithm based on orderly cluster and multi-scale kernel principal component analysis was presented.On the off-line stage,the similarity of the reference point was measured by analyzing the longest common subsequence of access point sequence in which reference point could be detected.The area was divided into several parts through an ordered cluster.On the on-line positioning stage,coarse location was applied to determine the subregion.The kernel principal component analysis model with the optimal scale was selected to process the fingerprint data of the subregion.The target node location was predicted using simple Bayesian weighted K nearest neighbor algorithm.Experimental results show that the proposed methods can effectively improve positioning accuracy and the positioning error within 1.2 m is 86.7%.