室内定位是打通信息覆盖“最后一千米”的关键技术。基于位置指纹的WiFi室内定位技术以其实现简单,成本较低,定位精度高等优点成为研究热点。鉴于现有的指纹定位法中的K近邻定位算法原理简单且定位精度不高,提出了一种基于熵加权的K近邻建筑内定位算法。首先,应用欧氏距离作为相似性度量指标优选近邻点的特征;然后,针对指纹数据库中大量数据样本,避免依赖多种主观赋权不足之处,深度挖掘数据自身内部特征,采用熵权法自适应调整各个匹配采样点的权重占比;最后,通过待测点定位实验进行误差分析。Matlab仿真结果表明:当误差在3 m内时,EKNN算法的定位精度高于WKNN、KNN算法;当K值相同时, EKNN算法最小平均误差低于WKNN、KNN算法。证明本文所提基于熵加权的KNN建筑内定位算法在楼房建筑内定位的稳定性和精确度,为更高精度的室内定位提供了一定的理论支撑和指导参考。

Indoor localization is the key technology to get through the “last mile” of information coverage. The WiFi indoor localization technology that based on position fingerprint is regarded as a research hot spot, because of its simple, low cost and high location accuracy. In view of the fact that the KNN (K-Nearest Neighbor ) localization algorithms in the existing fingerprint location method are simple and low location accuracy. To solve this problem, a KNN localization algorithm in buildings based on entropy weight (EKNN) is proposed in this paper. Firstly, Euclidean distance is used as a similarity measure to optimize the features of the nearest neighbor points; secondly, for a large number of data samples in fingerprint database, avoiding the lack of subjective weighting based on expert experience, based on the internal characteristics of the data, the entropy weight method is used to adjust the weight proportion of each matching sampling point adaptively; finally, The error analysis is carried out through the location experiment of the measurement point. Matlab simulation results show that when the error is within 3 m, the location accuracy of EKNN algorithm is higher than that of WKNN and KNN ; When the K value is the same, the minimum average error of EKNN algorithm is lower than that of WKNN and KNN. The stability and accuracy of the entropy-weighted KNN (Entropy KNN, EKNN) in-building location algorithm proposed in this paper has been proven to provide theoretical support and guidance for higher-precision indoor localization.

TN957. 52

国网重庆电力公司资助科技项目(5220051900NU)