K-means指纹定位的优化算法

摘要

K-means指纹定位可减少定位算法的计算量,提高定位的实时性已成为当前定位算法的一个研究热点.然而其聚类的随机性却给定位带来极大的不稳定性,对此提出使用两步聚类算法进行优化,根据AIC准则自动得到最优的聚类个数;针对最邻近算法定位误差大的情况,使用相关系数法确定相似度最高的子库,再估计最终位置.实验结果表明,优化后的算法不但改善了定位精度,也极大提高了定位的实时性与稳定性.%K-means fingerprint localization can reduce the complexity of localization,and improving the real-time of location has become a hot-spot of current localization algorithm.However,the randomness of clustering has resulted in great instability to the localization.In this paper,two-step clustering algorithm is proposed to optimize the clustering number according to the AIC criterion.Considering the nearest neighbor algorithm would result in great error,correlation coefficient method is used to determine the highest similarity of the sub-library,and estimate the final position.The experimental results show that the optimized algorithm improves not only the positioning accuracy,but also the real-time and stability of localization.