基于几何聚类指纹库的约束KNN室内定位模型  

A Constrained KNN Indoor Positioning Model Based on a Geometric Clustering Fingerprinting Technique

作　　者：刘春燕[1] 王坚[1] 

机构地区：[1]中国矿业大学国土环境与灾害监测国家测绘局重点实验室,江苏徐州221116

出　　处：《武汉大学学报（信息科学版）》2014年第11期1287-1292,共6页Geomatics and Information Science of Wuhan University

基　　金：国家863计划资助项目(2013AA12A201);江苏高校优势学科建设工程资助项目;中央高校基本科研业务费专项资金资助项目(2013RC16);新世纪优秀人才支持计划资助项目(NCET-13-1019)~~

摘　　要：针对室内环境基于RSSI定位不稳定问题,提出了以几何信息改进基于指纹库的KNN定位算法。根据室内几何布局建立了聚类指纹库,提出了表征点位几何特性的点散发性强度(geometric strength of sporadic,GSS)概念。利用最邻近样本点的GSS判别移动终端所在参考点RP控制网结构以动态选择KNN关键参数K,构建最佳多边形为约束准则自适应选取后K-1个邻近点,建立了基于几何聚类指纹库的约束加权KNN室内定位模型。结果表明,改进后定位模型可以更好地估计终端位置信息,其中几何聚类指纹库是改善定位准确性的关键,约束KNN能够有效地提高室内定位精度。The common algorithms based on RSSI presently available are unstable in the indoor envi- ronment. Hence, a constrained KNN positioning algorithm with geometrical information via clustering fingerprints is proposed to resolve this issue. Firstly, the geometric clustering fingerprints are built according to the structural layout. Then, the concept of geometric strength of sporadic （GSS） for a sample point＇s geometry characterization is introduced. The value of GSS is used for identifying the RP control network structure in which the mobile terminal is located to dynamically choose the key parameter K for KNN. When the nearest point（NP） is decided, an optimal polygon constraint condition is constructed to choose the latter （K-1） neighbour points. It can be summarized as a constrained KNN indoor localization model based on a geometric clustering fingerprinting technique. The results of series of tests indicate that the new algorithm can more effectively estimate the location of a mobile terminal. Clustered fingerprints plays a key role in improving the position accuracy, thus the impact of this new KNN algorithm should not be overlooked.

关 键 词：室内定位 聚类指纹库 几何特性 RP控制网结构 约束KNN 

分 类 号：TN911.23[电子电信—通信与信息系统]