关键词: 室内定位, 指纹数据库, 射频识别技术（RFID）, 聚类算法, 加权k近邻算法（WkNN）

Abstract: In order to solve the problem of high updating cost of indoor positioning fingerprint database, a RFID（Radio Frequency Identification） indoor positioning algorithm is designed to update the fingerprint database locally by region division. The algorithm divides the fingerprint map into several sub-regions by clustering algorithm, which selects a representative point for each sub-region to represent the fingerprint validity of the sub-region, and performs Weighted k-Nearest Neighbor（WkNN） algorithm or locally updates for sub-region database according to detecting the validity of all representative points. The experimental results show that the proposed algorithm can greatly improve the positioning accuracy and long-term positioning stability under the condition of low cost.

Key words: indoor positioning, fingerprint database, Radio Frequency Identification（RFID）, clustering algorithm, Weighted k-Nearest Neighbor（WkNN）