稀疏参考点下的室内定位方法

随着移动互联网的发展，人们对于室内的位置服务需求日益增加。基于Wi-Fi的指纹库室内定位算法具有成本低、定位误差小的优点，但指纹库信号采集需要消耗大量的时间和人力，本文对稀疏参考点下构建高效指纹数据库和高精度室内定位的方法进行了深入研究。本文改进了卡尔曼滤波有效解决了Wi-Fi的噪声和缺失点，设计了基于信号强度差分方差的无线接入点筛选策略来滤除信息量较低的接入点，提出了一种基于支持向量回归拟合的克里金插值算法（Kriging Interpolation Algorithm Based On Support Vector Regression， SVR-Kriging）进行指纹库的构建，最后通过接入点加权的K加权近邻法（AP weighted and Weighted K-Nearest Neighbor， AWKNN）完成定位。将该方法应用于实际的二维、三维定位场景，实验结果表明二维场景平均定位误差为1.01 m，三维场景平均定位误差为0.92 m。该方法解决了指纹数据库信号采集困难、接入点数据冗余的问题，有效地降低了定位误差。 

Indoor Positioning Method with Sparse Reference Points

? ?With the development of mobile internet， people's demand for indoor location service is increasing day by day. The indoor fingerprint database localization algorithm based on Wi-Fi has the advantages of low cost and good positioning accuracy， but the signal acquisition of fingerprint database takes a lot of time and manpower. In this paper， the methods of constructing efficient fingerprint database and high-precision indoor location under sparse reference points are studied deeply. This paper improves the Kalman filter to suppress the Wi-Fi signal noise and the missing points， designes the wireless access point selection strategy based on signal strength difference variance to filter out low information access point， and proposes kriging interpolation algorithm based on support vector regression （SVR-Kriging） to reconstruct the fingerprint database， finally uses the AP weighted and Weighted K-Nearest Neighbor （AWKNN） to get location. The method is applied to two-dimensional and three-dimensional location scenes. The experimental results show that the two-dimensional positioning average error is 1.01 m and the three-dimensional positioning average error is 0.92 m. This method solves the problem of signal acquisition difficulty and AP redundancy， and effectively reduces the positioning error.