Wi-Fi Fingerprint Indoor Positioning Method In The Sparse Reference Points

Compared with other positioning technologies,Wi-Fi indoor positioning technology has been widely studied because of its advantages of low cost and good positioning accuracy of existing network facilities.Fingerprinting Positioning is mainstream method in Wi-Fi indoor positioning technology,which uses the pre-acquired signal strength values in the offline phase as feature fingerprints in the positioning scene to predict the user's location.The signal acquisition work needs to consume a lot of time and human labor,how to ensure the accuracy of positioning under the premise of reducing the signal acquisition workload is the field of research hot spots.In this thesis,first consider the spatial correlation of Wi-Fi signal strength based on the construction of the fingerprint database in the off-line stage,and propose a universal Kriging interpolation algorithm based on the Wi-Fi signal channel loss model.In the acquisition of a small number of Wi-Fi data,the full use of a small amount of Wi-Fi signal fingerprint data to re-build high precision location with more reference point location of the Wi-Fi signal location fingerprint database.Then search the fingerprint database for the online search,which leads to the problem of time cost and waste of resources.The relationship between the spatial division of WiFi signal strength and the partition of physical space is analyzed experimentally.According to this correlation,Data for clustering.K-Nearest Neighbor(KNN)fingerprint matching algorithm with access point(AP)selection strategy,and then use the different spatial interpolation algorithm to build the location of the fingerprint database positioning accuracy experimental verification.This thesis use the position fingerprint database established by universal Kriging interpolation algorithm based on the signal propagation model for location.The experimental results show that,the sampling point interval of 5m,this method's average location error is only 0.1 lm higher than that of the raw data collected at the sampling point interval of 1m.And it is also 0.2m lower than that of the Gaussian process regression algorithm and 0.4m lower than the inverse distance weighting method.When the error accumulation probability is 50%,The positioning error with the universal Kriging interpolation algorithm based on Wi-Fi signal loss model is 3.2m,which is 0.4m and 0.7m lower than that of the Gaussian regression algorithm and the inverse distance weighting method respectively,and is 0.9m lower than the raw data collected at the sampling point interval of 5m before interpolation.To ensure a certain positioning accuracy conditions,the method of the reconstruction of the location of the fingerprint database in this thesis,can greatly reduce the offline phase of the data collection.