Research On Indoor Positioning Algorithm Based On WiFi And Fingerprint Database Matching

With the rapid development of Internet of things industry and wireless communication,the demand of location-based services in people's daily life is increasing,especially for indoor environment.However,in the indoor environment,the problem of non-line-of-sight and the interference of multipath effects have always restricted the development of indoor positioning technology,resulting in three inherent problems of low positioning accuracy,poor real-time performance and high cost,which have never been solved by a universal solution.In the current numerous indoor positioning technology schemes,WiFi technology,no matter in private or public places,its wireless network coverage is increasing year by year,with high universality and low cost advantages.This paper analyzes and compares domestic and international indoor positioning methods,determines the use of fingerprint matching methods based on WiFi,and proposes fingerprint matching optimization algorithms,which effectively solve the relationship between accuracy,real-time performance and stability.The main research contents of this paper are as follows:(1)Study and analyze the WiFi node layout scheme for indoor environment.In order to ensure that the RSSI sequences of AP nodes collected at the point to be measured have obvious differences,the experimental environment determines to arrange a WiFi node every 7 meters.The RSSI factors that affect AP nodes in the experimental environment are analyzed to determine a single point multiple multi-directional acquisition scheme.Construct an offline fingerprint database,process improved Gaussian filtering on it,and build a high-quality WiFi signal fingerprint database.(2)For the high-quality and high-density fingerprint information in the offline fingerprint database,the matching time will be increased during the real-time matching phase,and the positioning time will be extended.Since the K-means clustering algorithm needs to give the initial clustering center and the number of clusters,this paper uses the evaluation criteria and the Canopy clustering algorithm to determine the number of clusters and the initial clustering center,respectively.Canopy algorithm is used for "rough" clustering first,then K-means algorithm is used for "refined" clustering,and finally k clustering subsets are obtained.The experimental results show that the positioning efficiency is improved by about 95.05%.(3)In the real-time matching stage,first,the subset with the highest similarity is selected by correlation coefficient method to match,so as to improve the positioning efficiency under the condition of maintaining high positioning accuracy.Then in the fingerprint matching stage,combining traditional WKNN algorithm and Bayesian probability algorithm,this paper proposes an improved Bayesian probability optimization algorithm.The final experimental results show that the optimized matching algorithm after the reduced fingerprint database has an average positioning accuracy improvement of about 38.64% and an average running time reduction of about 93.51% compared with the WKNN algorithm without the optimized fingerprint database,which can effectively improve positioning accuracy and real-time performance.(4)Aiming at the user's actual application requirements,this paper designed and developed a positioning system,using Android Studio as the client's front-end development tool,to implement the fingerprint database construction in the offline phase and the positioning display function in the real-time matching phase;at the same time,the local management also uses Android Studio development tool to implement the fingerprint matching algorithm in the real-time matching stage,and finally estimates the position coordinates and returns them to the customer's front-end.Experimental results show that the proposed optimization algorithm and the designed positioning system can achieve the expected results.