Research Of BLE/PDR Indoor Fusion Positioning Based On Extended Kalman Filter

The rise of large and complex buildings has driven the growth of indoor positioning demand,and location services have gradually become well-known to the public.Outdoor positioning technology has matured,and research hotspots are increasingly turning to indoor and outdoor seamless positioning,and indoor positioning technology is the top priority.High-precision and high-stability indoor positioning systems are currently the hotspots and difficulties of research.Various positioning technologies have their own advantages and disadvantages.The RF signal accuracy is high,but it is limited by the dynamic changes of the indoor environment,while the inertial positioning is limited to the initial position and accumulated error.Therefore,the multi-source information fusion positioning method is the general trend.For this reason,this article mainly conducts detailed research on the fusion positioning method of Bluetooth positioning and PDR positioning technology.The main research contents are as follows:(1)Discuss the time and space propagation characteristics of Bluetooth signals,analyze the multi-peak distribution of Bluetooth signals,use principles and mean filtering for preprocessing,and construct a wireless signal attenuation model under occlusion and no occlusion.In terms of fingerprint positioning,weight Euclidean distance,study the influence of K value and outliers on positioning,and propose an adaptive fingerprint positioning algorithm based on DBSCAN-Meanshift.Compared with the WKNN algorithm,the positioning accuracy is increased by 28.67%,and the positioning stability is greatly improved.During the ranging and positioning process,the influence of the ranging error on the positioning is analyzed,and a distance correction algorithm is proposed to improve the positioning performance through continuous iteration,The positioning accuracy is increased by 0.75 m,and the positioning is more stable;through the comparison of fingerprint positioning and ranging positioning,follow-up experiments are carried out by using fingerprint positioning method.Aiming at the time-consuming problem of fingerprint database construction and update,a rapid construction and update method of fingerprint database based on inertial information is proposed.On the premise of ensuring positioning accuracy and stability,the efficiency of database construction is greatly improved.(2)In view of the three-dimensional positioning problem,verify the feasibility of Wi-Fi signal for floor area recognition,and assign floor area labels to the twodimensional positioning.The accuracy of the machine learning algorithm is verified,and the accuracy rate reaches 97.2%,and the matching efficiency is increased by 70%.Aiming at the time-consuming problem of fingerprint positioning,the Kmeans algorithm is used to divide the coarse-grained and fine-grained signal maps,and the machine learning algorithm is used to verify the accuracy of the division,and to determine the strategy of regional optimization.The division accuracy of the fingerprint database exceeds 97%,which meets the high-precision requirements while reducing the dimension of fingerprint matching.(3)Analyze the characteristics of the acceleration sequence and angular velocity sequence of pedestrians in different motion states,extract feature vectors through the window,and realize different motion state recognition based on the optimal SVM algorithm,with an accuracy rate of 95.15%;through motion state recognition,determine the corresponding Threshold,an adaptive threshold peak detection algorithm is proposed to complete the gait detection in different motion states,with an accuracy rate of 99.62%,which can adapt to the movement speed of the person;for the problem of the deviation of the heading estimation and the poor stability in actual sports,the design is designed The heading estimation method of Sage-Husa adaptive filtering based on UD filter can adapt to the dynamic changes of the system well,and the accuracy of heading estimation is also improved.(4)Aiming at the problem that traditional filters only perform sub-optimal estimation in the process of fusion positioning,this paper proposes an optimized positioning based on EKF based on the EKF algorithm positioning.The fingerprint cost item and the trajectory cost item are constructed by graph optimization theory,and then the EKF positioning result is globally optimized by the PSO algorithm.The improved fusion positioning algorithm can effectively solve the problems of uneven Bluetooth positioning and "rebound" phenomenon,while avoiding the problem of trajectory drift.Compared with the EKF algorithm,the positioning accuracy is improved by 0.59 m,the positioning performance is improved by 36.65%,the position estimation error is within2.5m,and the probability that the positioning accuracy is better than 2m is 89.06%,and the stability is better than Bluetooth positioning technology and PDR positioning technology.This thesis has 100 figures,17 tables,and 92 references.