Research On Subarea Clustering Indoor Location Algorithm Based On Improved Affinity Propagation Clustering Algorithm

With the development of information technology and pervasive computing,there is a growing demand for location-based applications and services,such as accurate positioning in indoor environments or real-time tracking of people or objects in buildings.Positioning can be divided into outdoor positioning,indoor positioning and hybrid positioning according to the positioning target environment.Global Navigation Satellite System(GNSS)is one of the most successful positioning systems in outdoor environment.However,satellite signals are easily blocked in closed structures,which makes it difficult for GNSS to carry out accurate indoor positioning estimation and indoor navigation.In recent years,scholars at home and abroad have developed many different indoor positioning schemes according to the different application requirements of end-users.The existing indoor positioning schemes vary in positioning accuracy,coverage,location update frequency and installation and maintenance.In the existing indoor positioning schemes,wlan-based indoor positioning technology has been widely used in people's production and life.In this thesis,based on the analysis and introduction of the existing indoor location technology,the location fingerprint location technology based on wlanis studied.Based on the analysis of the distribution characteristics of RSS signals,a sub-region partition algorithm and an improved ap nearest neighbor propagation algorithm are proposed.The main research and innovation contents of this paper are as follows:(1)The existing indoor positioning system and positioning technology are introduced.From the development history,research status,positioning accuracy,positioning cost and deployment difficulty,the existing positioning system and positioning technology are compared and analyzed.This paper introduces the indoor location technology and location system based on wlan,and analyzes the defects of WLAN indoor location technology.Finally,the location fingerprint location technology based on wlan is studied deeply.(2)The RSS signal is susceptible to the complex indoor environment,and the RSS signal value at the same location is random.The results of data preprocessing using mean filtering,median filtering and Gaussian filter algorithm are compared and analyzed.Gaussian filter algorithm is chosen to pre-process the rssi signal value in offline phase.(3)In order to solve the problem of large amount of data searching and low searching efficiency in the online positioning phase,the idea of dividing the sub-regions of the locating region in the offline positioning phase is put forward.The traditional meshing method does not take into account the regional distribution characteristics of RSS signals when dividing sub-regions.In view of this problem,this paper proposes a sub-region division method based on the rateof change of RSS signals to divide the location regions roughly.The rough-divided reference points are used as training sets and the svm algorithm is used for learning.The isolated points which are not divided into any regions are divided in detail by using the learned model.(4)In order to solve the problem that traditional clustering algorithms,such as k-means algorithm,need to allocate clustering centers artificially and determine the number of clusters,an improved Affinity Propagation algorithm is proposed.Traditional Affinity Propagation algorithm uses Euclidean distance to construct similarity matrix.However,Euclidean distance has the same weight in calculating the distance of reference points in signal space,and can not fully reflect the distribution characteristics of each AP component in different sub-regions.In order to solve this problem,the parameterized similarity matrix is calculated,and the weights are assigned according to the contribution degree of each AP in different sub-regions.Considering the effect of data point density on its becoming a cluster center,the setting of preference value in similarity matrix s(i,j)is improved by using the data density around reference points in each sub-region.In the offline phase,the improved Affinity Propagation algorithm is used for adaptive clustering in each sub-region.