Research On Several Problems Of Constructing Indoor Fusion Intelligent Positioning Platform

In the era of mobile Internet,"location" has become an indispensable and important information resource.Outside,Global Navigation Satellite Systems(GNSS)can provide low-cost location services to users around the world.Indoor,GNSS signals are blocked and interfered with,limiting the function of satellite navigation and positioning,and many indoor positioning technologies have emerged.However,the advantages and disadvantages of these positioning technologies coexist.Providing users with continuous,seamless,and high-precision fusion positioning in the complex indoor environment has become the research direction of people’s efforts.The research on indoor multi-source fusion positioning has theoretical and practical value.Based on the national key research and development program project "indoor intelligent fusion positioning technology",this thesis focuses on the existing problems and application requirements in the construction of an indoor fusion intelligent positioning platform,characterized by intelligence and hybrid.Theoretical and practical exploration has been carried out on the definition and identification of indoor scenes,floor identification,positioning in the non-cooperative environment,fusion positioning in multi ways,positioning system design and development,and other aspects.The research results are as follows:(1)Indoor scene recognition is an important function of the fusion intelligent positioning platform,which is used to drive the selection of positioning methods and fusion positioning in different scenes.Aiming at the lack of uniform specification for indoor scene description in the field of indoor positioning,the definition and description of indoor typical scene units are proposed from the perspective of indoor environment and signal conditions.On this basis,a scene recognition technology system including three stages of data collection,processing,and recognition is proposed.The process of scene recognition using a naive Bayes classifier and the data compression process based on the PCA algorithm are described in detail.The spatial topological relationship between scene units is also constructed to restrict the scene recognition system.Based on this theory system,an online scene recognition system is designed and developed.(2)Floor identification can quickly realize the determination of approximate area and narrow the space search range,and is also an important function of indoor fusion intelligent positioning platform in multi-floor complex environments.Aiming at the two cases of known and unknown base station distribution of floor,a KNN quadratic discriminant floor recognition algorithm and a floor recognition method based on DBSCAN clustering are proposed respectively.When the distribution of base stations on the floor is known,the proposed KNN quadratic discriminant floor recognition algorithm can avoid the problem that the current floor identification is susceptible to the deployment density of wireless base stations.In the case of unknown base station distribution of floor,the floor recognition method based on DBSCAN can distinguish the signal characteristics of different areas by using the existing fingerprint database information and realize floor recognition.To solve the problem that floor recognition results are prone to jump in the transition area of floors,an activity detection model based on air pressure is studied to assist floor identification and make floor identification and switching more stable.(3)The indoor fusion intelligent positioning platform needs to take into account high-precision positioning in non-cooperative scenarios.To solve the problem that the traditional range-based positioning is easily affected by the spatial layout of the base station,which leads to the gross error or difficult convergence outside the coverage area of the base station,a positioning method based on maximum posterior position estimation is proposed,which is suitable for location in the indoor non-cooperative space environment.The method establishes the optimal model of range-based positioning from the perspective of probability estimation,introduces the relationship between the spatial layout of the base station and the position of the estimated point into the optimal model in the form of prior information,and improves the positioning accuracy and stability by least squares iteration.(4)The implementation of an indoor fusion intelligent positioning platform needs to include different modes of fusion positioning.Research on absolute positioning fusion based on fingerprint and ranging is conducive to the platform to adapt to more scenarios.A new dynamic weighted fusion positioning method was proposed to solve the problem that the weights of traditional fingerprint positioning and range-based positioning should be determined in advance,determining the weight dynamically in real-time,and the indirect adjustment principle is used for adaptive weighted fusion positioning.In addition,a virtual distance fingerprint location method is proposed,which is based on the distance information and adopts the fingerprint matching method to locate.By integrating virtual distance fingerprint positioning and single point positioning methods,based on the dynamic weighted hybrid positioning model,the fusion positioning using the same set of observations and different methods is realized,and the positioning accuracy and stability are improved.(5)The relative positioning represented by PDR does not require external base station equipment so the fusion mode of absolute positioning and relative positioning runs through the whole indoor fusion intelligent positioning process.Based on the analysis of the causes of errors in the traditional fingerprint and PDR positioning methods,this thesis proposes a robust fusion positioning algorithm based on EKF to achieve the fusion of fingerprint positioning and PDR positioning.This method can realize fine-grained gross error detection and reduce the influence of pedestrian motion state on heading estimation Experimental results show that the method can effectively improve the accuracy,continuity,and stability of the positioning results.(6)Due to the complexity of the indoor environment,different positioning signals and methods can be used in different indoor scenes.Based on the theoretical research of an indoor intelligent positioning strategy based on scene recognition and multi-mode indoor positioning technology,this thesis proposes a cloud-terminal collaborative indoor fusion intelligent positioning scheme.The proposed scheme can adaptively switch positioning methods according to scene changes,realize complex scene change perception,pedestrian motion state detection,and adaptive fusion of multiple positioning methods.Develop and implement cloud services and terminal applications,and build an indoor hybrid intelligent positioning platform.The Winter Olympic venues carry out platform application demonstrations,test platform performance,and verify the technical system.There are 98 figures,35 tables,and 187 references in this dissertation.