Research And Implementation Of Indoor Positioning Technology Based On The Fusion Of Video And Inertial Sensors

Location Based Services(LBS)plays a very important role in fields such as unmanned driving,intelligent transportation,emergency rescue,and smart cities.For outdoor environments,the GPS system in the United States,the Galileo system in Europe,and the Beidou satellite system in China can already provide good location information.As satellite signals are easily blocked by buildings,the positioning performance of these mature positioning systems in indoor environments is not very good.With the development of society,people's demand for indoor positioning is increasing,and positioning solutions have gradually developed from a single positioning technology to a fusion of multiple technologies.There are roughly two main problems with the existing positioning technology solutions.First,the positioning accuracy is high,but a large number of additional hardware devices need to be deployed,and the positioning cost is too high,which is not convenient for practical promotion and application;the second is that widely deployed hardware devices can be used indoors.However,the positioning accuracy is low,and the positioning performance is easily interfered by the environment.The main work and innovation of this thesis are reflected in the following four aspects:Firstly,aiming at the existing problems of the existing indoor positioning technical solutions,this thesis proposes an indoor positioning solution that integrates surveillance video and mobile phone inertial sensor information.In large-scale indoor environments,surveillance cameras have been widely deployed as basic security equipment,and the improvement of computer processing capabilities and the development of deep learning technology have effectively improved the accuracy of video pedestrian detection and trajectory tracking.With the advent of the mobile Internet era,smart phones have become one of the electronic devices that people carry with them on their daily travels,and mobile phone inertial sensor positioning has the advantages of simplicity,efficiency,and user identification.Therefore,the surveillance video fusion mobile phone inertial sensor positioning solution can effectively solve the problem of difficult balance between positioning cost and positioning accuracy.The experimental results show that in the multi-pedestrian experiment scenario,the average positioning accuracy of the fusion system can reach 1.7 meters,which can meet the positioning needs of ordinary users.Secondly,in the process of pedestrian tracking based on surveillance video,when multiple pedestrians walk in parallel,the separation and merging of pedestrian detection bounding boxes will cause trajectory interruption.This thesis proposes an improved Markov model video pedestrian tracking scheme that combines image depth feature and trajectory motion feature.The experimental results show that the video pedestrian tracking scheme can effectively solve the problem of trajectory merging and separation,and effectively reduce the number of trajectory switching.Thirdly,in the matching stage of Pedestrian Dead Reckoning(PDR)trajectory and the video pedestrian trajectory,this thesis proposes a single-step matching based on feature weighting algorithm and multi-step matching based on dynamic time warping combined feature weighting algorithm.The experimental results show that these two matching algorithms can effectively improve the accuracy of trajectory matching.In the multi-pedestrian positioning scene,the average matching accuracy is about 81%.Finally,this thesis designs and implements a surveillance video and mobile phone inertial sensor fusion positioning system,developed a positioning app for Android phones and an online positioning server.The experimental results show that when a user initiates a location request in the location area,the positioning result returned by the server can be received and displayed online.