Research On The Key Technologies For Indoor Positioning System Based On Computer Vision

With the increasing complexity of building architecture and more frequent repetition of special scenes,demands for indoor positioning services boom.In the indoor environment,the Global Navigation Satellite System(GNSS)cannot locate accurately,and other indoor positioning technologies have problems such as high costs or the need for additional equipment.All the problems limits the popularization.One of the effective solutions for the above issue is to use portable monocular mobile terminals for positioning.Focusing on the solution,this thesis carries out a research on the key technology of computer vision-based indoor positioning system.This thesis first establishes a framework of computer vision-based indoor positioning system,and then studies the key technologies involved: the design and detection of special signs,camera position estimation,and progress calibration.In the end,experimental verification and analysis of the system are performed.The main work includes:1.To solve the problem of misjudgment of visual indoor positioning in highly repetitive scenes,an indoor positioning system,combining landmark and database method,is proposed.The system can realize decimeter-level indoor positioning without destructive modification of the scene.2.In the framework of the vision system in this thesis,A design scheme and an algorithm for specific sign generation,detection and recognition are introduced.In order to obtain the coordinate points in the three-dimensional space,the scheme provides up to five known three-dimensional spatial points through a single specific sign.3.In the framework,aiming to satisfy the needs of user camera position positioning,there are an improved two-step calibration positioning method and an advanced spatial resection positioning method.The new method reduces the amount of calculation.Additionally,the spatial resection method is applied to indoor positioning,achieving decimeter-level indoor positioning.4.On the basis of the visual positioning method,this thesis introduces an accuracy correction method based on multiple special marks.It greatly contributes to the accuracy in special scenes.The effects are verified by the following experiments.5.According to the vision system,an experimental environment is established.Thefeasibility of the vision positioning system is verified through actual scene experiments.Additionally,the positioning results and error sources are analyzed.