Research On Indoor Localization And Related Applications Based On 3D Information

With the popularization of mobile devices and the widely-equipped rich sensors,indoor localization based on mobile devices occupies an important position in indoor navigation,augmented reality and other location-based services.It involves a large potential market.However,there are many restrictive factors for the existing solutions in terms of localization accuracy,deployment costs and resource consumption,which severely limits their promotion in various applications.The indoor localization methods face three key challenges,i.e.,how to effectively sense the surroundings with the mobile sensors and improve the localization accuracy;how to take advantage of environmental characteristics to reduce deployment costs;how to design the localization algorithms and the related data structure to simplify the users' operations and reduce the resource overhead;This thesis focuses on the representation,collection and storage strategy of 3D information and the methods to sense 3D information for accurate orientation estimation and indoor localization by taking advantage of mobile devices.The contributions are summarized as follows:1)Based on the OpenStreetMap,this thesis proposes a compatible template-based representation method for 3D information,which achieves the 3D expansion and solves the problems for 3D information,including large storage overhead,lack of semantic in-formation and the challenges in detailed information collection;In the aspect of 3D information collection,a web-based information collection system is designed for the editing,updating and storage of 3D information;Finally,based on the locality and re-peatability of the distribution for 3D information,the method to multiplex the template and the existing data is designed to effectively reduce the overhead of updating 3D in-formation.2)In view of the complex electromagnetic environment,the poor accuracy of sen-sors and the high computational complexity in image processing for indoor orientation estimation,this thesis proposes a method which fuses images and inertial sensors for accurate orientation estimation.Based on the orientation from vanishing points detec-tion,the inaccurate orientation from sensors is corrected.In the contrast,the data from sensors are used as constraints to assist image processing.Finally,the new method suc-ceeds in improving the accuracy,reducing the resource overhead and enabling real-time applications on mobile devices.3)Aiming at the problems of high deployment cost,low localization accuracy and large resource overhead for indoor localization systems,this thesis chooses inherent 3D information as reference points and designs an indoor localization method based on the constraints of 3D information;Through image processing,the method effectively identifies 3D objects,classifies objects and calculates potential locations based on the relevant features.Finally,accurate indoor localization is achieved based on 3D con-straints.The resource overhead for data processing is reduced through multi-modal information fusion.4)For indoor localization systems in the relevant applications with high compu-tational complexity,poor user friendliness and low real-time performance,this thesis combines the existing sensor-based localization and designs the mechanism to fuse sensor-based localization and 3D information based localization,which simplifies the computation and the resource overhead in continuous localization for indoor navigation.A feedback and adjustment strategy is proposed to guarantee the accuracy.Furthermore,by fusing the sensor data,this thesis optimizes the above algorithms on mobile devices and realizes real time localization based on the streaming video.An augmented reality application is designed based on the optimized system.In summary,by studying the above issues for indoor localization,leveraging the rich sensors of mobile devices,this thesis proposes the solution for the whole system of indoor localization based on 3D information.The main contributions include the strat-egy for representation,collection and storage of 3D information,the methods for ori-entation estimation and indoor localization based on 3D information,the optimization for the indoor localization system and related applications through multi-modal sensor fusion.As a result,we improve the accuracy,reduce the deployment cost,optimize the resource overhead,simplify the users' operation for the proposed indoor localization system and finally apply it in indoor navigation and augmented reality.