支援協作擴增實境的霧結構單目定位與建圖系統

定位與建圖技術在近年為自動車、機器人、虛擬實境、擴增實境等應用提供了良好的解決方案，他能夠提供裝置理解真實物理環境並且準確定位於其中，使得自動導航、避開障礙物、虛擬化環境、放置虛擬物件等能夠實現在移動裝置上。ORB-SLAM為視覺SLAM之一項大成研究之一，他為一個完整的SLAM系統，包含前端處理、後端優化及建圖，他的健壯性使他能夠應用在隨意陌生的環境下並且有著良好的性能表現。霧運算為雲端運算擴展到網路邊緣的一種新型網路架構，他將在雲端的功能去中心化，並且將分散的運算能力移到當地來更接近我們的終端裝置，使得資料能夠就近處理並減少終端運算資源消耗及運算時間。本論文主要探討ORB-SLAM技術在協作擴增實境上的應用，協作擴增實境讓使用者能取回環境中被其他人儲存的物件。我們的研究將致力於他的架構以及重新定位在真實環境中被儲存的擴增實境物件的位置。

關鍵字

霧運算；擴增實境；單目相機；即時定位及建圖

並列摘要

Simultaneous Localization and Mapping technique provides a good solution for automatic car, robot, virtual reality, augmented reality in recent years. It helps devices to recognize the physic environment and locate in it precisely. Making automatic navigation, obstacles avoidance, environment virtualization and virtual object displacement can be realized in mobile devices. ORB-SLAM is one of the most complete work of vison SLAM. It contains a complete SLAM system which includes frontend processing, backend optimization, mapping. Its robust design makes it can be implemented in random unknown environment and have good performance. Fog Computing is a new network structure that extends the Cloud Computing to network edge. It decentralizes the computation abilities on the Cloud and separates them to the local. In this thesis, we focus on the discussion of ORB-SLAM implementation in Fog network structure for Cooperative Augmented Reality purpose. Cooperative Augmented Reality allows user to retrieve Augmented Reality objects which created by other users in the environment. We focus on the structure and the ability to relocate the AR Objects to the original place where they are created in the real environment.