Please use this identifier to cite or link to this item: http://scholarbank.nus.edu.sg/handle/10635/123731
Title: COMPUTER VISION TECHNIQUES IN AUGMENTED REALITY SYSTEMS
Authors: KOU WEN
Keywords: Dense planar reconstruction, robust factorization, structure from motion, non-convex optimization
Issue Date: 14-Jan-2016
Citation: KOU WEN (2016-01-14). COMPUTER VISION TECHNIQUES IN AUGMENTED REALITY SYSTEMS. ScholarBank@NUS Repository.
Abstract: Augmented reality (AR) is a technique to augment virtual objects such as sound, video and graphics, etc in the real environment captured through sensors like camera and viewed from an AR device such as smart phone, glasses. Nowadays AR has been widely applied in many fields like education, art and entertainment. The development of computer vision techniques especially 3D reconstruction, object tracking are crucial for the development of AR system. In this thesis, the aim is to obtain a dense piecewise planar reconstruction of a static scene from multiple image frames based on a factorization framework. Integrating all the relevant constraints in a global objective function, we are able to effectively leverage on the scene smoothness prior afforded by the dense formulation, as well as imposing the necessary algebraic constraints required by the shape matrix. These constraints also help to robustly decompose the measurement matrix into the underlying low-rank subspace and the sparse outlier part. Numerically, we achieve the constrained factorization and decomposition via modifying a recently proposed proximal alternating robust subspace minimization algorithm. The results show that our algorithm is effective in handling real life sequences, and outperforms other algorithms in recovering motions and dense scene estimate. This novel planar reconstruction technique is especially beneficial for the reconstruction of indoor scenes, since artificial planes almost dominate the entire indoor scene. After we obtain the dense planar reconstruction, a simple inference based on plane geometry can be applied to infer the most likely support planes or obstacles, which is important for AR system and even for indoor navigation of robotic agent.
URI: http://scholarbank.nus.edu.sg/handle/10635/123731
Appears in Collections:Master's Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
Computer Vision Techniques in Augmented Reality.pdf1.69 MBAdobe PDF

OPEN

NoneView/Download

Page view(s)

92
checked on Nov 9, 2018

Download(s)

98
checked on Nov 9, 2018

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.