碩士 / 國立雲林科技大學 / 電機工程系 / 106 / Monocular Visual Odometry (MVO) based on dead reckoning positioning method not only can eliminate the deployment cost of positioning infrastructure of multilateration or proximity matching positioning methods, but also can avoid external electromagnetic interference or internal multipath echoes. MVO simply requires a low-cost and easy-to-install monocular camera to achieve indoor/outdoor positioning functionality for applications of robotics, portable computing, augmented reality, and automobile electronics. In order to improve the issues of positioning estimation drift and vulnerability to environmental variation of conventional Semi-Direct MVO, this thesis proposes Compact ORB (CORB) MVO positioning based on conventional Feature-Based MVO to raise the robustness of depth descriptors of feature extraction and simplify the mapping procedure of conventional ORB Visual SLAM. Experimental results show the positioning estimation of CORB MVO is more accurate, reliable, and instant than that of conventional Feature-Based MVO and Semi-Direct MVO, under various indoor/outdoor spaces. On the other hand, CORB MVO proposed by this thesis has been implemented into Android portable device seamlessly and smoothly to accomplish indoor/outdoor augmented reality navigation. This implementation is suitably applied to versatile augmented reality navigation services, like path directions, event guidance, merchandise seeking, social searching, and so on.
| Identifer | oai:union.ndltd.org:TW/106YUNT0441030 |
| Date | January 2018 |
| Creators | HO, MING-CHE, 何明哲 |
| Contributors | HO, CHIAN-CHENG, 何前程 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 55 |
Page generated in 0.0151 seconds