碩士 / 國立臺灣科技大學 / 機械工程系 / 106 / The three-dimensional (3D) reconstruction of real objects is a very important research sector required for inspection, navigation, medical and reverse engineering, security, object identifications, visualizations, and animations. It can be achieved via different methods through different sensors. In fact, choosing the optimal reconstruction method and sensor is not an easy task. It requires a lot of experience and trials and errors. In this work, three applications for 3D reconstruction have been developed. Each of them has been implemented with a different technique, sensor, and theory.
The first application recognizes and tracks a pen’s marker using stereo vision theory via two calibrated 2D cameras. The center of the marker position has been precisely detected and followed. Algorithm’s limitations and errors' models have been determined. This application could be utilized in the future through attaching these 2D cameras to a robot to follow human hand trajectory for remote processes.
The second application uses the KINECT® sensor to perform Simultaneous Localization and Mapping (SLAM) navigation. Graph-based SLAM has been developed in an indoor scenario. By moving a hand-held camera around the environment, features have been detected. Then, matching these features and estimating their 3D positions. After that, align the generated 3D clouds and close the loop to construct the 3D map for this environment. This experiment could be applied by a navigation robot to scan various indoor environments including dangerous ones.
The third application reconstructs the 3D model for a server and renders its 2D images in respect to selected virtual camera positions. Then, moves a real camera to these selected positions and captures real images. Comparing the real and virtual images, system accuracy has been assessed. This idea was automated to perform an Automatic Optical Inspection (AOI) for a real industry case. The main target is to inspect the assembling, labeling and component placement on a manufactured server. Developing such an autonomous inspection system makes the inspection process faster, easier and cost-efficient.
These studies open the door for various robotic vision applications in industrial fields. They also promote the functions of the sensors and reconstruction methods for specific applications.
| Identifer | oai:union.ndltd.org:TW/106NTUS5489170 |
| Date | January 2018 |
| Creators | ALAA MOHAMED MAHMOUD AHMED, ALAA MOHAMED MAHMOUD AHMED |
| Contributors | Chyi-Yeu Lin, 林其禹 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | en_US |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 99 |
Page generated in 0.0016 seconds