 |
Dao Thanh |
   |
| Project:
|
| |
| Supervisor:
|
| |
| Host University:
|
| |
| Personal introduction |
top |
| A friendly hello from VietNam. I joined the Vibot program with my background on doing a 3D reconstruction project back in my undergraduate final year in UNSW, Australia and with my excitement for discovering Europe. The journey was not as always nice and easy as we expected, we had hard days (and nights) with this heavy loaded course, but in the end, it was a good time many good memories and lots of things learnt. To know more about my academic and work experience, it's probably best to click on my CV above. I'm looking for a job in Europe so if you want me, here I come. To know more about me personally, you have to get more personal :). A green light, I'm a friendly person so dont hesitate to contact me. Ciao |
|
| |
| Project Description and Objectives |
top |
|
Reconstructing the 3D model of objects from stereo images in underwater condition is a challenging task due to attenuation of light and unknown structures. This master thesis attempts to address this problem by first applying some image enhancing techniques to remove the effect of illumination and reduce the noise. In the next step, the corresponding SIFT features are then detected and matched between the images. From the matched points, epipolar geometry is computed to remove outliers and recover the transformation between cameras. Finally, the surface reconstruction is performed by linear triangulation. The result of the process is a set of 3D points with robust features that can be used for applications like robot mapping or 3D SLAM.
The method was tested on a set of underwater images, with the knowledge about the camera parameters. The result of testing shows that the methods can reconstruct a good 3D approximation of the underwater object with small error and fast performance. |
|
| |
| |
| With the aim of developing a complete system that performs 3D reconstruction for large scale under water scene using efficient and advanced techniques of in the field of computer vision, this master thesis have successfully implemented an algorithm that: given a set of underwater images, it can return a 3D reconstruction of the scene in form of 3D cloud point for surface modeling. It also have achieved the goal of finding a set of 3D points with robust features (SIFT) which can be utilized for other purposes such as mapping or Visual SLAM. |
| 
|
|
| |
| |
| |
| |
|
|
| |
