Point Cloud Registration and Procedures for Optical Acquisition | Tim Felle Olsen
| | Abstract | Digital 3D scanning of general objects often include manual processes. No scan can be done in one sweep since the object will always be held or rest in a way, which obscures parts of the object. This project seeks to reduce the manual par by automating a software based alignment of partially overlapping scans of an object.
The goal of the project is to describe and implement a global alignment algorithm, ultimately get the algorithm implemented as part of the DTU 3D scanning software.
The proposed algorithm is developed by Zhou et al. 2016 and boasts global efficient 3D registration without the need of a local initialisation, like the classical Iterative Closest Point algorithm do. The full mathematical background of the algorithm is explained in detail. Additionally thorough testing of the algorithm have been performed to find eventual limitations, which is an important aspect of the project. Tests are performed both on a small model of the Stanford Bunny with some modifications and large scans obtained at DTU Imaging.
The implementation performs well on small datasets where parameters can be tweaked easily, however it is not robust enough to handle real scanned datasets. The main faults of the implementation lie within the Fast Point Feature Histogram estimation and correspondence matching. The performance of the implementation can be optimised further.
The implementation by the original authors performed well in all tests and computation time were quite efficient, so with further development it is possible to integrate the method in a 3D Scanner pipeline. | | Type | Master's thesis [Academic thesis] | | Year | 2019 | | Publisher | Technical University of Denmark, Department of Applied Mathematics and Computer Science | | Address | Richard Petersens Plads, Building 324, DK-2800 Kgs. Lyngby, Denmark, compute@compute.dtu.dk | | Series | DTU Compute M.Sc.-2019 | | Note | | | Electronic version(s) | [pdf] | | Publication link | http://www.compute.dtu.dk/English.aspx | | BibTeX data | [bibtex] | | IMM Group(s) | Image Analysis & Computer Graphics |
|