Bibliographic Details
| Summary: | The evolution of 3D scanning devices and innovation in computer processing power and storage capacity has sparked the revolution of producing big point-cloud datasets. This phenomenon has becoming an integral part of the sophisticated building design process especially in the era of 4th Industrial Revolution. The big point-cloud datasets have caused complexity in handling surface reconstruction and visualization since existing algorithms are not so readily available. In this context, the surface reconstruction intelligent algorithms need to be revolutionized to deal with big point-cloud datasets in tandem with the advancement of hardware processing power and storage capacity. In this study, we propose GPUMLib – deep learning library for self-organizing map (SOM-DLLib) to solve problems involving big point-cloud datasets from 3D scanning devices. The SOM-DLLib consists of multiple layers for reducing and optimizing those big point cloud datasets. The findings show the final objects are successfully reconstructed with optimized neighborhood representation and the performance becomes better as the size of point clouds increases. |
|---|
| Main Authors: | Lee, Wai |
|---|
| Other Authors: | Hasan, Shafaatunnur; Shamsuddin, Siti; Lopes, Noel |
|---|
| Subject: | Surface reconstruction, self-organizing map, deep learning, parallel computing, point clouds |
|---|
| Year: | 2017 |
|---|
| Country: | Portugal |
|---|
| Document type: | article |
|---|
| Access type: | open access |
|---|
| Associated institution: | Universidade Politécnica da Guarda |
|---|
| Language: | English |
|---|
| Origin: | Repositório Científico da Universidade Politécnica da Guarda |
|---|