A Contribution to Semi-Automatic Segmentation of Point Clouds
The increase the emerging technologies and the low cost related with 3D data acquisition have allowed researchers to open new doors of research in domains like building information modeling (BIM), in-process inspection, virtual simulation, reverse engineering, among others. However, these new doors...
- Autores:
-
Tamayo Quintero, Juan David
- Tipo de recurso:
- Fecha de publicación:
- 2015
- Institución:
- Universidad Nacional de Colombia
- Repositorio:
- Universidad Nacional de Colombia
- Idioma:
- spa
- OAI Identifier:
- oai:repositorio.unal.edu.co:unal/61999
- Acceso en línea:
- https://repositorio.unal.edu.co/handle/unal/61999
http://bdigital.unal.edu.co/60820/
- Palabra clave:
- 0 Generalidades / Computer science, information and general works
62 Ingeniería y operaciones afines / Engineering
Nubes de puntos
Imágenes 3D
Modelos dentales 3D
Procesamiento de imágenes
Entornos Urbanos
Segmentacion 3D
Point cloud
3D images
3D dental models
Image processing
urban environments, 3D segmentation
- Rights
- openAccess
- License
- Atribución-NoComercial 4.0 Internacional
Summary: | The increase the emerging technologies and the low cost related with 3D data acquisition have allowed researchers to open new doors of research in domains like building information modeling (BIM), in-process inspection, virtual simulation, reverse engineering, among others. However, these new doors carry out with them some challenges regarding data transmission, processing, storage, all of them currently being active research topics. This thesis aims to advance point cloud processing– specifically in segmentation– in three directions: First, we propose a hybrid technique for carrying out a semi-automatic segmentation using NARF and Min-Cut. Our technique can be modified to be used in different places or fields, for instance: in this thesis the technique was applied on urban and indoor environments, and dental models (medical field). Second, we conducted several tests to the hybrid technique and propose a method ology for the segmentation of dental models. In order to establish the methodology we used an exploratory study in segmentation of dental models, where it was tested by different algorithms (region growing, RANSAC, Min-Cut and the hybrid technique). Third, we developed another methodology based in the hybrid technique for segmenting objects in 3D scenes, aimed towards outdoor and indoor environments. A quantitative evaluation was carried out on a point cloud consisting of about 30 million points, with diverse objects of interest such as trees, cars, chairs, buildings |
---|