Alinhamento de imagens de profundidade na reconstrução 3D de objetos de forma livre

AUTOR(ES)

Landecir Alves de Albuquerque

DATA DE PUBLICAÇÃO

2006

RESUMO

The 3D reconstruction of digital models from range images acquired by non-contact digitizers is currently the state of the art for the construction of models. The approach consists of three stages. In the first one the real world is sampled in the range images, which are nothing more than maps with information on measured distances from a viewer to points in the scene. The great advantage of these images is that the 3D information is explicit, saving the work of retrieving it, which represents a serious obstacle for many applications and is inevitable when using intensity images. The subsequent step is the alignment or register of images, which is due to the fact that the scanner is limited only to acquire portions of surface at a time, as a result of the unidirectional nature of the sweeps. It leads to need to acquire multiple range images varying poses between the scanner and the object. The images are acquired on the sensor system of coordinates and the process of alignment is to bring them together within a global or common system of coordinates. The last step is the integration of views and reconstruction of the surface of the virtual object. In this dissertation is presented the current state of technology all stages of 3D reconstruction. The research has focused on the second stage of reconstruction, since this is an issue of extreme importance to the quality of the final model, because a misaligned creates distortions in the model that disapprove it for many applications that require precision, such as those present in engineering, where is the common need to build threedimensional models with sufficient accuracy to be used in manufacturing systems or numerical simulation performance of machines in operation and components such as turbines and flows in non-circular ducts. In this sense is also presented an implementation of ICP algorithm for alignment of multiple range images acquired with laser scanning. The methodology used was subject the algorithm to tests with real range images and synthetic data contaminated with varying percentages of noise.

ASSUNTO(S)

sistemas de indicação visual tridimensional sistemas de varredura engenharia mecanica

Documentos Relacionados

• Reconstrução 3D a partir de imagens de profundidade baseada em modelos deformaveis
• Reconstrução 3D a partir de seções transversais de objetos
• Reconstrução 3D de imagens tomográficas de raios-x de arcos coronais em ambiente paralelo.
• DESENVOLVIMENTO DA ARTRODESE DO TORNOZELO BASEADA NA TOMOGRAFIA COM RECONSTRUÇÃO 3D
• 3D image registration of the human brain