Inhaltszusammenfassung

3D objects modelling from 2D images is a major task in many application fields including robotics, virtual and augmented reality, and entertainment. According to the equipment used to acquire and process data, a reconstruction process can be based on laser or optical scanners or on digital images. Laser scanner provides a highly detailed and accurate representation of even complex surface shapes but generally, they are expensive, the data processing consumes much time and requires expert inte...3D objects modelling from 2D images is a major task in many application fields including robotics, virtual and augmented reality, and entertainment. According to the equipment used to acquire and process data, a reconstruction process can be based on laser or optical scanners or on digital images. Laser scanner provides a highly detailed and accurate representation of even complex surface shapes but generally, they are expensive, the data processing consumes much time and requires expert interaction with the machine. On the other hand, image based multi-view stereovision techniques provide nowadays an exciting alternative for 3D scene reconstruction because images can be easily obtained. Given a set of 2D images, we propose a novel approach for 3D urban scene reconstruction with low cost by extending unstructured spatial data normally generated by stereovision approaches by linear structures found in the same images. Our suggested system try to resolve two proposed issues: (1) The generation of dense 3D point clouds from 2D images (2) The integration of the available structured 2D data for the improvement of the detection/reconstruction of urban structures.» weiterlesen» einklappen

Autoren

Klassifikation

DDC Sachgruppe:
Informatik