ETH Zurich : Computer Science : Pervasive Computing : Distributed Systems : Education : Student Projects : Abstract

Rapid Object Reconstruction for Product Augmented Reality (B)

Status: Abgeschlossen

Obtaining 3D models of objects in general requires a lot of work; usually the models are either generated by hand using professional drawing tools, by using expensive 3D-scanners, or by using complex offline reconstruction methods that build a 3D model from a series of photographs.

The goal of this project is to create an application that allows the user to spontaneously scan textured objects and obtain their 3D models with a single camera. The idea is inspired by the ProFORMA system by Qi Pan et al. The scanning process is interactive: the camera is stationary during the process and the user rotates the object in front of the camera. The background and the user's hand are filtered out using geometry constraints. The user can see the actual state of the reconstructed model reprojected on the camera image. The system also supports the user in the reconstruction by showing orientation cues for previously unseen directions.

The outcome of the thesis is an implementation of the ProFORMA algorithm in OpenCV. Some initial building blocks are already available from a previous student project.

Qi Pan et al. - ProFORMA: Probabilistic Feature-based On-line Rapid Model Acqusition, BMVC, 2009