3D surface reconstruction of smooth and textureless objects
Jahanzeb Hafeez, Soonchul Kwon, Seung-Hyun Lee, Alaric Hamacher · 2017
In this paper we propose a system to reconstruct the 3D model of an object from a set of 2D overlapped images under the assumption that the object does not have distinct features. We explicitly consider the object that is made of one or a number of smooth surfaces that radiant constant energy isotropically in all directions. We designed a portable and low cost multi camera rig system which can capture the images from all cameras simultaneously. In order to provide the artificial markers, we used two small projectors that project patterns over the object before capturing the images. We compared the 3D model reconstructed from multiple images with a CAD model to evaluate the performance of the proposed system. A simple algorithm is also proposed to acquire the original texture or color of the subject.