Automatic indexation of cultural heritage 3D object

Abstract

There has been significant evolution in the fields of 3D digitization thanks to the development of 3D reconstruction and geometry processing. The results of digitization researches have been widely applied in many fields, especially in Cultural Heritage and Archaeology. Reconstruction, characterization and annotation of components forming 3D objects have become an effective tool for research, conservation and promotion of archaeological relics. The aim of this paper is to propose a process of 3D model reconstruction, segmentation and annotation on the basis of a enhanced corresponding 2D dataset. A machine learning method is used for the semantic segmentation of 2D images, thereby label, annotate and reconstruct a 3D model based upon links between distinctive invariant features, orientation of images, and depth map of images. The initial result as a data basis for research, reconstruction and identification of parts in 3D objects is applied in the reconstruction of archaeological relics, object identification, 3D printing, etc. Our work uses the data collected from the Museum of Cham Sculpture DaNang and the Myson QuangNam sanctuary in VietNam, to carry out the proposed method.