Automatic 3D Imaging and Modelling System with Color Information for Cultural Heritage Digitization

Abstract

Automatic digitization of three-dimensional (3D) complex objects in a real scene using fringe projection profilometry (FPP) [1-3] has become an important issue in many application areas, especially for the cultural heritage protection. Because of the limited field of view (FOV) of 3D sensor, it is difficult for any 3D sensor to digitize the whole object automatically through a single acquisition, especially for the objects with complex geometric shape and topology. This issue could be addressed with a strategy of mobile digitization by single 3D optical sensor moving around the object. And the iterative closest points (ICP) algorithm, tracking the 3D sensor using other instruments, or other methods is utilized to register the multiple range images [4]. It is difficult for this strategy to achieve automation. Also a larger workspace is required for moving the 3D sensor. In our paper, we focus on studying a strategy of automatically controlled digitizing process and a complete digitization of 3D complex objects. A strategy that utilizes a one-dimensional (1D) array of 3D optical sensors combined with a rotation turntable to get a complete digitization of complex object automatically is proposed. With this approach, two important tasks should be concerned: (1) the calibration of multiple 3D optical sensors, and (2) the automatic registration of multiple range images taken from sensor array. In addition, for digitizing the cultural heritage, the color information is always necessary. Then a high-resolution digital camera is employed to take the photos of the object from different viewpoints. The key issue for correctly mapping of color photos onto the object surface is to accurately determine the mapping relationship between the color photo and the surface model. This can be done with help of the photogrammetry technique. Afterwards, we suggest a texture blending technique that utilizes a composite-weight strategy to blend the color images within the overlapped region, resulting in a photorealistic model.