Semi-automated method for views integration based on alternative color and geometry characteristics

The pace of development of information systems nowadays demonstrates the magnitude of the demand for digitization of all aspects of our lives, such as medicine, industry, and documentation of cultural heritage. Digitization is the process of converting objects from the real world into their digital representations. In order to acquire complete and detailed information about the whole surface of an object, several 3D scans have to be taken from different perspectives. The resulting 3D object can be acquired in a form of a numerous amount of 3D point clouds overlaying each other. Sometimes, depending on a quality of a 3D scanner and surface properties, the point clouds can represent a noisy geometrical surface and an incorrect colour. Moreover, the directional point clouds are not perfectly aligned and a registration between them must be applied. The registration of the point clouds is a complex task which is not always possible to automate. Usually, the entire process of registration has to be supervised by a skilled operator. The registration is usually divided into two parts: initial and final matching. Initial matching is a more complex one and in this scenario, it is supported by the known system calibration, which includes, e.g., robotic arm, head of the scanner, sources of lights. Using ICP based algorithms afterward is usually enough to get appropriate final matching. The difficulty of point cloud registration increase accordingly to the number of directional clouds of points to integrate. The aim of this paper is to propose a methodology to decrease or even fully eliminate some of the presented registration issues encountered during the reconstruction of Museum of King Jan III’s Palace at Wilanow.