#TITLE_ALTERNATIVE#

• Main Author: WARDHANA DJATMIKO (15108040); pembimbing : Dr. Deni Suwardhi, ST., MT ; Dr. Ir. Agung Budi Hart, JUDA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/16191
• Institution: Institut Teknologi Bandung

Point cloud is a way to represent objects in 3D by using a set of points. By using point cloud, it is possible to reconstruct the geometry of an object. Point cloud can be made using a series of photographs that are taken from various angles, and then further processed using digital photogrammetric technology. In this paper, some softwares are used to generate point cloud and the results are compared. The software used are OSMBundler+PMVS2,Microsoft Photosynth, My3DScanner, Hypr3D, and Agisoft. The point cloud is then used to reconstruct the geometry of a building roof by using the concept of curvature and 3D linear regression. Curvature basically shows how much a surface of an object bends, therefore it is possible to extract points from a surface (mesh) based on the curvature value, specifically those points located on the edge of the object. Linear regression is then used to model the roof segmens by calculating the coordinate of end-points of each roof segments. The 3D linear regression concept used for the calculation is Orthogonal Distance Regression. From five softwares used to produce the point cloud, My3DScanner and Hypr3D produced the best geometry. On the modeling method, the result has not given accurate building geometry yet and is very dependent on the quality of the mesh. The main drawback of this method is the 3D linear regression concept which can only be used to model a straight roof segment.