Online surface reconstruction from unorganized 3D-points for the DLR hand-guided scanner system

Abstract

Hand-guided scanners allow for digitization by manually sweeping a laser beam over an object's surface. The result highly depends on the way the user handles the system and his ability to keep track of the parts of the surface that are already scanned. Processing and visualization during data acquisition are helpful in this context. In this paper, we propose an online surface reconstruction algorithm for the visualization of the DLR scanner system data. The algorithm successively generates a triangle mesh by incrementally inserting 3D points. Point neighborhoods are used to limit the point density, to estimate the surface normal at the inserted point, and to locally retriangulate the mesh. A dynamic data structure for fast neighborhood search without restrictions to the amount of vertices or the object size and with low complexity is introduced. Finally, results with the hand-guided scanner system are presented.