Reconstruction of Polygonal Faces from Large-Scale Point-Clouds of Engineering Plants
Hiroshi Masuda · 2014
The recent progress of mid-range and long-range laser scanners makes it possible to capture dense point-clouds of manufacturing plants. 3D models of manufacturing plants are useful for simulating the reorganizing of production lines, but a raw point-cloud does not have structures suitable for shape manipulation, and the data size is too large to handle for common PCs. Therefore, it is often required to convert point-clouds to simplified mesh models. So far, researchers have studied shape reconstruction of pipe structures by detecting cylindrical surfaces from large-scale point-clouds. In the reconstruction of pipes, the shape of a pipe can be easily reconstructed only by estimating the length of the cylinder when a cylindrical surface is extracted. On the other hand, planar faces may have very complicated boundaries, because surfaces are often be occluded by other surfaces in manufacturing plants. In addition, point-clouds by mid/long-range laser scanners are very noisy and contain a lot of outliers, and the boundary of each planar face becomes a perturbed curve. In this paper, we propose an efficient method for constructing simplified polygonal faces without the loss of detail shapes.