A scalable approach for 3D mesh generation

We present an algorithm for creating a CAD model of an existing physical object from a cloud of points. The points can be obtained by one of the surface reconstruction cues; laser scanner, stereo imaging or CT scans. The algorithm has a linear complexity and can be applied to processing big clouds. It also can handle redundant as well as missing data. The output of the proposed mesh generation algorithm can be in the form of STL files that are suitable for most rapid prototyping machines. The proposed algorithm can be broken into three phases. In the first phase, large data sets are grouped into planar cross sections along the object central axis. Each data section is processed separately. In the second phase, a 2D active contour is used to fit a curve and resample the data points. In the third phase, the data points in the consecutive layers are connected to each other to construct the 3D mesh from a set of triangle patches. A detailed description of the algorithm is presented in this paper.