Primitive Shape Extraction for Objects in Point Cloud

摘要

Shape plays an important role in object recognition and shape modeling. In this paper, a comprehensive method is proposed to analyze and implement crucial primitive shape extraction including line, plane, cylinder and sphere. Linear shape is extracted by the proposed alpha-shape based method, and an algorithm that robustly computes the boundary of holes from incomplete objects is presented. The plane primitive shape is extracted by analyzing neighboring point distribution and the parameters that define the primitive shape. Reasonable primitive shape features is extracted accurately for further shape modeling. Experimental results demonstrate that our method could deal with different objects efficiently in different shapes.