HIERARCHICAL PRIMITIVE SURFACE CLASSIFICATION FROM TRIANGULATED SOLIDS FOR DEFINING PART-TO-PART DEGREES OF FREEDOM

摘要

Mesh segmentation is the process of organizing a set of data points into connected areas defined by known surface primitives or surface equations. Our approach is a hierarchical method based on clustering and Mamdani's fuzzification system. First, a clustering algorithm is used to isolate regions of small and irregular oriented triangles, which make up a large portion of the total polygonal faces. Then, using fuzzification rule sets, the remaining triangles are made into meaningful primitives: cylinder, cone, sphere and flat. The end goal of this work is to use the primitives to define the degrees of freedom between mating parts in an assembly when such information is unknown or lost. The result of this process has proven to be accurate with well-defined borders with no need of additional post-processing steps.