Lou, Shan (2013) Discrete algorithms for morphological filters in geometrical metrology. Doctoral thesis, University of Huddersfield.

In geometrical metrology, morphological filters are useful tools for the surface texture analysis and functional prediction. Although they are generally accepted and regarded as the complement to mean-line based filters, they are not universally adopted in practice due to a number of fatal limitations in their implementations —they are restricted to planar surfaces, uniform sampled surfaces, time-consuming and suffered from end distortions and limited sizes of structuring elements.

A novel morphological method is proposed based on the alpha shape with the advantages over traditional methods that it enables arbitrary large ball radii, and applies to freeform surfaces and non-uniform sampled surfaces. A practical algorithm is developed based on the theoretical link between the alpha hull and morphological envelopes. The performance bottleneck due to the costly 3D Delaunay triangulation is solved by the divide-and-conquer optimization.

Aiming to overcome the deficits of the alpha shape method that the structuring element has to be circular and the computation relies on the Delaunay triangulation, a set of definitions, propositions and comments for searching contact points is proposed and mathematically proved based on alpha shape theory, followed by the construction of a recursive algorithm. The algorithm could precisely capture contact points without performing the Delaunay triangulation. By correlating the convex hull and morphological envelopes, the Graham scan algorithm, originally developed for the convex hull, is modified to compute morphological profile envelopes with an excellent performance achieved.

The three novel methods along with the two traditional methods are compared and analyzed to evaluate their advantages and disadvantages. The end effects of morphological filtration on open surfaces are discussed and four end effect correction methods are explored. Case studies are presented to demonstrate the feasibility and capabilities of using the proposed discrete algorithms.

Final_Thesis_-_May_2013.pdf - Accepted Version

Download (6MB) | Preview


Downloads per month over past year

Add to AnyAdd to TwitterAdd to FacebookAdd to LinkedinAdd to PinterestAdd to Email