ACCURATE MATCHING OF 2-DIMENSIONAL SHAPES USING THE MINIMAL TOLERANCEZONE ERROR

Automated visual geometric inspection of machined parts depends upon t he ability to determine the underlying parameters of a shape model tha t will best fit the reference shape to the real shape captured. This p aper presents a parametric approach for the matching of two-dimensiona l profiles, which are composed of straight-line segments and circular arcs, based on the tolerancing requirements defined by the AINSI stand ards. The shape matching problem is formulated as a minimax optimizati on model and a procedure for solving this optimization model is develo ped. Experimental results have shown the proposed algorithm to be cons istently accurate and extremely fast. (C) 1997 Elsevier Science B.V.