Describes a CAD-model-based machine vision system for dimensionalinspection of machined parts, with emphasis on the theory behind thesystem. The original contributions of the work are: the use of precisedefinitions of geometric tolerances suitable for use in imageprocessing, the development of measurement algorithms correspondingdirectly to these definitions; the derivation of the uncertainties inthe measurement tasks; and the use of this uncertainty information inthe decision-making process. Experimental results have verified theuncertainty derivations statistically and proved that the errorprobabilities obtained by propagating uncertainties are lower than thoseobtainable without uncertainty propagation
展开▼