Geometrical constraints for object recognition using geneticalgorithms

摘要

In this paper we describe how different types of constraints canaffect the performance of genetic algorithms (GAs). The success of a GAapplication depends on efficient constraints to provide a cleardirection for GA search. Our application integrates geometricalconstraints with a GA to guide the pattern matching process for imageregistration and object location. Two types of constraints, namely,local and global constraints are considered Although both types ofconstraints are useful in constraining the pattern matching searchspace, the former type is less effective than the latter type.Intuitively one would expect that the combination of both types ofconstraints should be more powerful than each of them used alone. Yetour experimental result proves to the contrary. We describe the wholeprocess of integrating geometrical constraints with a GA for patternmatching and analyse the results