Multiple relational graphs mapping using genetic algorithms

This paper presents a genetic algorithm (GA) based optimization procedure for pattern classification in a model-based recognition system using the attributed relational graph (ARG) matching technique. The test scene may contain multiple overlapped instances of any model objects. The proposed algorithm matches a scene ARG to all model ARGs simultaneously in one genetic search. In this study, potential solutions are represented by integer strings. The population is initialized randomly and the GA is applied on this initial population to obtain the solution. We present experimental results to demonstrate the procedure.