Meta-GA Multi-Window Human Eye Detection

摘要

This paper reports on the application of meta-GA in abstraction of eyes from a facial image. The search model and evaluation function developed for the detection are presented. Therninvestigation was in two major parts. First, experimentally obtained optimal parameters were used in a single-GA search using the proposed model. Secondly, meta-GA was used in the determination of the optimal parameters for the search GA and the investigation repeated using these parameters. The results obtained for the two phases show a marked improvement on earlier detection results using a single GA and simpler search model thus suggesting an advantage in using the new model. The results from the meta-GA search show that heuristic setting of search parameters through GA while solving an optimization problem, in this case eye abstraction from a facial image, has great possibilities and should in fact ease the adaptation of GA to different optimization problems.