Intellectual Eye Detection Using GA Multi Window

Autor: Komoda, Daisuke, Nishimura, Hiromitsu, Tomikawa, Takehiko
Jazyk: japonština
Rok vydání: 2005
Předmět:
Zdroj: 神奈川工科大学研究報告.B,理工学編. 29:55-62
ISSN: 0916-1902
Popis: application/pdf
There have been a variety of eye detections reported in the past, however, many of approaches require the geometrical relations among the parts in facial image. We have reported the way of eyes' detection by window-pair chasing in a facial image based on Genetic Algorithm. Here, symmetry of eyes was utilized in order not to fall into incorrect locations while finding eye locations, although there were some Probiems remained for its detecting capability. in this paper, our improved version, so caiied Multi-Window model is introduced. The Multi-Window, three windows are used in this experiment; the window pair for two eyes and the rest one for either mouse or nose. Thus, our trial applying to sample images resulted in the recognition rate of more than 99%. On the other hand, parameters to activate GA must be decreased and/or optimized without any help of human in the future.
Databáze: OpenAIRE