ファジィ推論とニューラルネットワークを用いた人物画像の認識に関する研究

概要

論文の詳細を見る
This paper proposes an image extraction technique of the human face by color image processing, fuzzy inference, and neural network. The image processing technique provides a method for extraction of skin color and measurement of detailed information of the human face. The fuzzy inference estimates the face direction based on the configuration of the extracted face candidates, while the neural network distinguishes faces and other objects by detailed information and face direction. Compared with the previous works which extract data on the human full face, this method makes it possible to recognize not only a full face but also profiles. Further, we show how this method works for the illusion problem. The extraction accuracy of this system is 80 percent at this moment. According to the experimental results, this system will be useful for the vision systems of autonomous robots.
一般社団法人日本機械学会の論文
1993-02-25