ニューラルネットの感度特性を用いた要因特定法と真珠色彩識別への応用

DOI Web Site Web Site 被引用文献5件 参考文献12件

書誌事項

タイトル別名
  • Factors Identification Using Sensitivity of Layered Neural Networks and Its Application to Pearl Color Evaluation
  • ニューラル ネット ノ カンド トクセイ オ モチイタ ヨウイン トクジョウホ

この論文をさがす

抄録

This paper proposes a new method for extracting the sensibility (kansei) of human experts, and applies it to the design of a pearl color evaluation system. A factors identification method using sensitivity of layered neural networks finds the influential factors as the input units which have the greatest influence on all outputs. This method has been applied to the network mapping between spectral reflectance of pearls and the pearl color categories evaluated by experts, and allowed some specific wavelength bands to be identified as the essential factors contributing to the evaluation. Based on this result, a pearl color evaluation system has been designed using multi-bands images (MBI) acquired by combining a monochrome CCD camera with several interference filters corresponding to the identified major wavelength bands factors. Moreover, the identification method has been applied to find the influential factors of each image of MBI and finally the experimental results of discrimination using those factors show that this system is able to give evaluations of pearl color successfully as an expert does.

収録刊行物

被引用文献 (5)*注記

もっと見る

参考文献 (12)*注記

もっと見る

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ