PLS回帰分析を用いた近赤外光物質判別及び最適波長選択手法

抄録

In this paper, we present a proposal of material discrimination method based on the spectral reflectance properties in the near-infrered (NIR) wavelength range and informative wavelengths selection to discriminate multiple target materials such as the human skin, plant and asphalt. A material is detected from its image by using only two or three wavelengths corresponding to the absorption wavelength and the other wavelength based on the spectral reflectance property of the material. However discrimination between more than two materials are not always made by using only absorption wavelengths of each material. Even if the materials were discriminated, whether or not the wavelength was informative is not known. Here we discriminated three material regions by using multiple NIR wavelengths. In this proposed method, binary classifiers are generated from each material by Partial Least Squares (PLS) regression analysis, and three materials are discriminated by predicted values obtained by the classifiers. Also, the informative wavelengths are selected by Variable Influence Projection (VIP) obtained by PLS regression analysis. As a result of the experiment, three materials were discriminated with high accuracy by using only five wavelengths.