ドイ リョウイチ
  土居 良一   社会学部 社会学科   教授
■ 標題
  Doi 2016. Enhancing multispectral discrimination among vegetation types with a new pseudo-color imaging method
■ 概要
  In this study at the Sakaerat Biosphere Reserve of Thailand, 15 derivative grayscale images were generated from grayscale images, for redness, greenness, and blueness, among 7 Landsat grayscale images. The use of all 22 grayscale images provided an additional principal component and a larger number of pixel clusters. As evergreen forest is the natural vegetation type of Sakaerat, all pixels in the image were grayscaled based on the principal component score as an indicator of the clusters' evergreen forest- likelihood. Using separate sets of the 7 Landsat images and all 22 grayscale images. Dunnett's t-test completely discriminated values of evergreen forest-likelihood for seven 300 m x 300 m plots with different vegetation types, favored by a difference in the pattern of mean separation between the image sets. With all 22 grayscale images, the evergreen forest plot was more significantly discriminated from that of fire-protected deciduous forest compared to the Landsat images alone. Thus, differences in visible reflectance revealed by the derivative grayscale images quantified the degree of ecological restoration more strictly than the conventional Landsat images. The proposed imaging method would thus improve the real-time observation of forest and other canopies when used together with multispectral sensors.
  単著   Siva Lus   Instituto Nacional de Investigacao Agraria e Veterinaria, I.P.      2016


Copyright(C) 2011 Daito Bunka University, All rights reserved.