Application of Formosat-2 Multispectral Image on the Classification of Forest Landuse

碩士 === 國立嘉義大學 === 森林暨自然資源研究所 === 94 === This study aimed to investigate the landuses in Chilan Forest Area and the spectral characteristics (SC) by Formosat-2 (Formosa Satellite 2), and the spectral variations of landuses with multi-spatial spectral signatures through combining space concepts. The i...

Full description

Bibliographic Details
Main Authors: Cheng-Yi Tsai, 蔡正一
Other Authors: Chinsu Lin
Format: Others
Language:zh-TW
Published: 2004
Online Access:http://ndltd.ncl.edu.tw/handle/77038564838620949492
Description
Summary:碩士 === 國立嘉義大學 === 森林暨自然資源研究所 === 94 === This study aimed to investigate the landuses in Chilan Forest Area and the spectral characteristics (SC) by Formosat-2 (Formosa Satellite 2), and the spectral variations of landuses with multi-spatial spectral signatures through combining space concepts. The important conclusions were: the signal to noise ratios (SNRs) of the original image were 68.54, 23.41, 18.64, and 6.93 for the blue, green, red and near infrared band respectively, and the later SNRs of the atmospheric corrected image would been decreased. After the expanded vegetation index, there was well spectral signatures separated among Chamaecyparis formosensis, Cryptomeria japonica, broad-leaves mixed stands (BLMS), and conifer and broad-leaf mixed forests (CBLMF) in the area exposed directly under the sun (AEDS). The minimum transformed divergence (TD) was 1950, but the terrain effect might so confuse this result to reduce the classification accuracy of landuses. When the spatial scales were enlarged(from 8m to 40m), the spectral transformed divergence between BLMS and CBLMF would reduce; however, that of Cryptomeria japonica and CBLMF ascended. In addition, the TD of other landuses were not significant change. There was no significant direct or inverse proportion between multi-spatial spectral signatures and the overall accuracy; therefore, different results of space image analysis would be influenced by terrain effect. In a word, the best overall accuracy of bare soil, water body, Chamaecyparis formosensis, Cryptomeria japonica, BLMS, and CBLMF in Chilan Forest was 88.06%, and Kappa-hat coefficient was 0.8710.