期刊名称：Revista Raega: O Espaço Geográfico em Análise
印刷版ISSN：1516-4136
电子版ISSN：2177-2738
出版年度：2019
卷号：46
期号：3
页码：48-62
DOI：10.5380/raega.v46i3.67098
摘要：The complexity of pixel composition of orbital images has been commonly referred to the spectral mixture problem. The acquisition of endmembers (pure pixels) direct from image under study is one of the most commonly employed approaches. However, it becomes limited in low or moderate spatial resolutions due to the lower probability of finding those pixels. In this way, this work proposes the combined use of images with different spatial resolutions to estimate the spectral responses of the endmembers in low spatial resolution image, from the obtained proportions derived from the spatial higher-resolution images. The proposed methodology was applied to products provided by PROBA-V satellite with spatial resolution of 100 m and 1 km in the Pantanal region of Mato Grosso state. Initially, the fraction images (proportions) were generated from the 100 m dataset using the endmembers selected directly in the image, considering the higher probability of finding pure pixels in such images. Following the spectral responses of the endmembers in 1 km were estimated by multiple linear regression, using the proportions of the endmembers in the pixels derived from 100 m images. For the evaluation, the endmembers fraction images were compared and field data was used. These analyses indicated that the spectral responses estimated allowed to improve the results with regard to error, to variability, and to the identification of endmembers proportions, considering that inadequate choice of pixels considered as pure in low spatial resolution images can affect the quality of the fraction images for operational use.
其他摘要：The complexity of pixel composition of orbital images has been commonly referred to the spectral mixture problem. The acquisition of endmembers (pure pixels) direct from image under study is one of the most commonly employed approaches. However, it becomes limited in low or moderate spatial resolutions due to the lower probability of finding those pixels. In this way, this work proposes the combined use of images with different spatial resolutions to estimate the spectral responses of the endmembers in low spatial resolution image, from the obtained proportions derived from the spatial higher-resolution images. The proposed methodology was applied to products provided by PROBA-V satellite with spatial resolution of 100 m and 1 km in the Pantanal region of Mato Grosso state. Initially, the fraction images (proportions) were generated from the 100 m dataset using the endmembers selected directly in the image, considering the higher probability of finding pure pixels in such images. Following the spectral responses of the endmembers in 1 km were estimated by multiple linear regression, using the proportions of the endmembers in the pixels derived from 100 m images. For the evaluation, the endmembers fraction images were compared and field data was used. These analyses indicated that the spectral responses estimated allowed to improve the results with regard to error, to variability, and to the identification of endmembers proportions, considering that inadequate choice of pixels considered as pure in low spatial resolution images can affect the quality of the fraction images for operational use.
