Mapping Paddy Growth Stage Based-on Hyperspectral EO-1 Hyperion Using Pixel Purity Index Endmember Extraction Algorithm

Abstract

Hyperspectral remote sensing is widely used in monitoring vegetation because it provides a high spatial and spectral resolution. Thus, its ability to distinguish between various objects on earth. However, many problems arise in processing Hyperspectral data. In this paper, Pixel Purity Index algorithm is used in addressing this issue. PPI is an endmember extraction method which is widely used in hyperspectral data processing because it can handle mixed pixel on large resolution images as well as reduce the dimensionality of the data. In vegetation mapping, determining the wavelength plays an important role in object reflectance analysis. In this study, paddy reflectance characteristics are examined, where results show that the characteristics of paddy occur in wavelength 447, 701, 1024, and 1104 nm. In the region of VIS-red edge (447- 701nm), curve below value of 0.1 can be used to distinguish among paddy and non-paddy vegetation reflectance, while the combination of the increase/decrease value in red edge-NIR (701-1027nm) and NIR-SWIR (1027-1104nm) range, can be used as a reference for analyzing the reflectance of rice growth, wherein the mean value of red edge-NIR can be used as predictors in distinguishing the paddy growth stage.