Comparative classification approach in hyperion imagery

TLDR: The results showed that the sub-pixelbased classification produces a better distribution of Iron ore than the per pixel-based classification.

Abstract: The traditional approaches to estimate the Iron ore involves large manpower, cost and time. Iron ore identification is necessary due to the rapid increase in construction work, industries and population. Hyperspectral Imagery analysis used to estimate the Iron ore precisely depends on the spectral signature. The spectral signature of Iron ore shows huge absorption in 865 nm due to the presence of Iron content in the sample spectra. Hyperspectral imagery contains a large number of spectral bands and involves various processing steps such as identification of the calibration bands, absolute reflectance generation, data dimensional minimization, Iron ore endmembers extraction and classification. The radiance imagery absolute reflectance bands are carried out using FLAASH atmospheric correction module. The noiseless pure pixels are obtained using data dimensionality reduction techniques as spectral data reduction and spatial data reduction. The comparative analysis is performed between sub-pixel (LSU) and per-pixel (SAM) classification. The results showed that the sub-pixelbased classification produces a better distribution of Iron ore than the per pixel-based classification.