Iterative algorithms for abundance estimation on unmixing of hyperspectral imagery

Hyperspectral sensors collect hundreds of narrow and contiguously spaced spectral bands of data organized in the so-called hyperspectral cube. The spatial resolution of most Hyperspectral Imagery (HSI) sensors flown nowadays is larger than the size of the objects being observed. Therefore, the measured spectral signature is a mixture of the signatures of the objects in the field of view of the sensor. The high spectral resolution can be used to decompose the measured spectra into its constituents. This is the so-called unmixing problem in HSI. Spectral unmixing is the process by which the measured spectrum is decomposed into a collection of constituent spectra, or endmembers, and a set of corresponding fractions or abundances. Unmixing allows us to detect and classify subpixel objects by their contribution to the measured spectral signal. In this research, two new abundance estimation algorithms based on a least distance least square problem and compare it with other approaches presented in the literature were developed. Algorithm validation and comparison are done with real and simulated HSI data. HSI Abundance Estimation Toolbox ( HABET) was implemented in the ENVI/IDL environment. Application of the unmixing algorithm for remote sensing of benthic habitats is presented.