Application of unsupervised end member detection algorithms for spectral unmixing of hyperspectral data for mangrove species discrimination

Abstract

The Sunderban Biosphere Reserve of West Bengal, India is an ideal locale where hyperspectral image data may be successfully utilized for accurate mapping of nearly 94 mangrove species that exist there. The present study is the first attempt to use hyperspectral data in the Sunderban eco-geographic province to enable species level discrimination of mangroves. As priori knowledge of mangrove species distribution in most of the densely forested islands of the Sunderbans is not available, this paper applies unsupervised automated target detection algorithms such as N-FINDR and ATGP for detection of end members (mangrove species) from the hyperspectral image data. The pixels comprising of either homogeneous or mixed mangroves species are unmixed using both constrained and unconstrained linear mixing model and the fractional abundance images of the detected species generated. It has been found that the abundance images generated after unconstrained linear unmixing shows more accuracy with use of end members generated by N-FINDR algorithm as compared to that of constrained linear unmixing with ATGP as well as N-FINDR. The sub pixel classified results have led to the identification of species dominant in Henry's Island to be Avicennia Marina, Avicennia Officinalis, Excoecaria Agallocha, Ceriops Decandra, Phoenix Paludosa and Aegialitis. The area also comprises mixed patches of Ceriops-Excoecaria Agallocha as well as Aegialitis-Avicennia Marina var aquitesima in many places.