Analysis of end member detection and subpixel classification algorithms on hyperspectral imagery for tropical mangrove species discrimination in the Sunderbans Delta, India

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 dominant mangrove species that exist there. This study attempts to apply and analyze automated end member detection algorithms such as N-Finder (N-FINDR) where N is the number of end members and automated target generation process (ATGP) on Hyperion data of the study area to enable species-level discrimination of mangroves. The identified end members have been further unmixed using constrained and unconstrained linear unmixing and the fractional abundance images of individual end members generated. It has been found that classification results generated by unconstrained linear unmixing with N-FINDR algorithm shows higher accuracy than the unconstrained classification results of ATGP. The classified output of unconstrained linear unmixing also shows higher accuracy than constrained linear unmixing with ATGP and N-FINDR. The subpixel classification output identifies dominant species on the study area to be Avicennia Marina , Avicennia Officinallis , Excoecaria Agallocha , Ceriops Decandra , Phoenix Paludosa , Bruguiera Cylindrica , and Aegialitis. The results also identify mixed patches of Ceriops-Excoecaria Agallocha and Aegialitis-Avicennia Marina var aquitesima in many places. The accuracy assessment of subpixel classification showing mangrove species distribution has been done by generation of confusion matrix and calculation of kappa coefficient of field data collected during ground survey. It has been observed that unconstrained linear unmixing with N-FINDR has been more successful in detecting more target species as compared to the other algorithms applied.