Band selection algorithms for foliar trait retrieval using AVIRIS-NG: a comparison of feature based attribute evaluators

Abstract

Interband information overlapping enhances redundancy in hyperspectral data. This makes identification of application-specific optimal bands essential for obtaining accurate information about foliar traits. The current study investigated the performance of three novel Band Selection (BS) algorithms (i.e. the Chi-squared-statistics based attribute evaluator (CSS), the Recursive elimination of features-based attribute evaluator (REF) and the Correlation-based attribute subset evaluator (CBS)) in identifying the spectral bands of Airborne Visible/Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) from visible and Near Infrared (NIR) regions that are sensitive to variation in Chlorophyll Content (CC). Identified bands were employed to formulate Hyperspectral Indices (HIs) by incorporating combinations of Blue, Green, Red, and NIR regions. CC models were built by establishing a linear fit between ground CC and HIs. For all the three BS algorithms, optimum bands varied for visible and NIR regions. REF-HI (NIR,R), REF-HI(NIR,R + G), CSS-HI(NIR,R) and CSS-HI(NIR,R + G) had the best correlation with CC. HI(NIR,R) is identified as the best HI and REF the best BS algorithm for retrieving CC.

Keywords:

• Leaf chlorophyll content
• forestry
• hyperspectral indices
• AVIRIS-NG
• band selection

Acknowledgements

First author is thankful to Department of Science and Technology and Science and Engineering Research Board for National Post-Doctoral Fellowship (PDF/2017/002620). Authors extend sincere thanks to Ministry of Environment, Forest and Climate Change and Gujarat Forest Department for permitting ground-truthing in the study area. Authors thank Indian Space Research Organization for support. Authors are also grateful to forest officials for their helping hands during the field sampling.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

First author is thankful to Department of Science and Technology and Science and Engineering Research Board for National Post-Doctoral Fellowship (PDF/2017/002620). Authors extend sincere thanks to Ministry of Environment, Forest and Climate Change and Gujarat Forest Department for permitting ground-truthing in the study area. Authors thank Indian Space Research Organization for support. Authors are also grateful to forest officials for their helping hands during the field sampling.