Spatial and spectral optimization of object definition for leaf area index and aboveground biomass estimation from airborne spectroscopic images

Abstract

Hyperspectral images are a useful source to retrieve estimations of vegetation parameters like leaf area index (LAI) and aboveground biomass (AGB). These estimations are more reliable when using image objects rather than individual pixels as spatial units. Object-based image analysis groups neigbouring pixels based on spectral similarity. Two issues to be considered then are: 1) how similar the pixels should be, and 2) which spectral bands should be included for object definition