Applying Spectral Invariants to High Spatial Resolution Spectral Imaging of Vegetation

Abstract:

Optical remote sensing has a well-established role in the monitoring of vegetation for economic and environmental purposes. Most current applications are based on multispectral medium-resolution satellite imaging data, where the pixel size ranges approximately from 10 m to 300 m. In recent years, high spatial resolution spectral imaging data collected by commercial satellites, Unmanned Aerial Systems and traditional aircraft-based instruments has become increasingly available. However, the current vegetation mapping algorithms, developed for medium-resolution data, are based on solutions to the radiative transfer equation (RTE) with infinitesimally small scatterers. As such, when the size of a canopy element size exceeds that of a pixel, RTE-based algorithms are unable to realistically describe the spectral properties of the measured radiance scattered from a canopy. We propose the first physically-based algorithm for vegetation mapping with high spatial resolution resolution data utilizing the theory of spectral invariants, which has already been successfully applied to medium-resolution data. The theory of spectral invariants is a computationally efficient method that uses wavelength-independent (spectrally invariant) parameters for characterizing canopy spectral properties. Expanding the theory of spectral invariants to apply for high-resolution data will allow the retrieval of the true spectral response of leaves. This poster provides an overview of the theoretical framework of spectral invariants applied to high-resolution data, and showcases its application using simulated high-resolution spectral imagery.