The application of spectral invariants for discrimination of crops using CHRIS-PROBA data

Abstract

Numerous studies have demonstrated the ability of hyper-spectral data to discriminate crop types, however most methods rely on empirical data and are therefore site specific. In this brief proceeding we provide a physically based approach for separation of crop types using multiangle hyperspectral data. We use the radiative transfer theory of spectral invariants which allows for the parameterization of the canopy reflectance into two spectrally invariant and structurally varying parameters — recollision and escape probabilities. The spectral invariant parameters are retrieved from the CHRIS/PROBA multiangle hyperspectral sensor. We present the spectral invariant parameters in spectral invariant space. The horizontal axis provides information about macro scale features such as plant shape and size as well as ground cover. The vertical axis provides information about microscale features such as leaf density as well as portion of sunlit to shaded leaves. These features allow for the natural separation of crops. In addition we illustrate the potential for further separation of crop types based on angular information. Results suggest that multiangle information is important for canopies with similar structural features in the nadir direction.