Sgavetti M., Gaibazzi A., Longhi I., Meli
S., Giardino C., Capelletti R., Baraldi A., Crotti M.
ABSTRACT
Noisy spectra were simulated from laboratory 2000-2500 nm reflectance spectra of polymineralic sand samples and synthetic composite spectra, using MIVIS calibration parameters and measured solar irradiance both tith and without diffuse atmospheric radiance at the sensor. Noise content variation in output datawith varying sensor parameters and atmospheric conditions was analysed, as well as noise effetcs in spectral feature identification. Scaled random noise with a normal distribution was added to teh convolved spectra. The rms of the ratio of noisy to original spectrais assumed to represent noise content in output data. Noise content variation with SNR is expressed by a set of curves with a significant dispersion of noisy spectra for low SNRs. For individual SNRs, noise content ha an inverse linear relationship with sample albedo. Data dispersion occurs when noise content is compared with reflectance contrast and band depth. No clear relationship resulted for noisy to original spectra match values or noisy spectra classification probability with either band depth or band depth referred noise, due to spectra aliases.