Quantitative Prediction of Material Properties Using Reflectance Spectroscopy: A Multivariate Chemometrics-based Approach
Dan Shiley, ASD Inc.
This paper was presented at the Art, Science and Applications of Reflectance Spectroscopy Symposium sponsored by ASD Inc. and IEEE GRSS, February 23-25, 2010 in Boulder, Colorado.
Proceedings of ASD and IEEE GRS; Art, Science and Applications of Reflectance Spectroscopy Symposium, Vol. II, 10pp, Boulder, CO, www.asdi.com.
Author: Dan Shiley
Affiliation: ASD, Inc., Boulder, CO USA
Abstract
Spectral reflectance-based multivariate models that utilize large numbers of spectral points to predict sample composition have been widely accepted for quantitative and qualitative analysis within agricultural, pharmaceutical and industrial markets. However, multivariate models have not been widely used for remote sensing applications or for quantitative modeling of minerals in rocks. In order to develop a multivariate calibration model we need to understand the error of the reference measurements so that realistic expectations of model performance can be established. Lab error and its effect on the calibration model will be discussed. Calibration and validation sets must contain the diversity, both spectral and compositional, that would be expected to be encountered in new samples. Additional considerations for the development of multivariate calibration models include sample presentation, creation of the calibration and validation sets, model statistics, calibration monitoring and performing updates to the calibration model.
Full article link