Improving the Discriminatory Power of a Near-Infrared Microscopy Spectral Library with a Support Vector Machine Classifier