Premium
Linear quantification calibration of crystallinity at subpercent and its evaluation based on spectral and spatial information inherited in Raman chemical images
Author(s) -
Wu Jianping
Publication year - 2014
Publication title -
journal of raman spectroscopy
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.748
H-Index - 110
eISSN - 1097-4555
pISSN - 0377-0486
DOI - 10.1002/jrs.4516
Subject(s) - calibration , crystallinity , univariate , raman spectroscopy , mean squared error , pixel , root mean square , correlation coefficient , statistics , mathematics , remote sensing , biological system , analytical chemistry (journal) , artificial intelligence , computer science , chemistry , optics , physics , geography , multivariate statistics , chromatography , crystallography , biology , quantum mechanics
In this paper, the author reported two methods to extract spectral or spatial information inherited in the Raman chemical images for linear quantification calibration of crystallinity. The two approaches reported quantification results according to the spectral mean score of overall pixels or the spatial percentage of the pixels with a score greater than and equal to the threshold of the chemical images, respectively. From this study, it can be concluded that, first, sampling method for data collection in mapping has to be optimized to achieve linear quantification calibration through simple univariate analysis approaches. Second, the ordinary way of evaluating/validating a linear quantification technique by best linear correlation coefficient ( R 2 ) and root‐mean‐square error of calibration is disputable and has to be reconsidered. Lastly, with further consideration of root‐mean‐square relative error of calibration and predicted crystallinity at subpercent, it was found that the spectral mean score method cannot generate reliable quantification results at subpercent crystallinity. Copyright © 2014 John Wiley & Sons, Ltd.