Skip to main content
ARS Home » Research » Publications at this Location » Publication #198775

Title: PREPROCESSING OF NIR DATA FOR STANDARDIZATION

Author
item Sohn, Mi Ryeong
item Barton Ii, Franklin
item Himmelsbach, David

Submitted to: Eastern Analytical Symposium
Publication Type: Abstract Only
Publication Acceptance Date: 6/19/2006
Publication Date: 11/1/2006
Citation: Sohn, M., Barton II, F.E., Himmelsbach, D.S. 2006. Preprocessing of nir data for standardization [abstract]. 45th Eastern Analytical Symposium. Paper # 709.

Interpretive Summary:

Technical Abstract: Calibration transfer has been an issue for a long time in spectroscopy application field and is still remained as an important subject. This paper describes a transfer of NIR calibration model between two instruments for determining fiber content in flax stem. The instruments are the same type from same manufacture but have different spectral responses. Calibration samples (n=199) were scanned on a master instrument and a calibration model was developed with preprocessed data. Independent prediction samples (n=34) were scanned on both master and slave instruments. To find the best standardization condition, we used two sets of transfer samples (15 sealed reference standard powders from FOSS and 5 samples selected from prediction set), two standardization procedures, and three correction methods (DS, PDS, DWPDS), and the results were compared. The two standardization procedures differ in their sequences; one is generating a correction file with the raw spectra of the transfer samples, converting the slave spectra, preprocessing with the same treatment as used for the calibration model, applying to the model, in order. The second procedure is preprocessing of the spectra of the transfer samples, generating a correction file, converting the slave spectra, applying to the model, in order. The best calibration transfer was conducted from the use of the subset samples as transfer samples, preprocessing prior to standardization of the transfer data, and DWPDS method.