Skip to main content
ARS Home » Research » Publications at this Location » Publication #104794

Title: USE OF SPECTRAL WINDOW PREPROCESSING FOR SELECTING NEAR-INFRARED REFLECTANCE WAVELENGTHS FOR DETERMINATION OF THE DEGREE OF ENZYMATIC RETTING OF INTACT FLAX STEMS

Author
item Archibald, Douglas
item Akin, Danny

Submitted to: Vibrational Spectroscopy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/1/1999
Publication Date: N/A
Citation: N/A

Interpretive Summary: Fiber flax production in the United States has declined to near zero, despite the facts that flax is readily produced in many parts of the country and the U.S. public is the worlds largest consumer of linen. Retting is a critical step in the conversion of flax stems to flax fiber, and current research on enzymatic retting and flax processing is poised to produce a system that is more efficient and more compatible with the U.S. textile processing infrastructure. Part of that effort is the development of objective methods for determining whether the flax has been sufficiently retted. The study outlines an instrumental optical approach that simultaneously measures structural and compositional characteristics of the flax stems, while requiring very little sample preparation. Furthermore, the method was made to be resistant to the kinds of variations likely to be produced by real measurement conditions. The focus is on determination of sets of wavelengths, because this allows low-cost instruments to be developed. A novel approach was used to solve this technical problem, and it has use beyond this particular application.

Technical Abstract: A near-infrared reflectance method suitable for air-dried intact flax stems was developed for determination of Fried's test degree-of-retting scores. Performance statistics for preferred wavelength sets from 2 to 12 wavelengths are reported for the spectral range from 1432 to 2468 nm. Spectral properties of enzyme-retted flax specimens were measured under varying conditions of hydration, stem orientation, and optical geometry. Calibration models were constructed to be insensitive to these effects. For the 12-wavelength model, the correlation (R**2) between predicted and measured is 0.946 and root mean-squared (RMS) cross-validation error is 0.20 for smoothed Fried's test scores spanning the four visual score levels, 0 to 3, from under- to over-retted. This is better than the RMS repeatability error of the Fried's test (0.25). Furthermore, even the smallest wavelength set has excellent success in classifying enzyme-retted flax stems as either under- or over-retted. Wavelength sets were produced by a novel three-stage procedure. In stage one, five locally optimal spectral windows were determined by a computational/visualization procedure that evaluates many spectral window widths and positions. In stage two the five models were converted to minimal-sized wavelength sets, and in stage three the resulting list of key wavelengths was pared down to smaller recommended sets by cross-validated multiple linear regression testing of all-possible combinations of sets or individual wavelengths. This procedure was able to find wavelength models that substantially outperform full-spectral models of similar dimensionality, and furthermore the method facilitated observation of the most relevant spectral variations.