PREDICTION OF WET-MILLING STARCH YIELD FROM CORN BY NEAR-INFRARED SPECTROSCOPY

Near-infrared reflectance (NIR) spectroscopy was used to predict the w et-milling starch yield obtainable from yellow dent corn. Spectral dif ferences between samples with high and low starch yields were observed in two wavelength regions, 2,265-2,325 and 1,050-1,125 nm. The higher -wavelength region corresponds to known carbohydrate absorption bands; however, the origin of the shorter-wavelength differences is less cle ar. Different combinations of sample handling and spectral treatments were also evaluated, and the most successful instrument calibration us ed a multiterm linear equation of second-derivative reflectance terms obtained from whole-kernel corn. Multiple correlation coefficients for the various calibrations ranged from 0.8 to 0.9, with reproducibility of the laboratory wet-milling procedure being a limiting factor. Appl ication of the optimized calibration to a validation set containing sa mples from two crop years gave a bias-corrected standard error of pred iction equal to 1.41%, a result equal to or better than the standard e rror of the reference method.