A A C C   2 0 0 0   A n n u a l   M e e t i n g

The objective of the study was to develop a near infrared reflectance calibration to be used for determining amylose content in maize kernels. The instruments used were the Foss NIRS Systems 5000 and 6500. Both instruments were operated using the reflectance module except that the 5000 used the small sample ring cups and the samples had to be ground. The 6500 used the rectangular samples holder on a transport mechanism. A total of 1045 corn samples with varying amylose contents were aqcuired from a Breeding/Research Center. These samples were scanned whole to collect spectral data using the NIRS Systems 6500. They were then ground using a Brinkman Retsch Grinder, Model ZM1 and scanned on the NIRS Systems 5000. Using complex mathematical computations to calculate for mahalanobis distance and sample centering, 107 unique samples were selected from the original population. This set of 107 samples, called calibration set, was presumed to be an accurate representative of the original population. Thus, the set can be used to develop NIR calibrations which can be applied to accurately predict amylose contents of the original population, as well as other corn samples that may spectrally be comparable with this population set. The amylose contents of the 107 calibration set were determined using a wet chemistry procedure. The procedure used was an unpublished proprietary protocol which mimics the Corn Refiners Association Standard Procedure for amylose content (CRA IB-6). The wet chemistry amylose data, which ranged from 34% to 89%, were matched with the spectral data to begin the process of developing the amylose calibration. During the calibration development three different strategies were used, namely Principal Component Analysis approach (PCA), Partial Least Squares (PLS) approach, and the Modified Partial Least Squares (MPLS). Three calibrations were generated accorning to each approach. The PCA calibration had following statistics: amylose mean was 62%, SECV was 6.3, 1-VR was 0.795, and R(^2) was 0.82. The statistics for the PLS and MPLS calibrations were similarly as following: Amylose means were 61.87% and 61.79%, SECVs were 5.29 and 4.80, 1-VRs were 0.86 and 0.88, and R(^2)s were 0.900 and 0.93, respectively. When the amylose contents of the original population were predicted using these calibrations the results were very encouraging. More work is being conducted to predict other samples and the results will be reported during the presentation. The conclusion of the study is expected to show that near infrared spectroscopy can be used to accurately predict amylose contents in corn/maize kernels as well as in corn starch.