A technique is presented for rapidly determining the moisture content of hard red winter wheat, Triticum aestivum L., independent of grain bulk density. Complex admittance measurements from 1 to 110 MHz on bulk wheat samples were collected forpredicting moisture contents ranging from about 9 to 20, wet basis, in samples with densities ranging from 0.620 to 0.790 g/cm3. The data were transformed to density-independent function (DIE) values, and multivariate analysis was used to eliminatespectral and statistical outliers. Principal component analysis was also used to determine the optimum frequencies for predicting moisture content with the DIF values. Partial least-squares regression was then used to develop a calibration based onmeasurements at three frequencies of about 2, 25, and 80 MHz. Calibration was performed with 47 wheat samples representing six cultivars harvested in 1991 and 1992 with a standard error of cross validation of 0. 34 moisture content, and validation wastested on 146 different samples from the same cultivars. Predicted moisture contents compared well with values determined by air oven drying with a standard error of performance of 0.39 moisture content and a bias of +0.03.
展开▼
