Measurement Of Wheat Quality Traits By Near Infrared Transmittance Spectroscopy

Quality breeding is highly dependent upon the quality testing which should be characterized with rapid, accuracy, convenient and economic operation, and wide application. Near infrared transmittance spectroscopy (NITS) adopted by this study shows such advantages as rapid measurement, simple operation and several ingredients testing simultaneously. The objectives of this study is to investigate the possibility of using near infrared spectroscopy for rapid wheat quality testing In total, 426 wheat samples from the major wheat areas collected in 2001 and 2002 were used to measure moisture content, protein content, wet gluten content, dry gluten content, ash content, SDS sedimentation value, Zeleny sedimentation value, Farinograph parameter, and Extensograph parameter by NITS. Calibration models were established with data from chemical analysis and absorbed spectrum of calibration sets and then wheat moisture content, protein content, wet-gluten content, dry-gluten content, ash content, SDS and Zeleny sedimentation with satisfied results were obtained with higher coefficients of determination and lower SEC. r2 and SEC of their calibration sets and prediction sets was between 0.71-0.98 and 0.15~6.73, respectively. However, r2 of Farinograph parameter, and Extensograph parameter was relatively lower accompaning with lower SEC. Meanwhile, a set of representative individual wheat samples were used to predict the model and results showed that it was feasible to determinate wheat quality trait by NITS and could be used for early generation in wheat breeding program.Kernel hardness kernel hardness is an important criterion for wheat quality classification and marketing. It affects milling and processing quality through effecting on milling yield, flour granule size, water addition of tempering and starch damage. However, traditional methods for hardness testing was unacceptable in early generation, and NITScould provide a satisfying way. The additional objective of the study was to seek a better algorithm and way of spectra treatment for calibration and upset non-linearity. In experiment II, 583 wheat samples from major wheat area collected in 2001 and 2002 seasons were used to measure kernel hardness by NITS. Two types of algorithms including partial least squares and multiple linear regression and three treatments including log1/T, first derivative of log1/T and second derivative of Log1/T were compared by hardness model. Results showed that partial least squares was better than multiple linear regression and treatment of first derivative of logl/T is mostly suitable in all three treatments and its r2 of calibration sets and prediction sets was 0.81, 0.75, with SEC 11.47, 11.86, respectively. In addition, when calibration model by first derivative of logl/T and partial least squares was used, acceptable classification accuracies was achieved, 90% for hard type, 83% for soft type, and 63% for mixed type, respectively. A set of representative individual wheat samples were used to predict the model and results showed that it was feasible to determinate wheat hardness by NITS and could be used for early generation selection in wheat breeding program.