Nondestructive Determination Of Nutritional Quality In Intact Cottonseed Using Near Infrared Spectroscopy

Cotton is an important fiber crop in the world, and also the second important potential source for plant protein after soybean and the fifth oil-bearing crop. Also a large quantity of seeds is produced, i.e. with every1kg of fiber produced by the cotton plant, about1.65kg of seeds are produced. Usually, the content of protein, oil and amino acid is performed according to conventional wet chemical methods, which offer a high level of accuracy and sensitivity but are expensive, relatively labor and time consuming. The objective of this work was to establish accurate and stable calibration models based on NIR spectroscopy with suitable regression methods and variable selections, for simultaneously determining protein, oil and amino acid contents of shell-intact cottonseed. The main results and conclusions were listed as follows:(1) Total385samples of cottonseed were used in this experiment. Field trials were conducted in different cultivated environments in2008and2009. The protein, oil and amino acid contents of intact cottonseed were determined by the reference method, respectively. Standard normal variate (SNV) and Savitzky-Golay (SG) derivate were applied for spectra preprocessing. As variable selection technique, the Monte Carlo uninformative variable elimination (MC-UVE) method and successive projections algorithm (SPA) were presented in multivariate calibration. In addition, this paper presented an optimization approach for least-squares support vector machine (LS-SVM) parameters by genetic algorithms (GA). Compared with the optimal partial least squares (PLS) and LS-SVM models both with full-spectrum data, and MC-UVE-PLS models, the prediction performance of MC-UVE-LS-SVM models was validated to be much better.(2) The coefficient of determination (R2), residual predictive deviation (RPD) were0.959and4.871for protein,0.950and4.429for oil, the results indicated that the calibration models of protein and oil contents were achieved with good accuracy and robustness (R2>0.900, RPD>3.000). The optimal models of amino acids were achieved with R2from0.596to0.949and RPD from1.573to4.311, respectively. All the models except for Cys, Tyr, Ser, Met and Thr, were well-developed in the terms of the higher of R2and RPD. These results showed that it was possible to build robust and nondestructive models to quantify protein, oil and most amino acids contents in intact cottonseed using near infrared (NIR) spectroscopy.