Potential of Classifying Cotton Minicard Stickiness through Vis-NIR Spectroscopy as an Analytical Technique with DD-SIMCA as One-Class Classification.

Cotton stickiness, mostly resulting from honeydew depositions of whiteflies and aphids, presents a worldwide problem for cotton growers and processors consistently. To meet the challenge of measuring the cotton stickiness, a few direct and indirect techniques exist. Previous study showed that Fourier transform near-infrared (FT-NIR) spectroscopy can be used to detect Minicard stickiness in raw cotton from partial least-squares (PLS) analysis. In the present investigation, visible-NIR (vis-NIR) as an analytical technique was explored for potential classification of four-class Minicard cotton stickiness, in combination mainly with the data-driven version of soft independent modeling of class analogy (DD-SIMCA) as one-class classification. Both PLS prediction-based classification and DD-SIMCA models in different spectral regions were developed to optimize the identification efficiency. Compared to an optimal PLS prediction-based classification model indicating a four-class correct classification of 77.8% in the calibration set and 69.2% in the validation set from the 750-1850 nm NIR region, an optimal DD-SIMCA model from the same spectral region could reach an improved discrimination of >95.0%, with a 98.1% correct identification in the calibration set and a 96.2% success in the validation set. This observation emphasized that vis-NIR spectroscopy with an DD-SIMCA approach could be a rapid and nondestructive tool for screening the Minicard stickiness in cottons.