Classification of cotton foreign matter using color features extracted from fluorescent images

Abstract

The presence of foreign matter affects the quality and ultimately the monetary value of cotton lint. Current methods, such as the High Volume Instrument, compute the overall area of the foreign matter, but cannot identify the specific type. A fluorescent imaging system was investigated to classify the six types of botanical and seven types of non-botanical foreign matter under blue and ultraviolet (UV) light-emitting diode (LED) excitation lights, respectively. Two color models (RGB (red, green, blue) and HSV (hue, saturation, value)) were used and ratio images and single-channel images were examined. The F-values from the multivariate analysis of variance were used to select the three most contributing color features from the blue LED (R/G, H, V) and the UV LED dataset (B/G, S, V). The linear discriminant analysis model achieved classification rates of 80% or higher for bract, green leaf, hull, paper, plastic bag, plastic packaging, seed, and stem, and classification rates between 60% and 80% for bark, seed coat, and twine. The study demonstrated that fluorescence imaging is a promising tool to classify major types of cotton foreign matter and could be used for cotton classing.