Detection of synthetic foreign matter in cotton lint by shortwave infrared hyperspectral imaging

Abstract. The presence of various synthetic foreign matter in cotton lint seriously degrades the commercial value of cotton lint and further reduces the quality of textile products for consumers. This research was aimed to investigate the potential of the liquid crystal tunable filter near infrared hyperspectral imaging technique for inspection of synthetic foreign matter on the cotton lint surface. The synthetic foreign matter studied in this paper included module cover, plastic mulch, drip irrigation belt, plastic bags in gray, red, brown, and green, as well as the poly woven bag. Hyperspectral images of these foreign matter on top of the lint surface were acquired from 480 samples (60 for each type of synthetic trash) using the in-house built near infrared hyperspecrtal imaging system with wavelength range from 900 to 1700nm. The spectra of the foreign matter samples were extracted from the regions of interest manually and the characteristics were determined. Hierarchical cluster analysis was used to congregate various foreign matter and cotton lint based on their spectral characteristics. According to the results of the hierarchical cluster analysis, the various foreign matter and cotton lint samples were defined into different groups ranged from two to nine. Finally, linear discriminate analysis was utilized to differentiate various foreign matter and cotton lint under differently defined groupings. The total misclassification rate was 0.78% in Leave-one-out cross-validation under the nine defined groups. The results indicated the misclassification rate of the synthetic foreign matter and cotton lint samples was reduced with the increase of the defined groups. The results showed a great potential of the shortwave hyperspecrtal imaging technique in detection of various synthetic foreign matter on top of the lint surface.