Continued work to develop a low-cost sensor to detect plastic contamination in seed cotton at the gin

Authors: Whitelock, Derek; Armijo, Carlos; Delwiche, Stephen - Steve; Kim, Moon; Hughs, Sidney

Submitted to: National Cotton Council Beltwide Cotton Conference
Publication Type: Proceedings
Publication Acceptance Date: 2/7/2017
Publication Date: 6/1/2017
Citation: Whitelock, D.P., Armijo, C.B., Delwiche, S.R., Kim, M.S., Hughs, S.E. 2017. Continued work to develop a low-cost sensor to detect plastic contamination in seed cotton at the gin. National Cotton Council Beltwide Cotton Conference, January 4-6, 2017, Dallas, TX. p. 389-397.

Interpretive Summary: Contamination of cotton from plastic trash that collects in cotton fields and other man-made “trash” introduced during harvesting and ginning is one of the most significant threats to U.S. cotton world market share. Mixed with cotton lint, contaminants slow spinning mill capacity, increase mill waste, and render finished yarns and garments unmarketable. In the past, U.S. cotton has enjoyed its status as “contamination-free”, but recently, some foreign mills have considered purchasing cotton from countries other than the US due to the perception that contamination levels in U.S. cotton have risen. Lost markets tied to contamination are difficult to recapture and the textile industry has a long memory. The first line of defense against these contaminants should be in the field before the raw cotton can be introduced into the gin machinery. If contaminants enter the gin, they should be eliminated early in the process. However, current gin machinery does not effectively remove contamination from the cotton stream. Research was conducted to develop an inexpensive and reliable technology to detect cotton contamination at the cotton gin and separate the detected contamination from the cotton flow. Hyperspectral imaging was used to measure ultraviolet fluorescence and visible/near-infrared/short-wave infrared reflectance of different contamination samples with a seed cotton background. Data analyses showed differences in spectral response that could be exploited. Specifically, round module wrap, module tarp, and bale twine exhibit distinct differences from seed cotton in the visible/near-infrared spectra due mainly to visible color. Also, greases and oils have very different ultraviolet fluorescence responses than seed cotton. These results will lead to work to better quantify the differences between seed cotton and the contaminants, and allow future work to design, construct, and test a prototype sensor to detect contaminants in a bat of seed cotton. For U.S. cotton to maintain its status as “contamination-free” and continue relationships with foreign mills, the industry must strive to prevent contaminants from entering the cotton stream and to eliminate them when they do slip in.

Technical Abstract: Contamination of cotton from plastic trash that collects in cotton fields or introduced at the gin or due to mishandling at the warehouse is one of the most significant threats to US cotton world market share. For U.S. cotton to maintain its status as “contamination-free”, the industry must strive to prevent contaminants from entering the cotton stream and to eliminate them when they do slip in. The main objective of this research was to develop an inexpensive and reliable technology to detect cotton contamination at the cotton gin. First, hyperspectral scans of typical cotton contaminants were taken using a spectrophotometer that allowed for scans of contaminants with a seed cotton background. The contaminant samples included conventional module tarps, bale bagging, yellow round module wrap, hay baling twine and netting, plastic mulch, shopping bags, bale strapping, baggies, burlap, grease, and oil. Using spectral bands from the scans that showed distinct differences between cotton and contaminants, statistical analyses correctly classified all of the cotton and contaminant samples as Cotton or Not. Hyperspectral imaging was then conducted using two systems that measured the ultraviolet fluorescence, and the visible, near-infrared, and short-wave infrared reflectance of the contaminants. Analyses of the images indicated that visible reflectance may be effective in detecting colored plastics in seed cotton. However, detecting contaminants that are white or opaque will likely require using reflectance measured at longer wavelengths. Also, there were dramatic differences in ultraviolet fluorescence between oil and grease, and seed cotton. These results will lead to design, construction, and testing of a prototype sensor that utilizes ultraviolet light and visible color to detect the contaminants in a bat of seed cotton.