On-Gin Detection of Plastic Contaminants in Seed Cotton Using Hybrid Color and Laser-Scattering Images

Plastic contaminants contained in U.S. cotton can be in different materials, such as polyethylene (PE) in shopping bags and polypropylene (PP) in module wraps, in different colors (black, yellow, pink, brown, gray, white and clear), and in different sizes and shapes. To prevent these contaminants from entering the ginning process, it is important to detect and remove them prior to a gin stand. However, seed cottons contain other color objects (cotton burs, sticks, etc.) that could have similar color, size and shape as plastic contaminants but do not need to be removed at this stage. A sole color image analysis system cannot reliably discern plant particles from plastic particles in the cotton. with the support from USDA-ARS, we conducted a feasibility study to examine the potential for using laser-scattering patterns as a detection mechanism, and integrated the laser-scattering image system with the color image system USDA-ARS previously developed to detect both colored plastics and plant particles (cotton burs, sticks) using one camera system which is illuminated by four LED white lights and one green laserline simultaneously. In the field trials conducted in the USDA Cotton Gin Research Lab at Lubbock in Texas, we installed a prototype system on a gin stand to perform real-time, online inspection of cotton flows at speeds varying from 8 to 10 ft/s.  The experimental results showed that this hybrid image system could correctly detect 95% of color plastic contaminants and 50% of white and clear plastics, but generated false detections (overestimate) at a rate of ~0.06%. In the continuing development, we will focus on improving the efficacy of detecting white and clear plastics, and reducing the false detection rate by properly separating the white lighting and laser lighting in the camera enclosure.