Application of UAV Remote Sensing for Detecting Plastic Contaminants in Cotton Fields

Plastic contamination of cotton is a serious problem for the U.S. cotton industry and abroad, and therefore it must be addressed to maintain quality of cotton fiber for its marketability and industry sustainability. Most plastic contamination comes from plastic wraps on round cotton modules, plastic mulch used in crops production, and plastic that blows onto cotton fields like shopping bags. The issue has become common enough that the USDA AMS cotton program began implementing a new extraneous matter code for plastic contamination in bales in 2018. Manual detection of plastic in cotton fields is tedious and labor-intensive. In this study, we conducted an experiment targeted at detecting plastic contamination from shopping bags in fields using an Unmanned Aerial Vehicle (UAV). Two field tests were conducted at two locations (Weslaco, TX and College Station, TX) and two stages of the growing season (before and after defoliation) by manually tying plastic bags at randomized locations and three different heights (bottom, middle and top). A five-band multispectral camera was mounted on the UAV to collect aerial imagery, and an image processing algorithm was developed to detect presence and locations of plastic contaminants in cotton field. Mahalanobis distance supervised classification was used on top of eight textural features in the direction of maximum obtained accuracy (90^o). Initial results have shown that plastic contaminants due to shopping bags can be remotely detected with a class accuracy of 64.17%, overall accuracy of 97.74% and kappa coefficient of 0.81.