Mapping beach plastics using drone-based (UAV) imaging systems, in conjunction with East Lothian Council

Beached plastic and litter are the most common components of marine waste, with wet wipes comprising of increasing concern. In automating the process of detection, unmanned aerial vehicles (UAVs) prove an efficient method for mapping beached litter. UAVs paired with multispectral sensors are an accessible option for councils for the semi-automation of mapping beached plastic litter. The method of classifying plastic on a coastal section of Tyninghame beach was evaluated through simple and complex classes. Detection methods of pixel- and object-based classification methods, using the supervised classification method of support vector machine (SVM) were compared across the UAV surveys. It was found that segmentation and training samples do not transfer across spatial resolutions, as their performance is variable. Additionally, object-based image analysis (OBIA) proves highly effective at 89-98 % overall accuracy, compared to the pixel-based image analysis (PBIA) across the UAV surveys conducted. This research demonstrates OBIA and SVM as an effective combination for mapping and detecting beach plastics for the beach environment in East Lothian, Scotland for the given plastic litter samples tested. Future suggestions include the further discrimination of the litter classes to improve the accuracy of detecting specific plastic litter categories, through multispectral UAV surveys.