Detection of macroplastic on beaches using drones and object-based image analysis

Abstract

This pilot study demonstrates UAV-based image collection being a promising technology for plastic detection in beach zones and we recommend object-based image analysis and UAV-based data collection being further developed for marine litter detection. Objects of plastic at sizes down to 3.5 x 3.5 cm were segmented and classified successfully corresponding to approximately 4 times the pixel width of the classified images. Overall conclusions should be approached with some caution, due to the relatively few ground truth data points (13-37 geolocated plastic pieces) and their different characteristics regarding color and size.