Detection of floating marine plastic: Utilizing UAV remote sensing and object-based image analysis to derive an algorithm and test the accuracy against other models

Marine plastic is a prolific issue concerning the world’s oceans, yet methods of detecting and quantifying the true extent are not yet fully developed. However, remote sensing detection of plastic litter is currently in its foundation, although it is still necessary to realise the most effective process. Presently, regarding passive remote sensing, both satellite and airborne methods have been trialled. UAV remote sensing for plastic detection proves a cost-efficient method in which differing approaches of machine learning detection have been trialled. Of the methods, object-based image analysis (OBIA) has not yet been examined. Therefore, using eCognition software, UAV RGB imagery was tested using OBIA to test the accuracy of the results using sensitivity, positive predictive value and F-Score. Conversely, scoring 57.9%, 1.2% and 2.2% respectively, the method of OBIA is dissected for flaws in order to improve for the future, where larger sample sizes and multispectral imagery is suggested.