Combining Spectral Approaches and AI for Marine Litter Detection and Identification.

This study presents marine litter detection using Artificial Intelligence (AI) from drone imagery next to the required spectral considerations for plastic identification in aquatic environments. An artificial accumulation zone of marine litter has been created over a lake in Mol, Belgium and different flight scenarios have been tested in order to calibrate required parameters like altitude, camera setting, resolution of image and also checking performance of a designed AI model. Using Cascade-RCNN for identification, Average Precision (AP) obtained at 75 percent for detection of plastic samples. At the next step this research, developed model will be applied over rivers of Vietnam for the plastic detection and identification.