Spatial Quantification of Marine Litter Using Satellite and Drone Data through Empirical and Deep Learning Techniques – A Case Study from India

Abstract

Marine Litter is a major contaminant in the world's oceans. The mismanaged land-based garbage reaches the marine environment via rivers and creeks. Remote sensing techniques have the potential for detecting, classifying, and quantifying litter patches in the coastal zone. A case study for Chennai metropolitan city beaches attempted to quantify debris using different satellite sensors with specific spectral bands. Equally, the drone can detect and quantify macro litters (> 5 mm) at a relatively better resolution (2 to 2.5 cm). This study was carried out to evaluate the suitability of the drone images for detecting macro litter in the coastal segment. Image-based semantic segmentation deep learning techniques are applied. Estimating litter abundance with the spatial extent of natural and artificial litter on the beaches agrees with ground truth data.