Floating Litter Monitoring and Detection for Sustainable Marine Ecology

Marine litter poses a threat to health and balance to marine ecosystem. The adverse effect of litter on marine life includes entanglement of marine life, starvation, and habitat destruction. Cleansing floating litter offers a sustainable solution to protect and preserve marine life, ensures biodiversity, and supports the natural process that keeps the ecosystem thriving. In this paper, we propose a comprehensive framework for monitoring and detection of marine litter by integrating machine learning and computer vision techniques. The monitoring of litter is accomplished with drone-based surveillance system that recognizes litter with Histogram of Oriented Gradients (HOG) features. Mounting a Global Positioning System (GPS) on monitoring system renders approximate position of litter to the litter collection agents. The litter collection agents comprising detection systems that leverages benefits of transfer learning on YOLOv8 Algorithm. The detection system provides interface to collect floating litter safely avoiding marine life. The evaluations show that the proposed frameworks are reliable and have reasonable detection performance.