Waste Bottle Detection by Autonomous Surface Vehicles in Inland Water Bodies Using Light YOLOv9 Model

Plastic pollution in aquatic environments, particularly in lakes, is a critical environmental issue that poses significant threats to aquatic life and water quality. The accumulation of PET bottles on lake surfaces necessitates efficient detectionand removal strategies. This paper presents a novel applicationof the YOLO (You Only Look Once) algorithms for the detection of PET bottles on lake surfaces. Leveraging advanced computer vision techniques, challenges such as varying bottle orientations, colors, and complex backgrounds are addressed. The approach involves stages such as edge detection, threshold-based filtering, and post-enhancement techniques to improve detection accuracy and robustness. This research contributes significantly to real-time environmental monitoring by providing a scalable and efficient solution for detecting and mapping PET bottles. The integration of this technology with Autonomous Surface Vehicles (ASVs) offers a promising avenue for automated plastic pollution management in aquatic ecosystems. The findings underscore the potential of advanced computer vision techniques with YOLOv9c model in detecting PET bottles with outperforming metrics having map@50=89% and F-score=88.5%.