AI-Based Aquatic Plastic Waste Detection Using Underwater Image Processing and Deep Learning Algorithms

Abstract

Aquatic plastic waste has become a major environmental threat because plastic debris affects marine organisms, degrades water ecosystems, and reduces the ecological and economic value of aquatic environments. Conventional monitoring of underwater plastic pollution depends on manual visual inspection, diver surveys, net sampling, and image-based field observation, which are time-consuming, labor-intensive, and limited by water turbidity, illumination variation, occlusion, and large spatial coverage requirements. To address these limitations, this paper proposes an AI-based aquatic plastic waste detection framework using underwater image processing and deep learning algorithms. The proposed system combines underwater image acquisition, image enhancement, plastic object detection, semantic segmentation, and pollution severity assessment. Underwater images are first preprocessed using color correction, denoising, contrast enhancement, and normalization. A YOLO-based detection module identifies plastic objects such as bottles, bags, wrappers, fishing nets, and mixed debris. A U-Net segmentation module extracts pixel-level plastic regions for area estimation, while CNN-based feature learning supports classification of plastic and non-plastic objects. K-means clustering and Random Forest-based decision analysis are used to group monitoring zones and estimate pollution severity. The performance is evaluated using precision, recall, F1-score, intersection over union, mean average precision, and pollution severity classification. The proposed framework provides an efficient decision-support method for aquatic plastic pollution monitoring, underwater robotic surveys, environmental assessment, and sustainable cleanup planning.