Research on Underwater Object Classification Detection Based on EN-YOLOv8 Model

Image detection and recognition plays an important role in water environmental monitoring and management. By accurately identifying underwater objects, engineers are able to better assess the health of water bodies, monitor changes in ecosystems, reduce pollution risks, and ensure the safety of water facilities. The improved model YOLOv8 proposed in this paper performs instance segmentation tasks on the Trash Can dataset, and adopts a number of innovative techniques to improve model performance. The model realizes information aggregation of different spatial dimensions through improved EMA, which enhances the expression ability of features. The NWD loss function is specifically used to measure the similarity between small objects, and the model improves the accuracy of small objects by 2.4%. The integration of these methods significantly improves the detection effect of EN-YOLOv8, especially in the accurate identification of Marine debris. Experiments show that the EMA-NWD-YOLOv8 (EN-YOLOv8) model studied in this paper not only improves the detection efficiency and accuracy, but also provides strong support for ecological protection and scientific decision-making by accurately distinguishing underwater garbage, which has significant social and environmental value.