AI-Driven Identification of Underwater Vegetation Species Using YOLOv8 and OpenCV

This paper presents work identifying underwater vegetation species using computer vision techniques. Underwater videos were recorded in a challenging dataset that captures the Chilika and Pulicat Lake environment, characterised by turbidity and varying light conditions. OpenCV processed the videos for frame bifurcation and enhancement. An object detection model, YOLOv8, has been used to detect and classify different seagrass species that have been positively identified, thereby proving the model in underwater environments. The results have demonstrated that deep learning, when integrated with image processing can achieve environmental challenges in underwater monitoring. The model's accuracy is tested based on the usual performance evaluation metrics for detecting precision, recall, and detection rates. The proposed method showed high detection capability. In the last section of the paper, an h-index for Shannon diversity is computed. This gives some insight into the study's impact on conditions for environmental conservation. Results reiterate that AI approaches should be integrated into research on aquatic ecosystems, which supports decision-making for environmental management and data on conservation activities.