Analysis of Underwater Coral Reef Health Using Neural Networks

Safeguarding coral reefs requires extensive human effort. Divers, acting as underwater doctors, visually inspect the reefs, demanding expertise and significant resources. Digital methods of coral inspection include taking sequential images or video footage and using the technique of photogrammetry to create benthic maps for expert analysis. While improving scalability, this approach still relies on human interpretation, creating a bottleneck. Advancements in AI, particularly Convolutional Neural Networks (CNNs), offer exciting possibilities. These AI models, deployed on underwater vehicles like ROVs and AUVs, can auto-mate coral detection, classification, and segmentation. However, onboard resource constraints necessitate model optimization for computational efficiency and power consumption. This study showcases the potential of deep neural networks by implementing the YOLOv8 CNN model for automated coral health assessment. The model classifies coral into three cat-egories: healthy, partially bleached, and completely bleached, achieving an accuracy of 78% in validation data. This approach eliminates the need for diversification, significantly enhancing scalability and cost-effectiveness. Moreover, deploying the model on underwater vehicles enables real-time monitoring and inter-vention, potentially mitigating damage before it's too late. While promising, further research is crucial. Optimizing the model for even higher accuracy and resource efficiency is essential. Additionally, integrating data from other sensors, such as sonar or LiDAR, could provide a more comprehensive picture of reef health. Ultimately, real-world deployment and validation across diverse coral reef environments will solidify the practical value of AI-powered solutions.