Recognition of chlorophyll rings using YOLOv8

Chlorophyll acts as a critical indicator of marine phytoplankton and primary productivity, and its concentration distribution holds great significance for marine fisheries. With the rapid advancement of deep learning technologies, the application of neural networks in ocean science has become increasingly prevalent. This study employs the YOLOv8 model to identify elevated chlorophyll concentrations at the peripheries of mesoscale ocean eddies, known as chlorophyll rings. The YOLOv8 model demonstrates remarkable generalization capability and high accuracy in this application, surpassing other models such as Swin-Transformer and ResNet in both the quantity and precision. This paper conducts a statistical analysis based on the recognition results, revealing the spatial distribution characteristics of chlorophyll rings and examining the generation quantities of these rings under varying eddy radii and life cycles. The findings not only provide a robust tool for the analysis of chlorophyll concentration data but also yield new insights into the distribution and dynamic changes of chlorophyll within marine ecosystems.