海洋中尺度涡（10-100km）是活跃在大尺度背景环流之上的短暂而强烈的海洋信号，对海洋质量、动量、能量输送具有重要作用。但其检测与识别问题作为研究其动力学机制与生消规律的基础仍是一个瓶颈。虽然现有方法能够对简单情况下的中尺度涡进行基本提取，但对于复杂海流背景或复杂场景下产生的中尺度涡仍然很困难，尤其是对在碰撞、融合、分离过程中产生的不规则形状的中尺度涡提取更加困难。本项目针对中尺度涡检测与识别的关键问题展开研究：一是与背景环流的分离；二是不规则形状中尺度涡的识别，建立一个融合海流场分解理论与深度学习的中尺度涡检测与识别新方法。首先，利用海流场分解理论把任意复杂的海流场分解成包含中尺度涡的旋转部分和不包含中尺度涡的散度部分以便于中尺度涡的检测，然后利用深度学习算法对不规则形状的中尺度涡进行准确识别。本研究将利用AVISO卫星高度计数据和高分辨率HYCOM数据在西北太平洋海域进行实验验证。

Mesoscale eddies, 10-100km in space, are strong but short-life movement relative to the large-scale ocean circulations. They play significant roles on the transportation of ocean water masses, momentum and energy. However, the detection and recognition of mesoscale eddies, which is the foundation of exploring their dynamical mechanisms, is still a bottleneck for oceanographers. Although existing works can roughly extract mesoscale eddies from simple background currents, they fail at many complex cases especially with turbulent ocean currents and during the collision, fusion or separation of mesoscale eddies. The key of this problem is twofold: the separation of eddies from background currents and the recognition of non-regular eddies. In order for accurate and robust extraction of mesoscale eddies, we develop a novel mesoscale eddy detection and recognition framework based on the fusion of ocean current decomposition and deep learning. First, ocean current decomposition is designed for the segmentation of eddies from background surface currents. Then a sophisticated deep learning algorithm is developed for the robust recognition of mesoscale eddies of variant shapes and scales. Finally, we use AVISO satellite altimeter data and HYCOM data for the validation of our method in the Northwestern Pacific Ocean.