Abstract
Oceanic eddy is a common natural phenomenon that has large influence on human activities, and the measurement and detection of offshore eddies are significant for oceanographic research. The previous classical detecting methods, such as the Okubo–Weiss algorithm (OW), vector geometry algorithm (VG), and winding angles algorithm (WA), not only depend on expert’s experiences to set an accurate threshold, but also need heavy calculations for large detection regions. Differently from the previous works, this paper proposes a deep eddy detection neural network with pixel segmentation skeleton on high frequency radar (HFR) data, namely, the deep eddy detection network (DEDNet). An offshore eddy detection dataset is firstly constructed, which has origins from the sea surface current data measured by two HFR systems on the South China Sea. Then, a spatial globally optimum and strong detail-distinguishing pixel segmentation network is presented to automatically detect and localize offshore eddies in a flow chart. An eddy detection network based on fully convolutional networks (FCN) is also presented for comparison with DEDNet. Experimental results show that DEDNet performs better than the FCN-based eddy detection network and is competitive with the classical statistics-based methods.
Highlights
Oceanic eddy is a common phenomenon of seawater flow, which plays an important role in transporting energy and matter
In the flow the flow chart, an eddy presents a shape of intertwined streamline, and its boundary is not constructed by a closed curve, which requires the model to have the capacity of distinguishing details
We present an fully convolutional networks (FCN)-based eddy detection network to compare with our deep eddy detection network (DEDNet)
Summary
Oceanic eddy is a common phenomenon of seawater flow, which plays an important role in transporting energy and matter. Submesoscale eddy detection from HFR current data with a deep learning network is addressed. From the physics-based methods, geometry-based methods extract the eddy’s shallow features through one transformation and determine whether an eddy exists in the observed region according to these shallow features. The DL method offers another approach for eddy detection, which is able to extract eddies’ deep features from multiple layers, recombine features across different layers, and simultaneously identify the centers and boundaries [14,15]. A deep eddy detection neural network for HFR observation data, namely. In the pyramid scene parsing network (PSPNet)-like skeleton, we present an efficient deep detection network for HFR observation data, which considers both the global regional feature and the detailed geometry feature of an eddy.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
