Oceanic Mesoscale Eddy Detection Method Based on Deep Learning

Abstract

Oceanic mesoscale eddies greatly influence energy and matter transport and acoustic propagation. However, the traditional detection method for oceanic mesoscale eddies relies too much on the threshold value and has significant subjectivity. The existing machine learning methods are not mature or purposeful enough, as their train set lacks authority. In view of the above problems, this paper constructs a mesoscale eddy automatic identification and positioning network—OEDNet—based on an object detection network. Firstly, 2D image processing technology is used to enhance the data of a small number of accurate eddy samples annotated by marine experts to generate the train set. Then, the object detection model with a deep residual network, and a feature pyramid network as the main structure, is designed and optimized for small samples and complex regions in the mesoscale eddies of the ocean. Experimental results show that the model achieves better recognition compared to the traditional detection method and exhibits a good generalization ability in different sea areas.

Characteristics of Global Ocean Abnormal Mesoscale Eddies Derived From the Fusion of Sea Surface Height and Temperature Data by Deep Learning

“…Zhang et al, 2016). DL has been successfully applied to the studies of various ocean phenomena, such as internal waves, ocean fronts, and mesoscale eddies (Duo et al, 2019;Lguensat et al, 2018;Li et al, 2020;X. Sun et al, 2019;Xu, Cheng, et al, 2019;Zheng et al, 2020), and thus demonstrate the feasibility of extracting abnormal eddies based on DL.…”

Deep Regression Network-Assisted Efficient Streamline Generation Method

“…Bai et al [28] proposed a streampath-based region-based convolutional neural network(SP-RCNN) that detects ocean eddies from streampath images. Duo et al [29] presented a deep learning approach based on a deep residual network (ResNet) [4] and a feature pyramid network (FPN) [30] to detect oceanic mesoscale eddies.…”

DEDNet: Offshore Eddy Detection and Location with HF Radar by Deep Learning

“…Limited by the VG algorithm, the accuracy remains on a similar level as the geometry-based algorithm. Du and Wang proposed an eddy identification and tracking framework mainly based on feature learning with convolutional neural network and using the SLA data of Australia [ 29 ]. As a conclusion, the performance of the objective detection branch is limited by the VG algorithm, since it relies on the VG algorithm to determine the eddy’s center and boundary.…”

“…Wherein, 5000 flow charts are processed as the training database, and the other 760 flow charts are used as the testing database. Following the previous work [ 29 ], we also adopt python-eddy-tracker software (PET14) [ 33 ] outputs as the training database for our eddy detection algorithm. To extract the most valuable information from the flow chart, the input image is set to 832 × 576 pixels.…”