Abstract
An eddy is a circular current of water in the ocean that affects the fields of maritime transport, ocean data analysis, and so on. Traditional eddy detection methods are based on numerical simulation data and satellite images and their accuracy is affected greatly by manual threshold adjustment. In this paper, we present a new eddy detection approach via deep neural networks to improve eddy detection accuracy. First, we present a streampath-based approach to build a large-scale eddy image dataset from ocean current data and apply our dataset to eddy detection. Second, by combining the multilayer features in the neural network with the characteristics of the eddies, we achieve a competitive detection result with an mAP of 90.64% and an average SDR of 98.91%, which performs better than the previous methods. Third, through our enhanced eddy visualization approach, we solve the problem that eddies are difficult to detect in the sparse streampath region.
Highlights
The ocean is full of eddies with radii from a few kilometers to hundreds of kilometers
We have established a relatively more sufficient dataset, and our dataset will be published for future research
We present a deep learning-based method (SP-RCNN) for high-accuracy eddy detection, which shows a high accuracy with an mean average precision (mAP) of 90.64%
Summary
The ocean is full of eddies with radii from a few kilometers to hundreds of kilometers. Object detection approaches based on deep learning, such as faster region-based convolutional neural network (Faster RCNN) [6], you only look once (YOLO) [7] and single shot detection (SSD) [8], have achieved good results. These approaches can be applied to eddy detection applications, there is no satisfactory eddy image dataset. We propose an eddy dataset generation pipeline and a new approach to automatically detecting ocean eddies from flow field data, so-called streampath-based region-based convolutional neural networks (SP-RCNN).
Sign-in/Register to view highlights & summary 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 callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
