An Automatic Marine Mesoscale Eddy Detection Model Based On Deep Learning

The ocean is an important material foundation for the sustainable development of mankind.Relying on Marine remote sensing technology and information technology to realize the dynamic monitoring of Marine phenomena and constantly deepen the comprehensive understanding of the ocean,only in a reasonable control and efficient development of Marine resources.Marine mesoscale eddies are an important ocean phenomenon characterized by closed circulation and widely exist in the global oceans.Marine mesoscale eddies,which carry enormous energy,can significantly change the vertical distribution of nutrients and thermocline,and play a very important role in the distribution of plankton,energy and salt transport.The global climate and ecosystem are deeply affected by the change of ocean material and energy.Therefore,the automatic monitoring of marine mesoscale eddies is not only helpful to the study of ocean climate change,but also plays an important role in the effective development and management of marine resources and ensuring the safety of marine environment.Automatic detection of marine mesoscale eddies is an important method for monitoring and analyzing spatiotemporal variation of mesoscale vortexes.Marine remote sensing satellite has the characteristics of all-weather,large area,long distance,non-contact,fast and efficient in the observation of ocean phenomena.This observation method provides abundant data resources for the study of marine mesoscale eddies.Aiming at realizing automatic and accurate detection of marine mesoscale eddy,this paper proposes an automatic detection model suitable for marine mesoscale eddies according to the limitations of existing methods and the idea of deep learning.The main research contents are as follows:(1)An automatic marine mesoscale eddy detection model based on deep learningThe key to automatic and accurate identification of marine mesoscale eddies is to extract their high-level essential features.However,the complexity of marine mesoscale eddies aggravates the difficulty of feature extraction.Traditional methods cannot accurately express the characteristics of marine mesoscale eddies because they rely on artificial design features and expert threshold values,which seriously affects the accuracy of their identification.Aiming at the above problems,this paper constructs an automatic detection model suitable for marine mesoscale eddies,and realizes the stepby-step abstraction and expression of the essential features of the upper layer of marine mesoscale eddies,thus realizing its automatic and accurate detection.(2)Empirical analysis of marine mesoscale eddies automatic identification modelBased on the remote sensing data set of sea surface height anomaly constructed in this paper,the marine mesoscale eddies in the South China Sea area are analyzed empirically.Based on the model proposed in this paper,the extraction results of marine mesoscale eddies in the South China Sea were compared with other models,and the segmentation accuracy,attenuation rate of segmentation accuracy,visual segmentation effect and other aspects were analyzed comprehensively to prove the effectiveness of the model proposed in this paper.Finally,the marine mesoscale eddies in the South China Sea were extracted from the model in this paper,and the spatio-temporal variation characteristics of the extracted results were analyzed in detail.(3)The design and application of the system module for diagnosing mesoscale phenomena in the oceanBased on the research of marine mesoscale eddies automatic detection,design and build the concentration scale data management the mesoscale diagnosis,sharing,integration of mesoscale phenomenon test evaluation in the integration of mesoscale diagnostic system platform architecture,research and development oriented diagnosis analysis of the phenomenon of mesoscale information service prototype system of the business,build theory application sharing and knowledge service are the demonstration of the application.