Development of an algorithm based on deep learning for the classification of oceanic geophysical phenomena.
- Paper number
IAC-22,B1,IP,27,x73644
- Author
Mr. Lucas Nicolas Taipe Ramos, Peru, Universidad Nacional de Ingenieria, Peru
- Coauthor
Mr. Sergio Sosa Callupe, Peru, Universidad Nacional de Ingeniería (Lima, Perù)
- Coauthor
Mr. Jesus Antonio Tapia Gallardo, Peru, Universidad Nacional de Ingenieria, Peru
- Coauthor
Mr. Omar Enrique Blas Morales, Peru, Universidad Nacional de Ingenieria, Peru
- Coauthor
Mr. Juan Salvador Palacios Bett, Peru, Universidad Nacional de Ingeniería (Lima, Perù)
- Coauthor
Mr. Nilton Cesar Rojas Vales, Peru, Universidad Nacional de Ingeniería (Lima, Perù)
- Coauthor
Mr. Jhon Rocha Calderon, Peru, Universidad Nacional de Ingeniería (Lima, Perù)
- Coauthor
Ms. Medaly Eulogio Saenz, Peru, Universidad Nacional de Ingenieria, Peru
- Year
2022
- Abstract
The main objectives of the Sentinel-1 mission are the observation of Land, Marine and Atmospheric Monitoring for emergency management, security and climate change. One of the many phenomena that the European satellite can detect are geophysical phenomena, which include oceanic and meteorological features, where the main area of study is the open ocean. The development of new detection methods for these events is essential, as most of the events have an important role in the climate system. For this reason, we propose a new convolutional neural network architecture, GeophysicalNet, with improved feature extraction to identify each of the ten classes of geophysical phenomena present in the TenGeoP-SARwv database. The GeophysicalNet neural network is composed of non-sequential layers, residual layers, parallel layers and fully connected output layers with dropout. The neural network was trained with a fraction of the TenGeoP-SARwv database and using data augmentation; we also trained several classifiers based on standard We obtained the performance metrics of the GeophysicalNet neural network and traditional classifier models. The GeophysicalNet neural network proved to have higher accuracy in different performance metrics when compared to other open-source neural network architectures. The proposed model can detect anomalies present in the ocean, with this open access information we have a better control of maritime spaces and thus avoid environmental damage or natural disasters.
- Abstract document
- Manuscript document
IAC-22,B1,IP,27,x73644.pdf (🔒 authorized access only).
To get the manuscript, please contact IAF Secretariat.