SMU Data Science Review
Abstract
This study examines the use of a machine learning framework for predicting seafloor depth and coastline. The world’s oceans obscure the majority of the Earth’s surface and due to continual tidal motion, currents, and natural events, the seafloor changes constantly. Much of the world’s oceans remain unsurveyed by modern technologies or are under-surveyed using antiquated techniques. The increased availability and access of commercial imagery allows for the accurate prediction of bathymetric depths and the identification of coastline. DeepUNet is a sea-land segmentation deep learning model utilized to detect coastline. This study will modify the existing DeepUNet structure and preprocess the data using different techniques in attempt to increase accuracy of coastline detection. A RNN will then be utilized against preprocessed data in order to predict seafloor depths and then compared to an interpolated seafloor generated from nautical charting data. The results of this study indicate that derived bathymetry using deep learning techniques do not meet the IHO standards for inclusion in Safety of Navigation products. However, both tools allow the evaluation of areas in need of hydrographic surveying without the expenditure of expensive resources
Recommended Citation
Dickens, Kevin and Armstrong, Albert
(2019)
"Application of Machine Learning in Satellite Derived Bathymetry and Coastline Detection,"
SMU Data Science Review: Vol. 2:
No.
1, Article 4.
Available at:
https://scholar.smu.edu/datasciencereview/vol2/iss1/4
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License