A Geographic Segmentation Appoach for Sattellite - Derived Bathymetry

• Ano de defesa: 2022
• Autor(a) principal: Affonso, Juliane Jussara
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: University of New Hampshire (UNH)
• Palavras-chave em Português: SDB algorithm; Spatial filtering; Sensoriamento remoto
• Link de acesso: https://www.repositorio.mar.mil.br/handle/ripcmb/846486

Safety of navigation depends on our knowledge of seabed and its features, and, as such, any improvements in deriving bathymetry for nautical chart updating are of major importance. Satellite-Derived Bathymetry (SDB) is an alternative to traditional surveys using ship and airborne sensors, particularly for mapping remote and shallow areas, due to its reduced cost and the absence of navigational risks in very shallow and unsurveyed areas. However, the accuracy of SDB can be judged as relatively low for nautical charting purposes and, therefore, is mostly used for reconnaissance or/and for filling gaps in remote or very shallow areas. One of the reasons may be that the conventional approaches assume that bottom type and water clarity are constant and negligible within the entire image, and consequently, a single (global) and linear model is performed to retrieve bathymetric information. To address the spatial heterogeneity within a scene and with the aim to increase the accuracy and coverage of estimated depths, this work investigates the segmentation of the scene, both horizontally and vertically, into smaller spatial units, and accounts for water column parameters in the SDB equation. In practice, the main idea of the segmentation is to divide the image scene into small spatial units and then calibrate the model within each segment. The individual models use the same algorithm but varying model parameters from place to place. Also, to account for water column and sea bottom variations, an extended Dierssen model is applied. The performance of the methods is evaluated in two study areas in the Dry Tortugas, Florida, and St. Thomas East and Reserve, U.S. Virgin Islands. Overall, the results indicate that the accuracy of bathymetry may be improved when the area is divided into smaller spatial units, particularly with a vertical (by depth) segmentation of the scene. In detail, compared to the conventional global and linear approach, the accuracy in both study areas is increased by over 40% with segmenting the area and calibrating the water parameters within each spatial unit. Furthermore, as it is demonstrated with the two study areas, besides the improvements in the depth accuracy, the SDB coverage is increased with the extraction of bathymetry beyond the depth considered as the effective optical depth of the conventional global and linear approach. However, further work is recommended to investigate and verify the accuracy improvement demonstrated by the vertical segmentation and particularly that of the smallest utilized depth range of 1m. since questions are raised about the discontinuity of the models and their quantized depth predictions, and more precisely whether this is due to overfitting rather than an actual improvement in accuracy. Lastly, the results demonstrate that considering the water column and sea bottom heterogeneity for solving the global SDB model increases the accuracy of bathymetry estimates. Nonetheless, when the area is segmented into small spatial units, adding the water column contribution as a parameter to the equation did not produce a significant contribution.