Abessolo G. O., Almar Rafaël, Bonou F., Bergsma E.
Source
Journal of Coastal Research, 2020,
(No spécial 95), p. 1101-1105 ISSN 0749-0208
The accuracy of bathymetric measurements is crucial, especially to understand coastal processes. Video-based depth inversion methods have been widely developed in recent years, but they remain noisy, with typical errors due to the breakpoint optical and non-linear effects. Among the spectral and temporal approaches to video depth inversion, only the spectral approach applies an error criterion to identify erroneous data. Here, two error proxies are assimilated for the first time in the temporal approach, using a Kalman filter applied to 3.5 years (February 2013 to September 2016) of video images. Differences between filtered and unfiltered bathymetries were observed to be correlated with the proxies considered. A validation with field data on a 10-day experiment is performed between the original bathymetries and the filtered bathymetries. The results indicate that the mean square error can be reduced by at least 30%. Both proxies show good ability to correct depth estimates. Although the results are promising, validation of these approximations must be performed under various hydrodynamic and atmospheric conditions.
Plan de classement
Sciences fondamentales / Techniques d'analyse et de recherche [020]
;
Limnologie physique / Océanographie physique [032]
;
Informatique [122]
