Thursday 25 August 2005

G6

PTH0062
Depth estimation from hyperspectral imagery and robust techniques algorithms in coastal environments
Castillo, Elena¹, Bayarri Cayón, Vicente¹
1 University Of Cantabria, Spain
Author email: castille@unican.es
Accurate bathymetric algorithms from multispectral imagery have been searched for a long time and Principal Components Analysis (PCA) has been traditionally considered as a reliable method. This method implies an advance respect to other traditional methods such as spectral rationing since water's physical parameters are not needed. This article are shows the experience obtained after applying the new PCA robust method to CASI (Compact Airborne Spectrograpic Imagery) imagery with 36 spectral bands within 400 to 950 nm range. Two different correction levels have been considered aimed to assess whether the final product has been improved and two covariance estimator (volume minimum ellipsoide and minimum covariance determinant) has been used in the PCA robust method. On the other side a GPS-Echo sounder bifrecuency performing in RTK mode has been employed to calibrate and validate the data taking measures on site. The obtained components after applying classical and robust analysis have been calibrated using different nature fit models (lineal, least trimmed squared, Huber estimator) and validation processes will be shown.

Return to Poster Presentations