Dransfeld, S., Tatnall, A.R.L., Robinson, I.S. and Mobley, C.D.
Retrieval of optically active parameters from oceanic reflectance spectra using multi-layer perceptron and K-Nearest neighbour algorithms
In Proceedings of Current Problems in Optics of Natural Waters.
Bedford Institute of Oceanography., .
Artificial radiance sets, generated with the use of a biooptical and radiative transfer model
Hydrolight, corresponding to Case 1 and Case 2 waters, are used as inputs to Multi-layer Perceptron
and K-NN algorithms to study the algorithm’s retrieval capabilities for optically active constituents
in the water. The radiative transfer model Hydrolight has been used to produce 18,000 artificial
reflectance spectra representing various Case 1 and Case 2 water conditions. The remote sensing
reflectances were generated at the MERIS wavebands 412, 442, 490, 510, 560, 620, 665 and 682
nm from randomly generated triplet combinations of phytoplankton, non-chlorophyllous particles
and CDOM concentrations. These reflectances were then used to assess the performance of the KNearest
Neighbour and the Multilayer Perceptron algorithms, which were compared to some more
traditional band ratio regression algorithms that had been a popular choice for CZCS and SeaWiFS
imagery. The objective of the work was to establish the best kind of algorithm for this type of
Actions (login required)