The recognition of ocean red tide with hyper-spectral-image based on EMD

A new technique is introduced in this paper regarding red tide recognition with remotely sensed hyper-spectral images based on empirical mode decomposition （EMD）, from an artificial red tide experiment in the East China Sea in 2002. A set of characteristic parameters that describe absorbing crest and reflecting crest of the red tide and its recognition methods are put forward based on general pictre data, with which the spectral information of certain non-dominant alga species of a red tide occurrence is analyzed for establishing the foundation to estimate the species. Comparative experiments have proved that the method is effective. Meanwhile, the transitional area between red-tide zone and non-red-tide zone can be detected with the information of thickness of algae influence, with which a red tide can be forecast.