An ocean red tide monitoring method of the aerial remote sensing hyper-spectral image

The paper presents a method which uses hyper-spectral image data of different familiar dominant species to train different neural networks, then synthesizes the outputs of the networks with the same weight to recognize the red tide. It not only conquers difficulties such as the selection of training data and a network´s training method, but also improves the generalization ability of the network system effectively. A mass of comparison experiments prove that the method recognizes the red tide and the dominant species effectively. Furthermore, it distinguishes the transitional water area of the red tide using the algae´s intensity information, which enables forecasting of the red tide.