Study Of Red Tide Airborne Hyperspectral Remote Sensing Technology

Supported by national advanced technology research and development projects and national key research fundamental planning projects, to meet the demand of China's red tide marine airborne monitoring operation, the paper probes into three techniques: red tide airborne hyperspectral detection, red tide dominant species identification and extraction of red tide biomass distribution feature. Meanwhile, it verifies the detection and identification models by making use of red tide airborne hyperspectral data and synchronized monitoring materials collected for the first time in China. The thesis is divided into the following parts:1. After analyzing differentiation between the hyperspectral characteristics of red tide water and normal water, two airborne hyperspectral remote sensing models for red tide detection are put forward on the basis of hyperspectral characteristic reflection peak and artificial neural network.2. By analyzing features of hyperspectral data of red tide water, its eigenvector is calculated and three kinds of red tide dominant species hyperspectral identification models are built: SAM, SCM and SVM.3. Based on the multicomponent sea water remote sensing reflectance model, an exploration is made into by bringing forth two models: forward model, i.e. to simulate sea water remote sensing reflectance according to chlorophyll concentration, and reversal model, i.e. to calculate chlorophyll concentration according to sea water remote sensing reflectance. 4. Red tide airborne hyperspectral remote sensing detection model and red tide dominant species airborne hyperspectral identification model are tentatively put into application. In terms of the former, spatial distribution of red tide water and normal water are detected in the hyperspectral images taken on 12th and 25th August, 2001 by means of red tide feature reflectance peak detection model and artificial neural network detection model. The paper also points out that the models in question are not applicable to the hyperspectral image with a large area of sea glint. In terms of the latter, SAM, SCM and SVM have succeeded in identifying dominant species of the red tide occurring on 19th and 25th August, 2001 as Leptocylindrus danicus Cleve and Mesodinium rubrum. But misidentification happens to Noctiluca red tide occurring on 16th July, 2001.In the end, the paper points out the future work concerning red tide airborne hyperspectral remote sensing detection, red tide dominant species airborne hyperspectral identification and extraction of red tide biomass distribution feature.