Airborne Hyperspectral Red Tide Detection Method Based On Vegetation Index

Supported by a State Oceanic Administration project, to meet the demand of China's red tide marine airborne monitoring operation, the paper brings forward the conceptions of optimal bands combination and optimal vegetation index by embedding the vegetation index specific form into marine airborne hyperspectral remote sensing research realm, develops a model and method to detect red tide by airborne hyperspectral data, and carries out the detection model application using China's first batch of hyperspectral data. Main work in this paper includes:1. A new conception, significance, is brought forward to assess the image classification result, on the basis of which combination of the most sensitive bands and the most sensitive vegetation index for red tide detection arc determined. It's shown by the image classification results that significance is a reasonable and effective criterion for classification between two kinds of objects.2. A red tide detection model by airborne hyperspectral data is developed based on the combination of the most sensitive bands and several vegetation indexes, including ratio vegetation index (RVI), normalized difference vegetation index (NDVI), transformed vegetation index (NDVI) and difference vegetation index (DVI). Red tide and normal seawater can be classified by choosing threshold value of vegetation index image.3. The performance of red tide detection model by airborne hyperspectral data is assessed. It is manifested that detection results using images of four kinds of vegetation indexes are better than that based on image of single band. So it is feasible to detect red tide using vegetation index method when the sensitive bands combination is known.4. The influence of noise samples to the red tide detection result is assessed by mixing noise samples artificially when samples are chosen. It is shown that red tide detection model has the ability to resist noise to a certain extent, and the mix of noise samples does have impact to the red tide detection results.In the end, some research work that needs to be carried out in the future is pointed out as regards to the perfection of red tide detection model by airborne hyperspectral data.