As an effective survey tool for oil spill detection, the airborne hyper-spectral sensor affords the potentiality for retrieving the quantitative information of oil slick which is useful for the cleanup of spilled oil. But many airborne hyper-spectral images are affected by sun glitter which distorts radiance values and spectral ratios used for oil slick detection. In 2005, there's an oil spill event leaking at oil drilling platform in The South China Sea, and an AISA+ airborne hyper-spectral image recorded this event will be selected for studying in this paper, which is affected by sun glitter terribly. Through a spectrum analysis of the oil and water samples, two features --"spectral rotation" and "a pair of fixed points" can be found in spectral curves between crude oil film and water. Base on these features, an oil film information retrieval method which can overcome the influence of sun glitter is presented. Firstly, the radiance of the image is converted to normal apparent reflectance (NormAR). Then, based on the features of "spectral rotation" (used for distinguishing oil film and water) and "a pair of fixed points" (used for overcoming the effect of sun glitter), NormAR 894 /NormAR 516 is selected as an indicator of oil film. Finally, by using a threshold combined with the technologies of image filter and mathematic morphology, the distribution and relative thickness of oil film are retrieved.
INTRODUCEAs the development of economy, oil resources are more and more important for human. Followed, the oil spills in ocean from ships and oil platforms are increasing which have a serious influence to the marine environment. With remote sensing technology, several oil spill information such as distribution, thickness and type can be retrieved to help human clean oil film under scientific decisions.In oil spill remote sensing, thermal infrared camera, ultraviolet camera, microwave radiometer, multi-spectral scanners, synthetic aperture radar, laser fluorescence sensors have been used, but each of them has its own limitations [1] . With the widely use of the hyper-spectral technology, using hyper-spectral sensor to detect oil spills are constant attended by scholars. The hyper-spectral technology has great potential on the quantitative monitoring of oil distribution, type and quantity of spilled oil which effectively compensates the deficiencies of other sensors. *