Researches On Identifying Sea Surface Oil Slicks Based On Hyperspectral And Multi-spectral Remote Sensing Technologies

Oil spill accidents will cause oil leakage and the leaked crude oil will damage the marine environment.Monitoring the oil slicks formed by the leaked oil by remote sensing technologies is benefit for handling and warning the disasters.Airborne hyperspectral remote sensing technology can detect the floating oil slicks quickly and accurately,but its scanning scope is limited and it's expensive.Thus,airborne hyperspectral remote sensing technology is mainly used for detecting oil slicks under emergency.Satellite-based multispectral remote sensing technology can monitor the sea surface in a large scale with less cost,but it cannot distinguish different thicknesses of oil slicks or its detection performance is affected by the weather to a large extent.Thus,satellite-based multispectral remote sensing technology is mainly used for long time monitoring.Apparently,airborne hyperspectral remote sensing technology and satellite-based multispectral remote sensing technology are complementary for detecting oil slicks.Thence,to effectively respond to oil spills,this paper studied oil slick detection methods based on hyperspectral/multispectral remote sensing technologies.It is defective for traditional oil slick detection methods.Traditional methods utilizing hyperspectral remote sensing images focus on detecting thick oil slicks and cannot identify sheens which exist in a wide range of sea tables.Traditional methods based on spectral analysis utilizing hyperspectral remote sensing images cost too much time.They are not suitable for mapping oil spills rapidly.Traditional methods utilizing multispectral remote sensing images will misidentify seawater as oil slicks under the inhomogeneous marine environment,which will reduce oil slick detection accuracy.To deal these problems,this paper has done some researches:1)Evaluate the ability of seawater composition and hydrocarbon substance indices to detect different thicknesses of oil slicks and construct oil slick detection method based on them utilizing airborne hyperspectral images.Since thick oil slicks contain much hydrocarbon substances and their light transmittance is very weak,traditional spectral indices of hydrocarbon substances perform well for detecting thick oil slicks.However,thin oil slicks,such as sheens,have excellent light transmittance because they contain very little hydrocarbon substances.Thus,traditional spectral indices cannot distinguish sheens from seawater.Aiming at this problem,this research proposed a set of evaluation procedure to assess the ability of seawater composition and hydrocarbon substance spectral indices to detect different thicknesses of oil slicks.In addition,an oil slick detection method was established based on the evaluation results.AVIRIS images captured in the oil spill accident of Gulf of Mexico,2010 were used to test the hypotheses and methods proposed by this research.The results indicated that hydrocarbon substance indices were more suitable for identifying thick oil slicks,seawater composition indices were more suitable for identifying thin oil slicks and the spectral indices of hydrocarbons and seawater had a kind of complementarity for identifying oil slicks.Oil detection method established by the complementary spectral indices could accurately identify different thicknesses of oil slicks.This research revealed the ability of the seawater composition and hydrocarbon substance indices to identify different thicknesses of oil slicks.It provided available solutions for detailed oil slick mapping tasks.This study has been published in the journal Remote Sensing.2)Encode the spectra of seawater and different thicknesses of oil slicks into spectral DNA chains and then identify oil slicks through the spectral genes extracted from the DNA chains utilizing airborne hyperspectral images.Traditional methods based on spectral analysis need plenty of time to distinguish different thicknesses of oil slicks.There is much redundant spectral information which does not participate in the oil slick identification process in the airborne hyperspectral data.To deal this problem,this research adopted DNA encoding method to handle the multidimensional spectral information.Then,this research proposed two strategies to extract spectral genes from the encoded spectral DNA chains.Finally,the oil slicks in the hyperspectral images would be identified by the extracted spectral genes.Experimental results indicated that the proposed method could identify seawater and different thicknesses of oil slicks correctly and rapidly.In addition,the proposed method was suitable for hyperspectral images captured at different time and different sea areas.This research overcame the complex and time-consuming shortcomings of traditional spectral analysis-based identification methods.It provided an available solution for rapid and accurate oil spill mapping.One related research of this syudy has been published in the journal Remote Sensing and another has been finished.3)Detect oil slicks under the inhomogeneous marine environment by spectral indices which have complementary sensitivity for different seawater utilizing multispectral images.Spaceborne multispectral images can monitor a wide range of seas.The marine environments in multispectral images tend to be inhomogeneous caused by lighting conditions,foreign substance injection,marine life et al.The inhomogeneous marine environment will lead to reduced accuracy because the traditional oil slick detection indices usually misidentify the different seawater as oil slicks.To solve this problem,this research studied the sensitivity of different spectral indices to different seawater environments and assessed the detection ability of different spectral indices for different seawater.Based on this,this research proposed a kind of oil slick detection method which chose spectral indices whose sensitivities to inhomogeneous seawater were complementary to detect oil slicks in multispectral images.Landsat 8 and MODIS images came from different seas and different sources were used to test the hypothesis and method proposed in this research.Experimental results indicated that different spectral indices were sensitive to different seawater and the proposed method could effectively eliminate the false targets caused by uneven seawater environments.In addition,the proposed method was suitable for both Landsat 8 and MODIS images.This research broke through the limitation of traditional methods to treat the surrounding seawater as a homogeneous environment.It provided an available solution for detecting oil slicks under inhomogeneous marine environment utilizing multispectral images.A part of this research has been reported and published on the collected papers of 2018 International Workshop on Big Geospatial Data and Data Science(BGDDS 2018).