Classification of the Different Thickness of the Oil Film Based on Wavelet Transform Spectrum Information

Abstract

The detection of different thickness of spilled oil film is internationally recognized as one of the difficult problems. Hyperspectral remote sensing data can provide continuous spectrum, and be beneficial for the identification of oil film. The traditional detection method failed to make full use of different spectral characteristics of the oil film, so is unable to improving the identification precision of them. By deep digging the time-frequency information based on the wavelet transform, we can enhance the difference of the spectral characteristic between thick and thin oil film, so that we can find sensitive wavelength location of them. High frequency coefficient of wavelet transformation is able to indicating the sensitive spectral range of different thickness of oil film, and low frequency coefficient information is available for eliminating the noise of the image, consequently, enhancing the contrast between the categories. Based on this, this paper proposed a new method for classification of the various thickness of oil film based on wavelet transform spectrum information. Firstly, we extracted the sensitive wave bands of different thickness of oil film based on the analysis of the singularity of the high frequency wavelet coefficient curve. Secondly, we regenerated the new low-frequency wavelet coefficients image of sensitive bands which are regarded as sensitive bands in term of the analysis on the high frequency wavelet coefficients curve. Finally, we conducted the classification of the different thickness of oil spill film based on the new low-frequency wavelet coefficients image. The experiment has gotten good effect with Hyperspectral data obtained airborne from oil spill accident happening in Dalian in Bohai Sea on July 16, 2010.