Volume XLII-3
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-3, 2083-2087, 2018
https://doi.org/10.5194/isprs-archives-XLII-3-2083-2018
© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-3, 2083-2087, 2018
https://doi.org/10.5194/isprs-archives-XLII-3-2083-2018
© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

  30 Apr 2018

30 Apr 2018

OIL SPILL AISA+ HYPERSPECTRAL DATA DETECTION BASED ON DIFFERENT SEA SURFACE GLINT SUPPRESSION METHODS

J. Yang1,2, G. Ren2, Y. Ma2, L. Dong3, and J. Wan1 J. Yang et al.
  • 1School of Geosciences, China University of Petroleum, Qingdao 266061, China
  • 2The First Institute of Oceanography, SOA, Qingdao 266061, China
  • 3North China Sea Branch, SOA, Qingdao 266061, China

Keywords: Oil Spill Detection, Hyperspectral Data, Glint Suppression, Wavelet Transform, Enhanced Lee Filter

Abstract. The marine oil spill is a sudden event, and the airborne hyperspectral means to detect the oil spill is an important part of the rapid response. Sun glint, the specular reflection of sun light from water surface to sensor, is inevitable due to the limitation of observation geometry, which makes so much bright glint in image that it is difficult to extract oil spill feature information from the remote sensing data. This paper takes AISA+ airborne hyperspectral oil spill image as data source, using multi-scale wavelet transform, enhanced Lee filter, enhanced Frost filter and mean filter method for sea surface glint suppression of images. And then the classical SVM method is used for the oil spill information detection, and oil spill information distribution map obtained by human-computer interactive interpretation is used to verify the accuracy of oil spill detection. The results show that the above methods can effectively suppress the sea surface glints and improve the accuracy of oil spill detection. The enhanced Lee filter method has the highest detection accuracy of 88.28 %, which is 12.2 % higher than that of the original image.