Information Extraction of Marine Oil Spill with Collaborative Images of ASAR and MODIS
|School||Shandong University of Science and Technology|
|Course||Surveying and Mapping Engineering|
|Keywords||ASAR multi-source image fusion Gray-level Co-occurrence Matrix(GLCM) texture analysis|
Marine oil spill affects the ocean environment and ecology seriously, so the monitoring of oil spill has important significant for elimination of disasters and environmental assessment. With the development of remote sensing science and technology, remote monitoring is proved to be an important means of oil spill monitoring currently. This paper takes the Bohai sea surface oil spill as research object, discussed the method to extract oil spill information by using the fusioned image between ASAR and MODIS data. The paper focuses on the identification of fusion method and the algorithm for oil spill information extraction.The main results are summarized as follows:1) The speckle noise removal of space-borne radar ASAR image.Compared the results of ASAR image that filtered by 7 filters in different window size, and analysing the results according to 5 evaluation indexes, author determines the application of the Lee filter with 5*5 window size as the best filtering scheme. The choice makes ASAR image not only removing the noise in greatest degree, but also preserving edge information as much as possible, and provides a rich texture for subsequent information extraction based on texture.2) The fusion between ASAR and MODIS images.Statistical analyse the each band of MODIS image against the characteristics of marine oil spill information extraction, and calculate the OIF index for the bands which are identified initially, then eventually identified the best band combination with three unrelated bands that have maximum amount of information. Then use 4 methods of image fusion between MODIS image that with selected bands and the ASAR image after image data registration. Based on 7 integration of evaluation criteria, analyse the fusion results and determine that IHS is the best integration program, to achieve optimal coordination of data.3) Identification and extraction of oil spill information.Make texture analysis for the fusion image a, then calculate Gray-level Co-occurrence Matrix (GLCM) of the samples to analyse directional and window size of 8 texture features, and ultimately determine 5 kinds of texture feature with 9*9 window size. Based on the extracted texture feature vectors, classify the study area with 2 kinds of supervised classification including minimum distance and SVM, in order to realize the recognition and extraction of sea water, oil spill and like oil film. According to the classification results, SVM is the optimal classification method.