Offshore Oil Slicks Extraction by Landsat Data Based on eCognition Software in South China Sea

Extraction of oil slicks from remote sensing data is a basic work for investigation of oil spill and natural oil leakage. It is time-consuming and laborious when the traditional method of artificial interpretation is applied to tens of thousands of remote sensing data. In order to improve the efficiency of oil slicks extraction, eCognition, object-oriented image analysis software, is introduced to set up a rule set which can realize the batch processing of oil slick extraction. Landsat TM data of known oil seep site from Gulf of Mexico is used as training data to set up the rule set. Following the several processes such as image segmentation, hierarchical classification, training Sample selection, feature space optimization and nearest neighbor classification, oil slicks and several other sea surface targets were extracted in the format of shape within ten minutes. The result shows that the oil extraction based on eCognition rule set is faster and more accurate than manual interpretation. When the rule set was applied to the sea area of South China Sea, there were two significant discoveries. One is that the pollution of the white tiger oil field is very serious. Oil slicks can be detected on almost all available Landsat TM data. Another is that the first natural oil and gas seep is come to light in South China Sea.