Dynamic analysis of oil spill in Yangtze estuary with HJ-1 imagery

Abstract

This study was conducted to monitor oil spill changes in Yangtze estuary and to analyze their dynamic distribution by using HJ-1 charge-coupled device (CCD) imagery. First, the spectral response curve of the oil and other typical objects were analyzed to build the spectral feature space. Second, the classification algorithm of the polynomial kernel-based support vector machine (SVM) was studied to extract various levels of oil spill information including severe pollution, moderate pollution, and slight pollution. Third, the performance of the classification model was validated by comparison with other traditional approaches and the ground investigation data supported by the Shanghai Environmental Science Research Institute. Fourth, multi-temporal HJ-1 images were used to implement the classification with the polynomial kernel-based SVM algorithm. Finally, the oil-covered areas were calculated, the changes in spatial distribution were analyzed on the basis of the extracted results, and a statistical histogram was obtained. The results prove that the polynomial kernel-based SVM classification model has high accuracy with reliable performance for oil spill extraction. In addition, the dynamic analysis can be used to predict drifting trends and to provide important information for oil spill emergency response teams. Moreover, the HJ-1 satellite data can be applied to environmental monitoring.