DETECTION OF OCEAN SURFACE ANOMALY USING OPTICAL SATELLITE IMAGES

摘要

The observation of oil pollution on ocean surface is an important task for managing marine environment. Because the physical characteristic of oil is very different from sea water in optical spectra, anomaly is usually formed by oil slicks. In this study, the optical satellite images are used to detect this kind of anomaly on ocean surface. First, by assuming that the undisturbed ocean surface is the dominant background of the interested area, the RX algorithm is used to transform the original multispectral data and measure the degree of anomaly for each image pixel. Next, a Gaussian mixture model is used to characterize the distributions of background and anomaly respectively. Then, the Expectation Maximization (EM) algorithm is used to solve the needed parameters for the Gaussian mixture model. Finally, according to Bayes decision rule of minimum error, an optimized threshold can be found to extract the anomaly. Furthermore, with the solved distributions and the optimized threshold, the theoretical accuracy can be applied to evaluate the quality of the extracted results. Experiment results show that the proposed method can extract various kinds of anomalous patches in different site by corresponding optical satellite images.