Identification Of Marine Spilled Oil Based On Remote Sensing Image

With the development of global economy and the improvement in oil demand, theoffshore oil transportation and offshore oil exploration have become more and more busy,which increase the likelihood of marine oil spill. In recent years, marine oil spill is aserious threat to the sustainable development of the marine environment and coastal citieseconomy. More and more countries begin to actively carry out oil spill remote sensingmonitoring.Marine oil spill monitoring is to analyze the visible and near infrared spectral bands.Synthetic aperture radar (SAR) is capable to monitor the occurrence of the marine oilspill timely, accurately and comprehensively with all-weather, all-time, large coverage,protected from the weather and so on. It has become one of the most important methodsin marine oil spill monitoring.The computer technology can fast and accurately access remote sensing imageinformation based on visual interpretation. In this paper, the main work is to extract oilspill information and summarize a series of steps of the digital processing SAR oil spillimage, as follows:The first step, this dissertation discusses SAR oil spill image preprocessing andmakes a comparison of different filter methods and different filter window sizes whenfiltering the SAR image. We adopt the Gamma Map filter method in5*5window sizewhich is the best result.The second step, texture features are introduced to distinguish oil spill in SAR imagebased on visual interpretation and many parameters are determined such as direction、stepsize and window size by gray-level co-occurrence matrix method. The focus of this paperis to select texture features using two different methods of determinant and matrixeigenvalue in order to reduce the amount of computation. We both select Mean、VAR、CON、DIS and ENT texture features, which improve the accuracy of screening results.The third step, three classification methods, the Mahalanobis distance, maximumlikelihood and support vector machine (SVM) are compared. The results show that the SVM method can distinguish the oil and background seawater more accurately andeffectively.