A Study Of Marine Oil-spill Identification System Based On Information Fusion

Marine is the treasure trove of resources, is the economic lifeline of coastal states.China is a maritime country, the jurisdictional waters is vast, So preservation of marine clean and marine resource is a major issue bearing on our national economy.While the oil-spill is a serious marine environment disaster.As human activities on marine resources and increasing access to offshore oil exploration , the harm to the marine environment are increasing. Scientifically and effectively monitoring of marine oil pollution, accurately and rapidly identification the spill-oil is the prerequisite to solve this problem.However, due to two main features the marine oil spill image has : one is under the influence of wind, waves and currents , marine oil-spill due has dynamic characteristics, so there is no fixed shape of the oil-spill image.Another is due to the contrast of water and oil- spill is little, the boundary of oil and water in image is not very clear, has a certain ambiguity. The information fusion technology have advantages of increasing the information utilization of systems, improve the credibility and accuracy of synthesis-information, improve detection and identification of targets, reduce information acquisition costs, and enhance environmental adaptability.Therefore,use the information fusion technology to identify marine oil spill in the marine oil spill,has a great significance to quickly and accurately identify and solve the problem.This subject use the image pattern recognition technology and information fusion techniques to real-time monitoring and identification of marine oil spill. use Opencv to get video frame from the camera, then through the experimental data introducting pre-processings such as grayscale, filtering de-noising, center aligned and histogram equalization,Established feature extraction system based on GLCM , comparison of classification identify methods, we mainly choose the minimum risk efficient Support Vector Machine (referred to as SVM) identification method and improving it,in order to overcome the traditional method of complex error and low efficiency caused by the marine environment.For the complexity of the marine environment, this paper also proposed a semi-automatic identification method separate from the template, Moreover, using the feature level fusion and decision-making level fusion on identification method to overcome some shortcomings under certain conditions and single method,thereby maintaining accuracy and stability;using uniform scaling and optimal algorithm for selecting the best template template matching recognition to improve the accuracy of identification.