Research On The Technology Of Spilling Oil Identification

In recent years,with the development of social production technology and industrial level,the inland river transportation system is increasingly perfect,however,frequent water pollution accidents and traffic accidents are coming.The marine oil spill detection technology has made remarkable achievements all around the world,but in terms of inland river spilling oil detection it is still inadequate.This paper is based on the study of the development of Zhongshan inland river intelligence platform,the real-time video oil spill monitoring and identification are carried out in the actual needs of the inland river area.This paper includes the following three parts:(1)Inland river video acquisition and preprocessing.First of all,this paper introduces about the environment of video stream acquisition and image preprocessing,the preprocessing include reducing image data volume,removing solar light reflection from the water surface and enhancing the color of the image.As the color is the key factor in the oil spill detection,this paper pay more attention to the contrast of three color enhancement methods,experiments and analysis.(2)Inland river image feature extraction and selection.In this paper,the research of oil spill image are carried out from three aspects:color feature,texture feature and geometric feature.The feature extraction and data analysis are carried out by using color moments,high purity color,gray level co-occurrence matrix,color co-occurrence matrix,Hough transform,Shi-Tomasi and so on.After analyze the experiment data,we can determine the difference of each feature about the inland river image,which are contain inland spilling oil images,inland water surface images and other images of the inland river,and then determine the characteristics of each feature.(3)Inland river oil spill identification.According to the role of the six features in the classification,a directed acyclic graph support vector machine(DAG-SVMS)is constructed.Finally,we can optimize the DAG-SVMS classifier according to the classification results.The experimental results show that the combination of high purity color feature and gray level co-occurrence matrix can identify the spilling oil image efficiently.In this paper,we can screen three kinds of images which include the obvious spilling oil image of inland river,the not obvious oil image and other rich color image from inland river image by using the high-purity color features.Finally we can exclude the third category by using gray-level co-occurrence matrix.