Research On Marine Oil Spill Detection Method Using SAR Image Based On CFAR And Geometric Features

While marine industry to flourish brought serious pollution of the marine environment,mainly in the sudden oil spill pollution caused by oil drilling platforms,pipelines and ships,carrying oil leakage.At the same time,with the rapid development of satellite remote sensing technology,SAR with its all-weather,all-time,active emission of electromagnetic waves and other advantages,provides a powerful data support for marine oil spill detection,oil spill detection based on SAR images has become a hot technology applications in Remote Sensing Research.The current marine environmental monitoring department urgent need a automatic detection system with fast processing speed and high recognition precision,but has not yet seen,this mainly because:on the one hand the detection accuracy is not high for most algorithms,affected by the oil spill types,SAR imaging characteristics and viewing angle,ocean wind,etc.,many"false targets" to detect the presence of interference spill on SAR image;the other hand,some good algorithm are mostly dependent on a priori expert knowledge,more input parameters,automation level is not high.In this paper,the demand for the automatic detection of operational oil spill,abandon the traditional "one step" thinking,the use of "first crude extracts,after detailed classification,the first gray feature,after the geometric features" thinking,focusing on the preprocessing method,SAR image spill target segmentation based on CFAR algorithm,the study of oil spill dark spots classification combine geometric characteristics.Thesis completed work as follows:1.Combine with the characteristics of the oil spill and automated target detection algorithm theory,study the radiometric correction,geometric correction,speckle noise filtering and land masks and other pre-treatment process,and the selection of the optimal method.2.For defects of traditional CFAR algorithm which has the higher number of false targets,while for the dual-parameter CFAR algorithm which has high-precision to detect characteristics but computational complexity.In this paper,an improved dual-parameter CFAR spill target detection algorithm by simplifing calculation process and related criterion.The experiment proved that the missing rate in control,and the number of false targets reducing,while the processing speed is at least 2 to 3 times faster than the two-parameter CFAR algorithm.It is able to meet the automation needs of the SAR image oil spill detection of dark spots.3.To further eliminate interference of "false targets",in this article to improve the oil spill dark spots classification combine the geometric characteristics.By calculating the area,shape parameters of the dark spots to give the geometric characteristics,and using a weighted voting algorithm to complete similarity "scoring" for each dark spot oil spill,and finally detected dark spot spill into oil spill target,suspected oil spill target and target non-spill according to the total score,which make up the traditional "dichotomy" defects.Finally,experimental results show that the proposed algorithm improves the detection accuracy nearly 3 to 4 times.