Study Of Polar Sea Ice Detection Algorithm For C-Band Synthetic Aperture Radar

As the North and South Poles play a very import role in global climate changes,polar sea ice has gradually got attentions.Sea ice as an important factor affecting the climate not only has an important impact on global heat balance,but also on the temperature and salt balance of ocean.On the other hand,along with the melting of the Arctic sea ice,the open of Arctic waterway is possible.However,even in the summer,sea ice still exists in the Arctic waterway,threatening the safety of ships passing by.Therefore,monitoring of polar sea ice is of great significance.Synthetic aperture radar(SAR),with capabilities of weather and sunlight independent,as well as high spatial resolution,makes it suitable to monitor sea ice.Therefore,using spaceborne SAR to detect polar sea ice,i.e.to discriminate sea ice and open water,is an important research.In this paper,an algorithm of discriminating sea ice and open water from the C-band GaoFen-3 and Sentinel-1 SAR data was developed.The main research contents and preliminary conclusions are as follows:(1)Differences of sea ice radar scatter on C-band SAR horizontal-horizontal(HH)polarization and horizontal-vertical(HV)polarization data are studied.The variation of image gray values with the incidence angle and the variations of gray value under the effect of sea wind changing were analyzed.Based on the analysis,it was found that the gray value of HV polarization changes less with incidence angle than that in HH polarization,and the changes of HV polarization gray value with the effect of sea wind are smaller than that in HH polarization.These indicate that the HV polarization is relatively insensitive to the incidence angle and wind speed compared with HH polarization,and its gray values of the same object in the image are relatively consistent,which is more suitable to detect sea ice.(2)An algorithm using GaoFen-3 and Sentinel-1 SAR HV polarization data to discriminate sea ice and open water was developed.The method mainly uses image texture characteristics and support vector machine(SVM)classification to detect sea ice.In order to run the algorithm automatically without human interpretation,the improved watershed algorithm was proposed and used to automatically extract samples,and the samples are classified according to the obvious differences of sea ice and open water on image textures.Finally,sea ice detection was performed for the entire SAR image by classification model which was got by training samples with SVM.(3)The accuracy of the algorithm to discriminate sea ice and open water was further studied.In this paper,the accuracy of the algorithm was verified by comparing the detection results with visual interpretation.It shows the accuracy of detecting sea ice can reach more than 90%.(4)The applicability of the algorithm was also studied.The algorithm was tested by GaoFen-3 and Sentinel-1 SAR image.Most of the images can get good sea ice detection results,but there are also some images cannot.Some successful sea ice detection results have shown in the paper,and the reasons for unsuccessful cases were summarized.