Research On SAR Ice Separation Method Combined With Saliency Detection

Sea ice floes are affected by wind and waves,and their flowing areas pose threats to ship navigation,ports,offshore oil and gas exploration,etc.However,the size and shape of the floating ice in the ice flow area are different,some are in close contact with each other,and some are surrounded by many tiny pieces of ice.In addition,SAR coherent system will lead to speckle noise in the image,which makes it difficult to separate floating ice in SAR image.In order to solve the problem of complex background and multi-target ice floes in SAR images,this paper proposes a SAR ice separation method based on background suppression saliency detection.The validity and applicability of this method are verified by experiments.The main work and research contents of this paper are as follows:Firstly,SAR sea ice data are preprocessed and data sets are made,including noise.suppression,region of interest extraction,data set making and classification,etc.Then,the SAR sea ice data is analyzed for feature extraction.The color features(mean and histogram)and texture features(gray level co-occurrence matrix and LM filter bank)of the image are mainly extracted,and the optimal feature set composed of features suitable for SAR sea ice image is selected through experimental analysis for subsequent experiments.Secondly,aiming at the complex background problems and multi-target floating ice problems in SAR floating ice separation,a method of floating ice separation in SAR flowing ice region combined with significance detection is proposed.Through the analysis of many superpixel segmentation methods,the simple linear iterative method(SLIC)is finally selected to carry out multi-level superpixel segmentation on the image,and the random forest regression quantity is learned to obtain a preliminaty saliency map.However,when suspected floating ice contacts the image boundary in a large area,the detection accuracy of saliency detection method will be reduced.To solve this problem,this paper proposes a background suppression module to suppress the suspected ice on the boundary through boundary region infection,which makes up for the deficiency of saliency method in sea ice detection.Finally,comparative experimental analysis is carried out on SAR sea ice data sets.Compared with eight advanced saliency detection methods and five sea ice segmentation methods,and combined with PR curve,F_measure,MAE,MIoU,FWIou and other evaluation indexes,it is proved that the method in this paper has significantly improved the complex background problems and multi-target floating ice problems existing in SAR sea ice images,such as isolated floating ice,small broken ice,boundary suspected ice,etc.