Ship Detection Algorithm Research Of SAR Image

SAR target detection has a large number of applications in military,fishery and other fields.It is significant to carry out SAR image ship detection.The main contents of this paper are two aspects: SAR image sea-land segmentation and ship target detection.In this paper,SAR imaging mechanism,image characteristics,cause of the noise and image characteristics are first analyzed.The traditional methods are introduced in this paper,and the detection methods from aspects of sea clutter statistical characteristics and factors affecting the ship detection are analyzed.A method to detect ship target is proposed in this paper.The accuracy of sea-land segmentation is the prerequisite and guarantee of ship target detection.In this paper,a method considered of differences in edge and gray informations between ocean and land is put forward,which is combined with region growing.ROEWA operator could not determine edge direction when SAR image edges are detected.In this paper,the logarithm of the image is taken,the multiplicative noise model is transformed into the additive noise model,and then the thinking method of Canny operator is referenced to calculate the direction of the edge.The region growing point is selected in the sea,then the differences of ocean and land in edge and gray scale are used to realize sea-land segmentation.In ship detection,An improved CFAR operator is proposed for two questions,one is the weak target detection is failed detected in the traditional two-parameter CFAR detection algorithm,the other is the target could not be complete detected in traversing image.When the target segmentation is done,the adaptive threshold is proposed for each window in combination with the image histogram.In this paper,we first determine the position of the ship,and then do threshold segmentation of each detection window to extract the ship.The experimental results show that the improved ROEWA operator can detect edge of SAR image,and the algorithm proposed in this paper can accomplish sea-landsegmentation well.In the target detection,An improved CFAR detection of this paper could effectively reduce the false alarm and missed detection.The proposed adaptive detection threshold is less sensitive to noise than the Otsu method,and the robustness is good.