基于三聚类中心K-means算法的SAR船只检测方法

A ship detection method for SAR imagery based on three-centroid K-means clustering algorithm

为解决高分辨合成孔径雷达(SAR)图像中基于海杂波模型的CFAR船只检测方法适用性受限的问题,提出了一种基于三聚类中心的K-means船只检测方法.该方法将SAR图像划分为船只目标、海杂波及其他干扰3个聚类,利用K-means聚类算法求得最高聚类中心值,并将其作为检测阈值进行船只初步检测,然后结合分辨率和船只尺度等先验信息进行形态学滤波操作得到最终检测结果.基于实测数据的实验结果表明,所提方法无须海杂波的统计信息,且不依赖于SAR图像的分辨率,可有效地服务于高分辨率SAR图像中的船只检测任务.

Based on the three-centroid K-means clustering algorithm,this paper proposed a feasible ship detection method for the high resolution SAR imagery.The method proposed has nothing to do with the resolution of SAR imagery,and it needs no prior knowledge about the sea clutter.With the K-means algorithm,the method in this paper firstly divides the SAR data into three clusters,i.e.ship targets,sea clutter and other disturbance,then executes clustering operation to get the centroid of the three clusters,and set the maximum be the detection threshold to get the candidates,after then,a mathematic morphological filter is involved to obtain the final detection results.The experiment results based on 3-scene real RADARSAT-2data show that the proposed method could be a valuable one for ship detection in high resolution SAR imagery.