采用自适应背景窗的舰船目标检测算法

An Algorithm of Ship Target Detection Based on the Adaptive Background Window Function

针对合成孔径雷达(SAR)图像应用中,传统滑动窗口检测算法无法对近距、近岸或不同大小目标实现精确检测的问题,提出了一种自适应窗口的舰船目标检测算法。该算法首先利用阈值滤波实现海面的分离,然后通过对分离后待检目标的像素体积统计,剔除大体积陆地目标,得到待检目标,并根据其像素分布设置自适应窗口,通过对自适应窗口内目标像素和背景像素的分离统计,最终拟合得到待检目标附近背景的K-分布概率模型进行恒虚警检测。相对于传统滑动窗口检测算法,自适应窗口检测算法可实现对目标背景像素的精确统计及K-分布拟合,以及对舰船类目标的精确检测。实验结果证明,在相同的虚警概率条件下,对复杂海面情况的SAR图像进行舰船检测,自适应窗口检测算法比传统局部窗K-分布的恒虚警检测结果的品质因子高0.34。

An adaptive detection algorithm is proposed to solve on the problem in the application of the SAR images that the traditional target detection algorithms,which are on the basis of sliding window,cannot give an accurate detection to targets that are in short distances,near to the coast,or very different in size.The method separates the sea surface using threshold filtering,and then the targets to be detected are obtained by counting the pixel volume of targets and eliminating the most acreage of the land.Adaptive windows are set according to the pixel distribution.The K-distribution possibility model of the scenes near the targets is obtained by separating the counting of the target pixel from the counting of the background pixel on adaptive windows.The adaptive window detection algorithm has the advantages of achieving accurate statistics of background pixel of targets and K-distribution fitting over the traditional target detection algorithm.Experimental results show that the quality factor of the proposed adaptive window detection algorithm is 0.34 higher than that of the traditional target detection algorithm when the detecting is performed on the SAR images with the circumstances of the complicated sea surface and the same CFAR condition.