基于多层显著性模型的SAR图像舰船目标检测  

Ship detection in SAR image based on multi-layer saliency model

作　　者：扈琪 胡绍海[2,3] 刘帅奇 HU Qi;HU Shaohai;LIU Shuaiqi(College of Electronic and Information Engineering,Hebei University,Baoding 071002,China;Institute of Information Science,Beijing Jiaotong University,Beijing 100044,China;Beijing Key Laboratory of Advanced Information Science and Network Technology,Beijing 100044,China)

机构地区：[1]河北大学电子信息工程学院,河北保定071002 [2]北京交通大学信息科学研究所,北京100044 [3]现代信息科学与网络技术北京市重点实验室,北京100044

出　　处：《系统工程与电子技术》2024年第2期478-487,共10页Systems Engineering and Electronics

基　　金：国家自然科学基金(62172030,62172139);河北省自然科学基金(F2022201055);中国博士后科学基金(2022M713361);河北省高等学校科学技术研究项目(BJ2020030);模式识别国家重点实验室开放课题基金(202200007)资助课题。

摘　　要：针对合成孔径雷达图像舰船目标检测问题,提出了一种结合选择机制与轮廓信息的多层显著性目标检测方法。首先,利用非下采样剪切波和频谱残差法进行全局显著性区域提取。其次,提出了一种基于动态恒虚警率的活动轮廓显著性模型,逐步滤除候选区域的虚警,提取目标轮廓,从而实现目标的精确检测。所提方法能够由粗到细地快速捕获目标区域,从而实现高效、高分辨率合成孔径雷达图像舰船检测。最后,在真实SAR数据集进行了测试,与其他经典的舰船检测方法相比,所提算法不仅有效地抑制了海杂波的影响,而且在检测精度上有较大提高。Aiming at the problem of ship detection in synthetic aperture radar(SAR)images,a multi-layer saliency target detection method combining selection mechanism and contour information is proposed.Firstly,non-subsampled shearlet transform(NSST)and spectral residual method are used to extract the globally significant region.Secondly,an active contour saliency model based on dynamic constant false alarm rate(CFAR)is proposed to filter out the false alarms of candidate regions step by step and extract the target contour,so as to realize the accurate detection of targets.The proposed method can quickly capture the target area from coarse to fine,so as to achieve high-efficiency and high-resolution SAR image ship detection.Finally,the algorithm is tested on real SAR datasets.Compared with other classical ship detection methods,the proposed algorithm not only effectively suppresses the influence of sea clutter,but also greatly improves the detection accuracy.

关 键 词：SAR图像目标检测 非下采样剪切波变换 显著性检测 活动轮廓模型 

分 类 号：TN957.52[电子电信—信号与信息处理]