基于相对全变分的复杂背景SAR图像舰船尾迹检测  

Ship wake detection in SAR images with complex backgrounds based on relative total variation

作　　者：杨国铮[1,2] 禹晶[3] 孙卫东[1] 

机构地区：[1]清华大学电子工程系,北京100084 [2]北京市遥感信息研究所,北京100192 [3]北京工业大学计算机学院,北京100124

出　　处：《中国科学院大学学报（中英文）》2017年第6期734-742,共9页Journal of University of Chinese Academy of Sciences

基　　金：国家自然科学基金(61501008)资助

摘　　要：SAR图像舰船尾迹检测不仅能够反演运动舰船的航速航向信息,也有助于发现图像中弱小的舰船目标。现有的舰船尾迹检测方法对于简单背景SAR图像的检测效果较好,但复杂背景下的检测效果难以满足使用要求。提出一种基于能量泛函极小化的复杂背景SAR图像舰船尾迹检测方法。该方法采用相对全变分技术将图像分解为包含舰船尾迹的光滑成分和海背景纹理成分,通过剪切波变换高频系数重构增强光滑成分,再通过Radon变换检测光滑成分中的尾迹线。比对实验结果表明,本文所提方法对于复杂背景SAR图像的舰船尾迹检测效果明显优于现有的方法。The ship wake detection in SAR images is not only helpful in estimating the speeds and the directions of moving ships,but also in finding small ship objects. The existing ship wake detection methods achieve satisfactory results only for SAR images with simple backgrounds,but hardly work for SAR images with complex backgrounds. This work proposes a ship wake detection method for SAR images with complex backgrounds based on the relative total variation（ RTV）. The proposed method decomposes an SAR image into a cartoon component which contains ship wakes and an oscillating component which includes sea surface textures. Then the shearlet transform is used to process the cartoon component and some high-frequency coefficients are reconstructed to enhance the cartoon component. Finally the Radon transform is used for the enhanced cartoon component to detect ship wake lines. The comparison with the experimental results shows that the proposed methodfor the ship wake detection in SAR images with complex backgrounds obviously outperforms the existing methods.

关 键 词：合成孔径雷达 舰船尾迹检测 相对全变分 剪切波变换 RADON变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]