一种结合上下文与边缘注意力的SAR图像海陆分割深度网络方法  

A Sea-Land Segmentation Method for SAR Images Using Context-Aware and Edge Attention Based CNNs

作　　者：梁烽 张瑞祥 柴英特 陈金勇[2] 茹国宝[1] 杨文[1] LIANG Feng;ZHANG Ruixiang;CHAI Yingte;CHEN Jinyong;RU Guobao;YANG Wen(School of Electronic Information,Wuhan University,Wuhan 430072,China;CETC Key Laboratory of Aerospace Information Application,Shijiazhuang 050002,China)

机构地区：[1]武汉大学电子信息学院,湖北武汉430072 [2]中国电子科技集团公司航天信息应用技术重点实验室,河北石家庄050002

出　　处：《武汉大学学报（信息科学版）》2023年第8期1286-1295,共10页Geomatics and Information Science of Wuhan University

基　　金：国家自然科学基金(61771351);中国电子科技集团航天信息应用技术重点实验室开放基金(SXX19629X060)。

摘　　要：海陆分割对于合成孔径雷达(synthetic aperture radar,SAR)图像海洋目标检测、海岸线提取等任务具有重要意义。针对实际应用中多分辨率SAR图像海陆分割难题,提出了一种基于上下文与边缘注意力的海陆分割方法。该方法利用通道注意力机制融合不同尺度和层次的上下文特征,设计了边缘提取支路提供边缘信息,进一步提高了海陆边界的分割准确率。同时,构建了基于高分三号卫星数据的多分辨率SAR图像海陆分割数据集,该数据集涵盖了多个分辨率,包括港口、岛屿等多种海陆边界类型。并基于所构建的多分辨率SAR图像海陆分割数据集,对所提网络的有效性和各模块的作用进行了实验分析。实验结果表明,所提网络的整体预测准确率和平均交并比分别达到了98.21%和96.47%,能够较好地完成海陆分割任务。Objectives:Sea-land segmentation is of great significance for tasks such as ocean target detec⁃tion and coastline extraction in synthetic aperture radar(SAR)image.To solve the problem of sea-land seg⁃mentation of multi-resolution SAR image in practical applications,this paper presents a sea-land segmenta⁃tion method based on context and edge attention.Methods:The proposed method uses the channel atten⁃tion mechanism to fuse context features of different scales and levels,and designs an edge extraction branch to provide edge information for further improving the segmentation result of the boundary area.In addition,a sea-land segmentation dataset of multi-resolution SAR image based on the Gaofen-3 satellite data is pro⁃vided.The dataset covers multiple resolution images,including various sea-land boundary types such as ports,islands.Using this multi-resolution SAR image coastline segmentation dataset,we perform experi⁃mental analyses on the effectiveness of the proposed network and the contributions of each module.Results and Conclusions:Experimental results show that the proposed method can work well for the task of sealand segmentation,the overall classification accuracy and mean intersection over union achieve 98.21%and 96.47%,respectively.

关 键 词：合成孔径雷达 海陆分割 边缘提取 注意力机制 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置] P237[自动化与计算机技术—控制科学与工程]