基于加权混合SPF主动轮廓的SAR图像水陆分割  

Water-Land Segmentation in SAR Image Using Weighted Hybrid SPF Based Active Contour

作　　者：沈嘉程 韩斌 鲍秉坤 SHEN Jia-cheng;HAN Bin;BAO Bing-kun(School of Communications and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing,Jiangsu 210003,China;School of Computer Science,Nanjing University of Posts and Telecommunications,Nanjing,Jiangsu 210023,China)

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003 [2]南京邮电大学计算机学院,江苏南京210023

出　　处：《电子学报》2024年第6期2000-2010,共11页Acta Electronica Sinica

基　　金：国家自然科学基金(No.62325206,No.61936005,No.62201281);江苏省自然科学基金(No.BK20220392);南京邮电大学引进人才自然科学研究启动基金(No.NY222004)。

摘　　要：针对现有主动轮廓模型(Active Contour Model,ACM)无法进行准确高效的SAR(Synthetic Aperture Radar)图像水陆分割的难题,提出一种基于加权混合符号压力(Signed Pressure Force,SPF)主动轮廓的新模型.该模型将中值拟合中心和均值拟合中心结合起来,建立新的混合符号压力函数,以抑制干扰边缘的影响.通过设计基于绝对中位差的自适应权重,实现自动调节内外拟合中心的比例系数,以更合理地描述灰度差异.此外,计算平滑后图像梯度信息构建边缘引导函数并引入到SPF函数中,以进一步提高分割性能和演化效率.在实际SAR图像上进行分割实验,结果表明:与相关实验模型相比,提出模型在保持高效的同时能获得更加准确的分割结果,其中Dice相似系数(Dice Similarity Coefficient,DSC)至少提高6%.To solve the problem that the existing active contour models(ACM)cannot accurately segment the synthetic aperture radar(SAR)water-land images,a new ACM based on the weighted hybrid signed pressure force is proposed.The median fitting center and mean fitting center are combined to establish a new hybrid signed pressure function,which can suppress the influence of interference edges.The adaptive weights based on the absolute median difference are designed to automatically adjust the coefficients of the inner and outer fitting centers,which can describe grayscale differences more reasonably.Moreover,the gradient information of the smoothed image is calculated to construct an edge guidance function and introduced into the SPF function to further improve segmentation performance and evolution efficiency.Segmentation experiments are conducted on real SAR images and the results show that compared with the related ACMs,the proposed model can perform accurate segmentation with high efficiency and stability,and the Dice similarity coefficient(DSC)is improved by at least 6%.

关 键 词：水陆分割 SAR图像 主动轮廓模型 混合符号压力函数 绝对中位差 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] P237[自动化与计算机技术—计算机科学与技术]