联合显著性与MRF的SAR建筑物分割算法  

SAR Building Segmentation Algorithm Combining Saliency and MRF

作　　者：张磊 王小龙 刘畅 ZHANG Lei;WANG Xiaolong;LIU Chang(Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100190,China;School of Electronic,Electrical and Communication Engineering,University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院空天信息创新研究院,北京100190 [2]中国科学院大学电子电气与通信工程学院,北京100049

出　　处：《计算机工程》2022年第4期284-291,298,共9页Computer Engineering

基　　金：国家重点研发计划(2017YFB0503001)。

摘　　要：针对经典马尔可夫随机场(MRF)在进行高分辨率SAR图像分割时存在容易受到斑点噪声干扰等问题,提出一种基于建筑物指数相似度距离及MRF模型(BISD-MRF)的高分辨率SAR建筑物分割算法。基于较复杂SAR场景下建筑物目标可能呈现多种形态结构的问题,设计一种多尺度显著性建筑物指数(MSBI)方案来提取建筑物目标的显著性特征,并通过强度信息重构、纹理显著性提取、频谱显著性信息统计来分别提取不同类型区域的显著性信息,构建适用于SAR建筑物目标的显著性模型。在此基础上,将MSBI值引入到改进的基于改进余弦函数的势函数模型中,利用余弦函数对邻域像素MSBI值进行相似性度量,同时利用特征空间语义信息对像素及其邻域像素标签信息进行有效约束,以提升势函数模型对高分辨率SAR建筑物目标的表征能力。不同平台下的建筑物分割实验结果表明,与MRF、MBI、FRFCM等算法相比,本文算法分割性能平均提升了4.3~10.7个百分点,更适用于较复杂场景下高分辨率SAR建筑物的分割任务。Aiming at the problem that classic Markov Random Field(MRF)problem is susceptible to speckle noise during high-resolution Synthetic Aperture Radar(SAR)image segmentation,this study proposes a high-resolution SAR building segmentation algorithm based on a Building Index Similarity Distance and MRF(BISD-MRF).Considering that building targets in a more complex SAR scene may show multiple morphological structures,a Multi-scale Saliency Building Index(MSBI)scheme is designed to extract the saliency of building target features using intensity information reconstruction texture saliency extraction and spectral saliency information statistics;the saliency information of different types of target areas is extracted,and a saliency model suitable for SAR building targets is constructed.On this basis,the MSBI value is introduced to the improved potential function model based on the improved cosine function,which is used to measure the similarity of the MSBI value of the neighborhood pixels,and the feature space semantic information is used to measure the pixel and its neighborhood pixel labels.The information can be effectively constrained to improve the ability of the potential function model to characterize high-resolution SAR building targets.The results of building segmentation under different platforms show that the segmentation performance of the proposed algorithm improved by an average of 4.3~10.7 percentage points compared to traditional algorithms,such as MRF,MBI,and FRFCM,and it is more suitable for high-resolution SAR building segmentation tasks in more complex scenes.

关 键 词：高分辨率SAR图像 多尺度显著性建筑物指数 改进MRF模型 建筑物提取 势函数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]