面向显著灰度差异的SAR海冰图像特征点匹配研究  

FEATURE POINT MATCHING FOR SAR SEA ICE IMAGES WITH SIGNIFICANT GRAY LEVEL DIFFERENCES

作　　者：王妍[1] 郭浩[1] 黄超[1] 安居白[1] Wang Yan;Guo Hao;Huang Chao;An Jubai(Dalian Maritime University,Dalian 116026,Liaoning,China)

机构地区：[1]大连海事大学,辽宁大连116026

出　　处：《计算机应用与软件》2022年第7期227-234,255,共9页Computer Applications and Software

基　　金：中国国家海洋局的海洋非营利性行业研究专项资金项目(2013418025)。

摘　　要：SAR海冰图像易受斑噪声影响而导致特征显著性下降。利用一些传统的非线性特征描述符进行特征提取,可以克服使用线性特征在降噪时无区别滤波导致局部特征缺失的问题,但是当待配准图像存在显著性灰度差异时,这一类方法配准效果不理想。针对存在显著性灰度差异SAR海冰图像,提出一种新的配准方法。考虑到LDB描述子对灰度差异适应性的固有优势,以及GLOH描述子在配准主方向判断上的优势,将二者结合提出GLDB描述子,使得主方向的选择更加准确。在特征点匹配阶段,提出一种基于几何对应匹配和灰度差异对应匹配的两步匹配策略。通过实验对比在一定程度上验证了该算法的有效性。The sea ice image of SAR is susceptible to speckle noise, which results in a substantially decrease in qualities of characteristics. Using traditional non-linear feature descriptors for feature extraction can overcome the problem of local feature deletion caused by indistinguishable filtering when using linear features in noise reduction. However, when there is large gray difference in the image to be registered, the registration effect of this kind of method is not ideal. This paper proposes a novel registration method for SAR sea ice images with significant gray level difference. Considering the advantages of LDB descriptors in adaptability to gray difference and GLOH descriptors in judging the main direction of registration, we proposed GLDB descriptors by combining the two to make the selection of main direction more accurate. In the process of feature point matching, this paper put forward a two-step matching strategy based on geometric correspondence matching and gray difference correspondence matching. The effectiveness of this algorithm was verified to some extent by experimental comparison.

关 键 词：特征匹配 海冰 漂移跟踪 SAR 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]