摘要
SAR海冰图像分割对气候变化研究和航行安全保障具有重要意义。现有MRF分割算法仅能利用到单个极化SAR图像中的信息,易受相干斑噪声和地物信息不全面性的影响,不能准确有效地完成分割。为此,在充分利用单个极化SAR图像相邻区域之间的相似性的基础上,进一步融入多个极化SAR图像相同区域之间的一致性,由此提出了一种三维区域MRF(3DRMRF)的SAR海冰图像分割算法,能够实现SAR海冰图像的准确分割。通过RADARSAT-2和SIR-C获得的单视全极化SAR海冰图像的实验结果表明:和其他较先进的算法相比,所提出的算法优势明显,特别是具有更高的分割精度。
SAR sea ice image segmentation plays an important role in climate-change study and navigation safety. Ex- isting MRF segmentation algorithm can only use the information of single-polarized SAR image, and is easily affected by the interfering speckle noise and deficiency of feature information;and can not complete segmentation accurately and effectively. Aiming at this problem, on the basis of sufficiently using the similarity of adjacent regions in single- polarized SAR image, the consistency of the same regions in different polarized SAR images is introduced, and then a new sea ice image segmentation algorithm, called three-dimensional region-level MRF （3D-RMRF）, is proposed, which can achieve the accurate segmentation of the SAR sea ice images. The experiment results of acquiring the sin- gle-look full polarimetric SAR sea ice images using RADARSAT-2 and SIR-C indicate that compared with other ad- vanced algorithms, the proposed new algorithm has obvious superiority, especially has higher segmentation accuracy.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2013年第11期2551-2557,共7页
Chinese Journal of Scientific Instrument
基金
国家自然科学基金(41076120,61271381,61102154)
中央高校基本科研业务费专项资金(2012HGCX0001)资助项目
关键词
计算机应用
海冰
合成孔径雷达
图像分割
马尔可夫随机场
computer application
sea ice
synthetic aperture radar（SAR）
image segmentation
Markov random field（MRF）
