基于稀疏分解和改进MRF模型的SAR海冰图像分割方法
发布时间：2018-10-31 14:23
【摘要】：合成孔径雷达(SAR)海冰图像的精确分割是准确解译海冰分布信息的前提,但现有分割方法受相干斑噪声影响严重,分割误差大,解译结果可靠性低。提出一种基于低秩稀疏表示的SAR海冰图像分割方法,首先利用噪声分布的稀疏性,通过鲁棒性主成分分析提取图像的稀疏分量,再利用双边滤波增强图像细节信息。针对基于固定势函数的MRF分割模型无法准确反映图像区域间关联性的问题,根据贝叶斯置信传播算法建立基于交互势函数的MRF分割模型准确分割海冰图像。利用Radarsat系列卫星数据验证算法性能,结果表明:和传统算法相比,本文算法在保持分割图像连通性的同时,能增强图像的细节信息,具有更高的分割精度。
[Abstract]:The accurate segmentation of synthetic Aperture Radar (SAR) (SAR) sea ice image is the premise of accurately interpreting the sea ice distribution information. However, the existing segmentation methods are seriously affected by speckle noise, and the segmentation error is large, and the interpretation result is low reliability. A method of SAR sea ice image segmentation based on low rank sparse representation is proposed. Firstly, the sparse component of the image is extracted by robust principal component analysis (PCA), and then the image detail information is enhanced by using bilateral filtering. Aiming at the problem that the MRF segmentation model based on the fixed potential function can not accurately reflect the correlation between the image regions, the MRF segmentation model based on the interaction potential function is established according to the Bayesian confidence propagation algorithm to accurately segment the sea ice image. The performance of the algorithm is verified by using Radarsat series satellite data. The results show that compared with the traditional algorithm, the proposed algorithm can not only maintain the connectivity of the segmented image, but also enhance the detail information of the image, and has a higher segmentation accuracy.
【作者单位】： 内蒙古科技大学信息工程学院;
【基金】：国家自然科学基金项目(61261028) 国家海洋局海洋遥测工程技术研究中心创新青年基金(2014003) 内蒙古自治区高等学校“青年科技英才支持计划”青年科技骨干项目(NJYT-14-B11) 内蒙古自然科学基金项目(2014MS0610) 内蒙古科技大学创新基金(2014QNGG07);内蒙古科技大学教改项目(YJSJGX2015006)资助
【分类号】：TN957.52
[Abstract]:The accurate segmentation of synthetic Aperture Radar (SAR) (SAR) sea ice image is the premise of accurately interpreting the sea ice distribution information. However, the existing segmentation methods are seriously affected by speckle noise, and the segmentation error is large, and the interpretation result is low reliability. A method of SAR sea ice image segmentation based on low rank sparse representation is proposed. Firstly, the sparse component of the image is extracted by robust principal component analysis (PCA), and then the image detail information is enhanced by using bilateral filtering. Aiming at the problem that the MRF segmentation model based on the fixed potential function can not accurately reflect the correlation between the image regions, the MRF segmentation model based on the interaction potential function is established according to the Bayesian confidence propagation algorithm to accurately segment the sea ice image. The performance of the algorithm is verified by using Radarsat series satellite data. The results show that compared with the traditional algorithm, the proposed algorithm can not only maintain the connectivity of the segmented image, but also enhance the detail information of the image, and has a higher segmentation accuracy.
【作者单位】： 内蒙古科技大学信息工程学院;
【基金】：国家自然科学基金项目(61261028) 国家海洋局海洋遥测工程技术研究中心创新青年基金(2014003) 内蒙古自治区高等学校“青年科技英才支持计划”青年科技骨干项目(NJYT-14-B11) 内蒙古自然科学基金项目(2014MS0610) 内蒙古科技大学创新基金(2014QNGG07);内蒙古科技大学教改项目(YJSJGX2015006)资助
【分类号】：TN957.52
【相似文献】
相关期刊论文 前6条
1 德;;测量海冰厚度新技术[J];科技信息;1987年09期
2 刘锡安,谭世祥;辽东湾海冰样区数学模型的建立[J];仪器仪表学报;1990年04期
3 张晰;张杰;孟俊敏;苏腾飞;刘眉洁;;基于极化散射特征的极化合成孔径雷达海冰分类方法研究:以渤海海冰分类为例[J];海洋学报(中文版);2013年05期
4 于波;孟俊敏;张晰;纪永刚;;结合凝聚层次聚类的极化SAR海冰分割[J];遥感学报;2013年04期
5 岳前进,毕祥军,季顺迎,张涛;航海雷达识别与跟踪海冰试验[J];大连理工大学学报;2000年04期
6 郎文辉;常灿灿;杨学志;张杰;孟俊敏;;ScanSAR模式海冰图像的分割[J];遥感学报;2014年01期
相关硕士学位论文 前9条
1 刘立剑;基于C波段PolSAR的多特征融合的海冰厚度反演[D];大连海事大学;2015年
2 张盼;基于ScanSAR海冰图像的冰水解译[D];合肥工业大学;2016年
3 门鹏;面向大分辨率差异SAR图像的海冰漂移跟踪研究[D];大连海事大学;2017年
4 张s，
本文编号：2302517
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2302517.html
