基于SAR影像海冰动态特征的提取方法

Dynamic Feature Extraction of Sea Ice in SAR Imagery

海冰边缘区(Marginal Ice Zone,MIZ)海冰种类繁多,具有较强的动态性,充分了解其覆盖范围及变化对于研究海冰变化、环境变化以及更好地开展人类活动等方面具有重要意义。以Radarsat-2SAR影像为例,根据MIZ在影像中的不同尺度和方向上的曲线特征,采用能够将图像分解为多尺度多方向信息的曲波变换进行特征提取。利用中尺度曲波系数邻域内的均值和灰度共生矩阵(GLCM)的能量值之间的相互关系,设计并实现了一种基于SAR影像的海冰动态特征的提取方法。利用此动态特征得到的MIZ与采用海冰密集度数据定义的MIZ相比较,准确率得到大幅的提升。结果表明:该特征能够有效地描述海冰的动态程度,为MIZ的识别算法提供基础和新的研究思路,同时能够为海冰分析模型、环境预测模型等提供有效参数。

The Marginal Ice Zone （MIZ） consists of different ice types,which makes it very dynamic.The dynamic features of sea ice in SAR imagery show as numerous curves in random orientations and scales.According to these curve features, the paper use middle scales coefficients of the curvelet transform which gives an optimal sparse representation of singularities along smooth curves at multi-scale and multi-direction to design a dynamic feature extraction method in SAR imagery.The feature is related to the mean and GLCM energy of curvelet coefficients magnitude and its neighborhood.The MIZ getting from the proposed feature has an obvious improvement of accuracy comparing with the MIZ getting from the SIC data.The re- sults demonstrate that it is an effective way to extract dynamic feature of sea ice.It can be used as the first step of the detection of MIZ,also used as an effective parameter in sea ice analysis model and environment prediction model.