Abstract: The imaging mechanisms for optical and SAR imagery is different, which throws a big difficulty in the registration between them. Aiming at above problem, this paper proposes an registration method based on virtual searching window. Based on spatial features of registered imagery, the virtual searching window is constructed, and then the spatial features and gray statistical features are combined together to register optical and SAR imagery. While the algorithm’s searching efficiency is guaranteed, its precision is improved. Images of Radarsat-2 and ASTER with big differences in scale and angle are tested. Some features are extracted manually to check the transform model’s precision, the results indicate that errors are lower than one pixel, which effectively proves that the algorithm is robust to differences in scale and angle between optical and SAR imagery and has a high precision.

Key words: optical and SAR imagery, spatial feature, virtual searching window, normalized cross correlation coefficient

摘要： 光学和SAR影像的成像机理及像元表现形式互不相同，给两者的精配准造成很大困难。针对上述问题，提出基于虚拟搜索窗口的区域配准法，根据配准影像的空间特征，构建虚拟搜索窗口，将空间特征和灰度统计特征结合用于光学和SAR影像的自动配准，在保证算法搜索效率的同时提高配准精度。选取具有较大尺度和角度偏差的RADARSAT-2与ASTER影像进行实验，结果证明该算法对光学和SAR影像之间的角度和尺度偏差具有较强的鲁棒性，配准精度小于一个像素。

关键词: 光学与SAR影像, 空间属性, 虚拟搜索窗口, 归一化互相关系数

CLC Number: 

• TN751.1