SAR和光学图像精配准技术的研究  

Investigation on Fine Registration for SAR and Optical Image

作　　者：尤红建[1] 胡岩峰[1] 

机构地区：[1]中国科学院电子学研究所,北京100190

出　　处：《雷达学报（中英文）》2014年第1期78-84,共7页Journal of Radars

基　　金：国家重大型号工程项目资助课题

摘　　要：光学图像和SAR图像的配准是实现多源图像融合和复合分析的基础,该文介绍了一种光学图像和SAR图像精配准的方法。首先采用人工的方式选择3～4个同名控制点,在此基础上实现图像的粗略配准,保证SAR与光学图像尺度和旋转的归一化;然后在粗配准的光学图像上提取特征点,并分别采用不同的梯度算子提取SAR和光学图像的梯度强度,基于梯度强度进行归一化互相关,从而得到不同图像上的同名特征点;最后根据提取的同名特征点构建不规则三角网,并对每个三角网构成的小面元进行图像的精确配准,最终实现SAR和光学图像的精配准。通过采用星载图像的精配准处理和检验,证明了该方法的可行性,能获得较好的配准效果。The registration of SAR and optical remote sensing image is the basise for fusing of multi-source image and comprehensive analysis. In this paper a new fine registration method for SAR and optical image is proposed. Firstly, three to four corresponding points are selected manually to realize a coarse registration that eliminates the differences in scale and rotation. Many characteristic points in the optical image are detected and the corresponding points in SAR image are extracted using normalized gradient correlations based on the different gradients by operators. An irregular triangle network is constructed using these corresponding points and each triangle region is finely registered. Finally SAR image and optical image are finely registered. Experiment and processed results demonstrate the feasibility of this method.

关 键 词：SAR图像 光学图像 梯度强度 精配准 归一化互相关(NCC) 

分 类 号：O436[机械工程—光学工程]