Abstract: Matching multi-sensor images is one of the key steps in vision navigation and multi-sensor images fusion. A common image matching method is to match the features detected in both images separately. However, for the images with great difference in imaging apparatuses, for example, SAR image and optical image, it is hard to obtain corresponding features. In this paper, Multi-scale Feature Support（MFS） algorithm is proposed to match multi-sensor images. MFS algorithm only needs to detect multi-scale edge feature in one image and find the maximal support to this feature in the transform space of the other image. The MFS is calculated by normalizing the summation of gradient intensity. The matching result can be obtained by optimizing the support function through genetic algorithm. Experimental results show that the MFS algorithm can match SAR image and optical image effectively and efficiently.

Key words: multi-sensor images, multi-scale feature support, image matching, genetic algorithm

摘要： 异源图像匹配是视觉导航、多源图像融合分析的关键步骤之一，常用的匹配方法是分别从两幅图像中提取特征，再对特征进行匹配。但是对于成像机理差别较大的异源图像，如SAR图像和可见光图像，很难提取到同名特征。提出一种基于多尺度支持度的异源图像匹配方法，只需要从一幅图像中提取多尺度边缘特征，在变换空间中寻找另一幅图像对该特征的最大支持度。支持度的计算采用了标准化梯度强度和的形式。采用遗传算法对支持度函数解空间进行全局寻优来获取最优匹配点。实验结果表明，该方法能有效实现SAR图像和可见光图像的匹配，处理时间能够满足工程要求。

关键词: 异源图像, 多尺度支持度, 图像匹配, 遗传算法