基于强角点的SAR图像自动配准

图像配准技术在合成孔径雷达图像处理中占据至关重要的位置。针对合成孔径雷达图像配准对时间和精度的需求,提出基于角点特征的配准算法。提取两幅图像的角点构造三角形,计算三角形的角度,寻找两幅图像对应的相似三角形,用三角形的顶点坐标求得仿射变换系数,参考图像进行仿射变换,与待配准图像比较相似性,最大相似性对应的变换图像为配准结果。引入用于视频追踪的强角点,对该算法进行验证。实验数据表明该算法的快速准确,并具有一定的抗噪性。

Auto SAR Image Registration Based on Strong Corner

For the need in time and precision of SAR registration, the article advanced an algorithm based on corner feature that satisfied this need well. Extracted corners in both two images, constructed triangles using these corners, found the similar triangles by computing the angles of each triangle in two images and got the affine transformation coefficients by using the coordinates of six vertexes of two similar triangles in two images. Transformed the reference image using the affine transformation coefficients, compared the similarity of two images, and the transformed image corresponding to the maximum similarity was the registration result. Introduced the strong corner always used in video track into the experiment to verify the algorithm. The experiment data demonstrates that the algorithm is fast and accurate and has the ability of anti-noise to some extent.