基于匹配的遥感影像自动纠正方法研究

针对多源遥感影像人工几何纠正方法存在精度差、效率低等问题，提出了一套新的基于匹配技术的自动纠正算法流程.该流程基于仿射变换对待纠正影像和已地理编码的参考影像进行粗配准.在利用Harris算子从两幅图像中分别提取特征点的基础上，通过由特征点和小波影像金字塔引导的从粗到精的匹配策略获得控制点对，再利用多项式拟合迭代法剔除错误点对.在获得大量的高精度控制点对后，利用不规则三角网(TIN)仿射变换法对影像进行纠正，并对杭州地区多源遥感影像进行纠正试验.结果表明,使用该方法选取的控制点均方根误差(RMSE)可以控制在0.5个像素以内.

Study on image automatism registration methods based on matching

Aimed at resolving the problem of lower-precision and time-consumption in traditional manual geometry registration for multi-source remote sensing images,a new automatic imagery registration workflow based on image matching was proposed.An affine transformation model was employed for coarse registration between unregistered and reference geo-coded image.Based on the feature points extracted from two images by Harris operator,control point pairs between the two images were obtained by a coarse-to-fine image matching strategy guided by feature points and wavelet image pyramids,and big error pairs were expurgated using iterative polynomial error checker.After a great amount of high-precision control point pairs was gained,triangle irregular network(TIN) affine transformation was used to register images.Experiments were performed by registering multi-source remote sensing images of Hangzhou area.The results show that the root mean square error(RMSE) of the control points obtained by the workflow is lower than 0.5 pixels.