基于SURF算法的无人机航空图像自动配准研究

Automatic registration of unmanned aerial vehicle images based on SURF operator

针对实时性要求中SIFT特征配准算法耗时长的缺点,本文将SURF(Speeded Up Robust Feature,即加速鲁棒特征)算法应用于无人机航空图像的自动配准问题中。首先利用Hessian检测子检测特征点,再通过粗匹配和细匹配得到匹配点对,最后执行几何变换完成对图像的配准。通过与SIFT(Scale Invariant Feature Transform,即尺度不变特征变换)配准方法进行对比,结果表明SURF算法在满足精度的前提下具有比SIFT算法计算量小、速度快的优点,有一定的理论和应用价值。

To overcome the shortcoming of SIFT registration algorithm, SURF operator is introduced into automatic registration of UAV aerial images. Firstly, the feature points are extracted through the Hessian detector. Secondly, matched points are selected through the coarse matching and the fine matching. Finally, the geometry transform is taken to complete the registration. An real experiment is performed, in which the proposed method is compared with the automatic registration method based on SIFT. The results show that SURF method meets the needs of accuracy and is faster than SIFT as well. So, our new method is valuable in both theory and practice.