基于PCA-SIFT和马氏距离的SAR图像自动配准

Automatic Registration of SAR Image Based on PCA-SIFT and Mahalanobis Distance

针对SAR图像特征,提出了一种基于PCA-SIFT算法和马氏距离的SAR图像自动配准方法。首先,利用PCA-SIFT和马氏距离不变量理论相结合获得特征点及完成特征点配对;然后,利用RANSAC算法消除误匹配,并据此估计变换矩阵;最后,根据得到的变换矩阵进行图像间的映射,并对其进行重采样,实现SAR图像的自动配准。实验证明,该方法能够实现本文图像的配准,具有一定有效性和鲁棒性。

According to the characteristics of SAR image,a SAR image automatic registration algorithm based on PCA-SIFT and Mahalanobis distance was proposed.First it combined the PCA-SIFT and Mahalanobis distance invariant to obtain the feature points and complete matching.Next,it used RANSAC algorithm to cancel the wrong-matching,and estimated the transformation matrix based on this,and then accomplished the mapping among the images according to the obtained matrix,and made the resample on the mapping.Finally SAR image automatic registration was realized.Experiments showed that the method could achieve image registration in effectiveness and robustness.