SIFT特征匹配技术在自动相对定向中的应用

The application of SIFT feature matching method in the aotomatic relative orientation

提出了一种利用尺度不变特征变换(SIFT:Scale invariant Feature Transform)特征匹配技术应用于遥感影像自动相对定向的方法,SIFT特征是图像的局部特征,其对旋转、尺度缩放、亮度变化均保持不变,对视角变换、仿射变换、噪声均有一定程度的稳定性。文中给出了较为详细的理论分析和实验分析,结合小波变换建立影像金字塔采用分频道相关提高影像匹配的速度,利用最小二乘匹配方法提高了匹配的精度,取得了较好的实验结果。

A SIFT feature matching method applied to the relative orientation is proposed in this paper. The SIFT feature is invariant with the rotation, scale zoom, gray change, and is stable for the visual angle change, affine transformation and the noise. The detailed academic and experimental analysis is introduced. This method can improve the matching speed with the usage of the frequency channel correlation, and can enhance the matching precision with the lease squares image matching. Experiments show that this method is effective.