摘要： 通过分析校正变换的不唯一性，提出一种基于尺度的极线校正方法。该方法使用尺度不变特征变换算子分析特征点在尺度空间中的分布：在某一特定的尺度下会有更多稳定的特征点，使得图像的局部特征更加突出。对无人机拍摄的照片进行实验，结果表明，该方法能较好地消除垂直误差，且在校正优化最优尺度上能得到满足需要的视差图，从而根据视差恢复三维地形图。

关键词: 极线校正, 图像匹配, DoG算子, 尺度不变特征变换算子, 视差图, 三维恢复

Abstract: An epipolar rectification method based on scale is proposed through analysing the polysemy of rectification transformation. Matching algorithm of Scale Invariant Feature Transform(SIFT) descriptor is used to analyze the distribution of the scale space of the feature points, and that will cause more robust interest points in some particular scales than before of not optimized, and the local features of an image are then enhanced. Experiments are performed on images taken by unmanned aerial vehicle. The results demonstrate that the proposed method has strong application value. It is able to eliminate the perpendicularity errors, and also can generate satisfied disparity map in the optimal scale. Moreover, it can also recover 3D topographic map based on the disparity map.

Key words: epipolar rectification, image matching, DoG descriptor, Scale Invariant Feature Transform(SIFT) descriptor, disparity map, three-dimensional recovery

中图分类号: 

• TP391