无人机核线影像的稀疏匹配与稠密匹配

Sparse Matching and Dense Matching of UAV Epipolar Images

无人机影像转化为水平核线影像后,能够有效地减少同名点的搜索空间。在此基础上,本文使用SIFT算子进行了稀疏匹配,并用BP算法进行了稠密匹配。结果表明:(1)SIFT算子获取的同名点比较少,但是计算方法简单,同名点空间坐标精确,适用于大范围获取简要的空间三维信息;(2)BP算法计算复杂度高,可以获取地物大量的同名点,适用于小范围的地物三维重建。总体而言,两者各有优缺点,在实际的应用中可互补。

Converting UAV images to epipolar images, makes a good effect on reducing the search space of corresponding point matching. On this basis, SIFT operator based sparse stereo matching and BP algorithm based dense stereo matching were presented in this paper. The result indicated that：Less corresponding points, simple calculation and accurate spatial coordinates were shown in the results of SIFT operator, so SIFT operator was suitable to acquire summary spatial information in a big-scale area. The computation of BP algorithm was complex but a large number of same points were outputted, which indicated that BP algorithm applied to 3D reconstruction in a small range. In a word, each of them has its own advantages and disadvantages, and they can be complementary.