摘要
针对常规算法在倾斜立体影像匹配时的不足，本文提出了一种基于互补不变特征的高精度均匀匹配算法，其可实现特征匹配数量较多、空间分布均匀度较好、匹配点定位精度较高的影像匹配。首先融合Harris-Affine（仿射不变的Harris角点）和SIFT（Scale Invariant Feature Transform）互补不变特征，然后利用得到的单应矩阵对影像进行射影变换并消除影像失真，最后提取Harris特征进行双重约束归一化互相关（Normalized Cross Correlation，NCC）匹配。利用近景影像、无人机影像和AMC580影像进行的实验结果表明，本文所提算法是一种适用于倾斜影像的鲁棒匹配算法。
Since the conventional algorithm for stereo image matching of oblique images has many drawbacks, a high preci- sion and well-proportioned matching algorithm based on complementary invariant features is designed to realize image matching with more feature matching, better spatial distribution uniformity and higher positioning accuracy. Firstly, the complementary in- variant features of Harris-Affine and SIFT are combined to get the homography matrix, which are then used to carry out the projec- tive transformation of the image and eliminate image distortion. At Last, Harris features are extracted for hybrid constraints Nor- malized Cross Correlation matching. In order to verify the effectiveness of the proposed algorithm, experiments are carried out u- sing close-range image, UAV image and AMC580 image. The results show that it is a robust matching algorithm and can be ap- plied to oblique image.
出处
《测绘科学与工程》
2017年第1期40-44,71,共6页
Geomatics Science and Engineering
基金
基金项目：国家自然科学基金资助项目（52110295,61271421）.
关键词
倾斜影像
互补不变特征
双重约束NCC匹配
随机采样一致性
oblique stereo image
complementary invariant feature
hybrid constraints NCC matching
random sample consensus
