Abstract: Images with large rotation angle from low altitude Unmanned Aerial Vehicle(UAV) with weakness stability cannot realize automatically mosaic using the traditional methods with geo-referenced orthographic images. In this paper, an image mosaics technique with Scale Invariant Feature Transform(SIFT) and multi-resolution spline merging is proposed for low attitude UAV images. A fast and batch image distortion correction algorithm based on single image distortion look-up table is presented for multiple images. A novel homograph-constrained image matching algorithm based on SIFT feature is proposed to calculate the optimal transformation matrix between overlapping images. A multi-resolution spline mosaics technique with optimal transformation matrix of overlap images is presented. Experimental result shows that the proposed method can obtain a large number of stable correspondences and transformation matrix of overlap images for further seamless image mosaic, which can be useful for fast and automatic images mosaic from low altitude UAV with large rotate angle and instability.

Key words: Unmanned Aerial Vehicle(UAV), image mosaic, Scale Invariant Feature Transform(SIFT), optimal transformation matrix, multi- resolution spline

摘要： 低空无人机抗风能力弱、稳定性差，影像旋偏角大且存在突变，无法按照常规正射影像镶嵌方法获得全区域拼接影像。为此，提出一种基于尺度不变特征变换(SIFT)特征匹配与多分辨率样条融合的低空无人机影像全自动拼接方法。对非量测影像进行畸变校正，利用查找表设计多幅影像快速畸变校正算法。采用SIFT特征的单应约束影像匹配算法，计算相邻影像的最优变换矩阵。给出最优变换矩阵的多分辨率样条融合影像拼接算法。实验结果表明，该方法能够获得大量稳定的匹配点对，影像间几何变换关系稳定，得到的拼接影像无缝清晰，适用于大旋角、低稳定性的低空无人机影像非摄影测量快速拼接。

关键词: 无人机, 影像拼接, 尺度不变特征变换, 最优变换矩阵, 多分辨率样条

CLC Number: 

• TP391