摘要
文章提出了一种面向大区域、多航带无人机序列影像快速、无缝拼接方法:以影像序列的航带排布特性作为先验知识,加快影像间多度重叠SIFT特征点的匹配;通过Levenberg-Marquardt方法平差求解拼接区域各影像的变换参数;根据中心投影的像点位移规律与影像间的相对位置关系进行重叠区影像的优选与接边处影像的融合。实验结果表明该方法对于平地、丘陵、山地等不同地形的无人机影像数据都具有很好的适应性。
In the paper, a fast seamless unmanned aerial vehicle(UAV)sequential image mosaic meth- od was proposed for the images in large areas, in which sparse bundle adjustment was considered for the multi-stri Pimages. Particularly, the relative spatial position of the sequence images were input as prior knowledge, which would speed up the subsequent matching of the SIFT features. Then, the transforma- tion parameters between the images were solved by the Levenberg-Marquardt method, in which the sparse bundle adjustment was performed. At last, the optimal region of each image for the image mosaic was de- termined based on the displacement law of image points after central projection, and image fusion was per- formed on the overlapped areas. Experimental results showed that the proposed method could have a good adaptability for UAV image mosaic in plain, hill, and mountain terrains.
出处
《测绘科学》
CSCD
北大核心
2014年第11期129-132,共4页
Science of Surveying and Mapping