基于特征点和最优路径的无人机影像拼接方法

A Mosaic Method in Unmanned Aerial Vehicle Images based on Feature Points and Optimal Path

利用SIFT算法对无人机影像进行匹配,采用距离阈值的方法剔除匹配后距离较近的特征点。为了降低误差累积对影像拼接效果的影响,提出了最佳基准影像的选择方法,同时将Dijkstra算法引入到拼接路径的搜索中,利用影像间的投影转换误差构建搜索权阵。结果表明:最优路径的拼接方法能有效避免投影误差较大的影像对后续拼接影像的影响,减少了影像投影转换次数,改善了影像的拼接效果。

The SIFT algorithm was used in Unmanned Aerial Vehicle images matching,the distance threshold filter was used to remove too close feature points.In order to reduce the error accumulation which can effect on the results of image mosaic,and shorten the path splicing,an optimal base image selection method was introduced,meanwhile,the Dijkstra algorithm was introduced into the search of stitching path,and the project conversion errors between the two images were used to build the search array.The experimental results indicated that this optimal path method can avoid the influences of the image which has large projection error mosaic with the follow-up images effectively,and reduce the number of image projection transformation,which can improve the mosaic effect.