基于均匀特征匹配的无人机影像拼接  

UAV Image Mosaic Based on Uniform Features Matching

作　　者：马超[1] 赵西安[1] 王青松[1] 

机构地区：[1]北京建筑大学测绘与城市空间信息学院,北京100044

出　　处：《北京建筑工程学院学报》2013年第4期47-51,共5页Journal of Beijing Institute of Civil Engineering and Architecture

摘　　要：通过特征匹配的方法进行无人机影像拼接,不需要地面控制点坐标和内外方位元素.基于SIFT算法进行特征匹配,对于匹配中容易出现的误匹配点,通过合理控制算法中的阈值大小并采用唯一性约束和视差约束来进行剔除,采用分块控制匹配点数目来剔除密集匹配点对,使匹配点对分布均匀.使用RANSAC算法来计算影像之间的转换模型,实验证明这种方法具有很好的拼接效果.A new approach for mosaicing images in unmanned aerial vehicle （UAV） is proposed in the paper. Firstly, the image features are extracted by the multi-scale algorithm, which is a SIFT improved. Based on the uniqueness constraint and parallax constraint, the images are matched. The dense matched point-pairs are processed uniformly using the partitioning images. Secondly, the transformation model between images is built in the RANSAC. The experiment results show that the approach is effective for the UAV image mosaic.

关 键 词：影像拼接 均匀匹配 RANSAC算法 

分 类 号：P231.5[天文地球—摄影测量与遥感]