一种基于最小生成树的无人机影像拼接方法  

A method of UAV image mosaic based on minimum spanning tree

作　　者：张冬梅[1] 卢小平[1] 张航 陶晓晓 余振宝 ZHANG Dongmei;LU Xiaoping;ZHANG Hang;TAO Xiaoxiao;YU Zhenbao(Key Laboratory of Spatiotemporal Information and Ecological Restoration of Mines,MNR,Henan Polytechnic University,Jiaozuo,Henan 454003,China;Reconnaissance,Planning,Design and Research Institute,YRCC,Zhengzhou 450003,China)

机构地区：[1]河南理工大学自然资源部矿山时空信息与生态修复重点实验室,河南焦作454003 [2]黄河勘测规划设计院有限公司,郑州450003

出　　处：《测绘科学》2020年第5期42-48,共7页Science of Surveying and Mapping

基　　金：2016年国家重点研发计划项目(2016YFC0803103);河南省高校创新团队支持计划项目(14IRTSTHN026);河南省创新型科技创新团队支持计划项目。

摘　　要：针对无人机影像拼接速度慢精度低的问题,提出了一种基于最小生成树的无人机影像拼接方法。对采集到的序列影像进行特征点匹配,采用LM法和RANSAC算子剔除误匹配点,通过距离阈值使特征点均匀化;计算每一张影像的准则值,以准则值最小的影像为基准影像进行坐标转换;将最小生成树算法引入到拼接路径中,实现多张影像的拼接。实验结果表明,该方法提高了影像的拼接速率,有效避免误差累积对后续拼接影像的影响,改善了拼接效果。Aiming at the slow speed and low precision of uav image stitching,an image mosaic method based on minimum spanning tree is proposed. Firstly,feature point matching is carried out on the collected sequence images,LM method and RANSAC operator are adopted to eliminate the mismatched points,and the feature points are homogenized through the distance threshold. Then,the criterion value of each image is calculated,and the image with the minimum criterion value is used as the reference image for coordinate transformation. Finally,the minimum spanning tree algorithm is introduced into the Mosaic path to realize the Mosaic of multiple images. Experimental results proved that the proposed method can not only improve the image mosaicing rate,but also effectively avoid the influence of error accumulation on subsequent mosaicing images and improve the mosaicing effect.

关 键 词：无人机影像 影像拼接 透射变换 最小生成树 

分 类 号：P237[天文地球—摄影测量与遥感]