无人机视频影像的准实时融合算法  

An algorithm for quasi real time fusion of UAV video images

作　　者：耿中元[1,2] 袁继英 李英成 孙新博[1,2] 朱祥娥 冯亮 GENG Zhongyuan;YUAN Jiying;LI Yingcheng;SUN Xinbo;ZHU Xiang'e;FENG Liang(Key Laboratory for Aerial Remote Sensing Technology of National Administration of Surveying,Mapping and Geoinformation,Beijing 100039,China;China TopRS Technology Co.,Ltd.,Beijing 100039,China;Inner Mongolia Land Surveying and Planning Institute,Hohhot 010010,China;Chinese Academy of Surveying and Mapping,Beijing 100036,China)

机构地区：[1]航空遥感技术国家测绘地理信息局重点实验室,北京100039 [2]中测新图(北京)遥感技术有限责任公司,北京100039 [3]内蒙古自治区土地调查规划院,呼和浩特010010 [4]中国测绘科学研究院,北京100036

出　　处：《测绘科学》2019年第10期72-78,共7页Science of Surveying and Mapping

基　　金：国家重点研发计划子课题项目(2016YFC0803104,2016YFC0803109)

摘　　要：针对无人机视频影像可以实时地融合到地理信息系统中,但是无人机的位置和姿态数据存在较大误差,导致相邻的视频影像的同名地物点并不重合的问题,该文提出准实时的无人机视频影像融合算法,采用基于滑动窗口的光束平差方法来优化相机的外方位元素,这种方法比全部视频影像做整体光束平差的延时更低,而且可以显著地改善无人机视频影像的融合效果。算法使用优化后的相机外方位元素对视频影像做正射纠正,再把纠正后的影像叠加到三维大场景中。实验结果表明,这种方法可以在准实时的条件下融合无人机视频影像,并且相邻视频影像的同名地物点在融合结果中是重合的。Unmanned aerial vehicle(UAV)video images can be fused into geographic information system in real time.Homonymous points in adjacent video images are not coincident,because of the errors in the position and attitude data of the UAV.In this paper,aquasi real time fusion method of UAV video images was proposed to solve the problem.Exterior camera elements were optimized with the sliding window bundle adjustment.This method had a lower delay than the bundle adjustment of all video images,and could significantly improve the fusion result of UAV video images.Ortho-rectification of UAV video images was done with the optimized exterior camera elements,and then the result images were overlaid into the 3D large scene.Experimental results showed that UAV video images were fused in quasi real time with this method,and homonymous points in adjacent video images were coincident in the fusion result.

关 键 词：无人机 视频影像融合 光束平差 正射纠正 

分 类 号：P231[天文地球—摄影测量与遥感]