无人机图像全自动生成大比例尺真正射影像方法

随着无人机遥感技术逐步从研究开发阶段发展到实际应用阶段，采用无人机图像进行大比例尺的真正射影像生成方法的研究具有重要的现实意义．本文旨在探索将计算机视觉中获得巨大成功的多视三维重建技术应用到对无人机影像处理中，给出了一种基于运动恢复结构重建算法和多视图立体视觉算法全自动生成大比例真正射影像的方法．本文首先分析了无人机图像PMVS重建点云的特点，给出一种由基于面片多视图立体视觉稠密点生成数字表面模型的方法，然后详细介绍了包括正射影像图像坐标映射模型、可见性计算、基于Markov随机场能量优化的面片选择和匀光处理等真正射影像生成的关键步骤．实验结果以及与商业软件的比较表明：本文给出的方法在野外地形和城市区域均能获取有效的真正射影像结果．

Automatic generation of large scale true ortho-image from UAV images

With the rapid development of unmanned aerial vehicles （UAV） remote sensing technology, UAV finds its place in various applications where the large-scale orthoimage generation is a popular and important step. This paper is meant to apply hugely successful multi-view 3D reconstruction techniques in recent years in computer vision field to UAV images, and presents a method of automatic generation of large scale true orthoimage based on structure from motion （SfM） and multi-view stereo （MVS） technologies. This paper includes the following key parts： First, an approach is proposed to build a Digital Surface Model （DSM）, seemingly a necessary step for any good orthoimage generation, from the points cloud reconstructed by Patch-based Multi-view Stereo （PMVS）.Then the key steps of our orthoimage generation method are presented, such as reference system definition and transformation, the patch visibility computation, Markov Random Field（MFF） model based patch selection and seamless mosaicking. Finally, our orthimage generation method is tested with several sets of UAV images, and compared with some commercial software. The experiments show that our orthimage generation method performs satisfactorily for both terrain and urban area UAV images.