【作者】 张冬梅；

【导师】 卢小平；

【作者基本信息】 河南理工大学 ， 测绘科学与技术， 2020， 硕士

【摘要】 无人机具有机动性强、操作简便、成本低及数据获取周期短等优点,已成为当前大尺度影像获取及三维重建的主要手段。随着我国信息化进程的快速发展,对三维模型重建的需求越来越大,三维重建技术已成为智慧城市、卫生医疗、导航、考古、数字媒体、影视娱乐、虚拟现实及增强现实等领域的热点。因此,研究如何利用无人机影像进行快速高效三维重建的技术方法,具有重要的理论意义和实际价值。本文以无人机影像作为数据源,提出一种改进的基于序列影像的三维重建技术方法,为智慧城市建设和信息化建设提供技术支撑。在对无人机序列影像三维重建理论与算法进行研究基础上,在影像匹配过程中对RANSAC算法进行了改进,采用一种优化的样本采样方法和使用自适应阈值验证参数模型,通过实验验证表明了本文算法可有效提高剔除误匹配点的效果。对面片多视图稠密重建(PMVS)技术进行深入研究基础上,针对面片初始化时法向量存在的误差问题,提出对法向量进行校正的改进方法,该方法通过选取面片K邻域区间进行曲面拟合的方法求面片的法向量,并通过约束条件判断是否对法向量进行校正,使面片可以更好地符合物体表面模型,从而提高面片优化效率。在面片扩展阶段,选择空间点在影像上的投影距离中心点较近的面片作为初始面片进行重新选择和扩展,提高了结构模型的重建精度。利用无人机采集的两个区域中的序列影像进行三维重建实例验证,通过与原始算法相比,结果表明重建精度和重建效率有了显著提高。更多还原

【Abstract】 Unmanned aerial vehicles(UAV)have the advantages of strong maneuverability,easy operation,low cost and short data acquisition period.They have become the main means of large-scale image acquisition and three-dimensional reconstruction.With the rapid development of China’s informatization process,the demand for 3D model reconstruction is increasing.Research on 3D reconstruction technology has become a field of smart cities,health care,navigation,archaeology,digital media,film and television entertainment,virtual reality and augmented reality Hot spot.Therefore,it is of great theoretical and practical value to study how to use UAV images for rapid and efficient 3D reconstruction.This paper takes the UAV image as the data source,and proposes an improved three-dimensional reconstruction technique based on sequence images,which provides technical support for the construction of smart cities and information construction.On the basis of studying the theory and algorithm of 3Dreconstruction of UAV sequence image,the RANSAC algorithm was improved,and an optimized sample sampling method and an adaptive threshold were used to verify the parameter model.Based on the in-depth study of patch multi-view dense reconstruction(PMVS)technology,an improved method for normal vector correction is proposed for the error problem of normal vectors during patch initialization.The fitting method finds the normal vector of the patch,and judges whether to correct the normal vector through the constraint conditions,so that the patch can better conform to the object surface model,thereby improving the optimization efficiency of the patch.In the patch expansion stage,the patch whose spatial point is projected closer to the center point on the image is selected as the initial patch for reselection and expansion,which improves the reconstruction accuracy of the structural model.The sequence images in the two areas collected by the UAV are used to verify the 3D reconstruction example.Compared with the original algorithm,the results show that the reconstruction accuracy and reconstruction efficiency have been significantly improved.更多还原