基于包含多个刚体运动目标的单目视频三维重建研究

3D Reconstructing Dynamic Scene From Monoculartvideo Containing Multiple Moving Rigid Objects

基于单目视频的动态场景重建一直是一个非常艰巨的任务。目前大多数的研究主要集中在基于单目视频的静态场景重建,而对动态视频重建的关注比较少。此外,未知数量的运动目标和稠密重建增加了动态场景重建的难度。论文将提出一个完整的重建系统用于恢复包含多个刚体运动目标的单目视频中的动态场景结构。主要通过大尺度光流法获得特征点轨迹,并在此基础上对独立目标进行姿态估计,并提出了基于超像素的精确目标分割算法,目的是获得每一个实体的稠密重建。

3D reconstruction of dynamic scene from a single video has been a challenging task Whi le tradit ional methods can only recover the structure of a static scene from the motion of a moving camera, how to reconstruct the scene with dynamicobjects has been paid less attention In addition, the unknown number of moving objects and the dense reconstruction of each moving object make the problem more difficult. In this paper, a method for reconstructing monocular dynamic scene with multiple moving rigid objects captured by a single moving camera is proposed. It combines applying large displacement optical flow to get point trajectories, and gets the pose of dependent objects and their exact boundary. Finally the dense reconstruction of each object will be gotten.