Rigid Body Pose Estimation from Line Correspondences

In this paper, we aim to solve pose estimation of rigid body motion in real time with 3d lines model. According to the line's perspective projection model, we design a new error function expressed by the average integral of the distance between line segments to estimate parameters. Considering the continuely of motion, we restore cracked line segements with re-projection of Model lines Constrianted. Last, we proposal to estimate many frames jointly in framework of SFM and get better precious while bears slow speed. Comparisons on synthetic and real images demonstrate that baseline methods get accuracy estimations in complex environments. For plane objects, the precious of pose on x, y, z axes are better than 0.5m in 100m distance, and those of relative positions perpendicular to the optical axis and along the optical axis are better than 0.3%.