A Minimal Solution for Non-perspective Pose Estimation from Line Correspondences
In this paper, we study and propose solutions to the relatively un-investigated non-perspective pose estimation problem from line correspondences. Specifically, we represent the 2D and 3D line correspondences as Plücker lines and derive the minimal soluti
- PDF / 1,878,041 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 102 Downloads / 153 Views
Abstract. In this paper, we study and propose solutions to the relatively un-investigated non-perspective pose estimation problem from line correspondences. Specifically, we represent the 2D and 3D line correspondences as Pl¨ ucker lines and derive the minimal solution for the minimal problem of three line correspondences with Gr¨ obner basis. Our minimal 3-Line algorithm that gives up to eight solutions is well-suited for robust estimation with RANSAC. We show that our algorithm works as a leastsquares that takes in more than three line correspondences without any reformulation. In addition, our algorithm does not require initialization in both the minimal 3-Line and least-squares n-Line cases. Furthermore, our algorithm works without a need for reformulation under the special case of perspective pose estimation when all line correspondences are observed from one single camera. We verify our algorithms with both simulated and real-world data. ucker lines · Non-perspective · Gr¨ obner Keywords: Pose estimation · Pl¨ basis · Line correspondences
1
Introduction
Pose estimation is a well-known problem in Computer Vision. It refers to the problem of finding the rigid transformation between a camera frame and a fixed world frame, given a set of 3D structures expressed in the world frame, and its corresponding 2D projections on the camera image. The 2D image projection to 3D structure correspondences can be either points or lines, or less commonly a combination of both. A minimum of three 2D-3D correspondences is needed to solve for the camera pose. Solutions to the pose estimation problem have significant importance in many real-world applications such as robotics localization, visual Simultaneous Localization and Mapping (vSLAM)/Structure-fromMotion (SfM), and augmented reality etc. The pose estimation problem for a single camera from point or line correspondences is a very well-studied problem, and a huge literature of robust and efficient solutions had been proposed [2,6,9,20,21,23,24,28] since the 1850s. This problem is also commonly known as the perspective pose estimation problem. In the recent years, due to the increasing popularity of multi-camera systems c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part V, LNCS 9909, pp. 170–185, 2016. DOI: 10.1007/978-3-319-46454-1 11
Non-perspective Pose Estimation from Line Correspondences
171
for robotics applications such as self-driving cars [7,15,16,19] and Micro-Aerial Vehicles [10,11], many researchers turned their attentions to the so-called nonperspective pose estimation problem [4,13,17,18,22] from point correspondences. The main difference between the perspective and non-perspective pose estimation problem is that for the latter, light rays casted from the 3D points do not meet at a single point, i.e. there is no single camera center. Consequently, many new algorithms [4,13,17,18,22] for the non-perspective pose estimation problem from point correspondences were proposed. Despite the fact that the perspective pose estimation prob
Data Loading...
