Abstract
In this paper we present a novel feature-based RGB-D camera pose optimization algorithm for real-time 3D reconstruction systems. During camera pose estimation, current methods in online systems suffer from fast-scanned RGB-D data, or generate inaccurate relative transformations between consecutive frames. Our approach improves current methods by utilizing matched features across all frames and is robust for RGB-D data with large shifts in consecutive frames. We directly estimate camera pose for each frame by efficiently solving a quadratic minimization problem to maximize the consistency of 3D points in global space across frames corresponding to matched feature points. We have implemented our method within two state-of-the-art online 3D reconstruction platforms. Experimental results testify that our method is efficient and reliable in estimating camera poses for RGB-D data with large shifts.
Highlights
Real-time 3D scanning and reconstruction techniques have been applied to many areas in recent years with the prevalence of inexpensive depth cameras for consumers
Manuscript received: 2016-09-09; accepted: 2016-12-20 techniques have various popular applications, e.g., in augmented reality (AR) to fuse supplemented elements with the real-world environment, in virtual reality (VR) to provide users with reliable environment perception and feedback, and in simultaneous localization and mapping (SLAM) for robots to automatically navigate in complex environments [1,2,3]
A major limitation of KinectFusion is that camera pose estimation is performed by frame-to-model registration using an iterative closest point (ICP) algorithm based on geometric data, which is only reliable for RGB-D data with small shifts between consecutive frames acquired by high-frame-rate depth cameras [4, 5]
Summary
Real-time 3D scanning and reconstruction techniques have been applied to many areas in recent years with the prevalence of inexpensive depth cameras for consumers. Even though an ICP-based framework can effectively deal with RGB-D data with small shifts, it solves a non-linear minimization problem and always converges to a local minimum near the initial input because of the small angle assumption [4]. This indicates that pose estimation accuracy relies strongly on a good initial guess, which is unlikely to be satisfied if the camera moves rapidly or is shifted suddenly by the user.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.