2016/09/27 by Maksym Dzitsiuk, Dzitsiuk, Maksym, Jürgen Sturm +8
Computer Science · Engineering · #Advanced Vision and Imaging #Computer Graphics and Visualization Techniques #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Robotics and Sensor-Based Localization
paper · pdf · doi:10.48550/arxiv.1609.08267
openalex publication_date 2016/09/27 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
Creating 3D maps on robots and other mobile devices has become a reality in\nrecent years. Online 3D reconstruction enables many exciting applications in\nrobotics and AR/VR gaming. However, the reconstructions are noisy and generally\nincomplete. Moreover, during onine reconstruction, the surface changes with\nevery newly integrated depth image which poses a significant challenge for\nphysics engines and path planning algorithms. This paper presents a novel, fast\nand robust method for obtaining and using information about planar surfaces,\nsuch as walls, floors, and ceilings as a stage in 3D reconstruction based on\nSigned Distance Fields. Our algorithm recovers clean and accurate surfaces,\nreduces the movement of individual mesh vertices caused by noise during online\nreconstruction and fills in the occluded and unobserved regions. We implemented\nand evaluated two different strategies to generate plane candidates and two\nstrategies for merging them. Our implementation is optimized to run in\nreal-time on mobile devices such as the Tango tablet. In an extensive set of\nexperiments, we validated that our approach works well in a large number of\nnatural environments despite the presence of significant amount of occlusion,\nclutter and noise, which occur frequently. We further show that plane fitting\nenables in many cases a meaningful semantic segmentation of real-world scenes.\n