2017/09/12 by Robert Maier, Maier, Robert, Raphael Schaller +3 · 1 citation
Computer Science · Engineering · #Advanced Vision and Imaging #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Optical measurement and interference techniques #Robotics and Sensor-Based Localization
paper · pdf · doi:10.48550/arxiv.1709.03763
openalex publication_date 2017/09/12 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
State-of-the-art methods for large-scale 3D reconstruction from RGB-D sensors\nusually reduce drift in camera tracking by globally optimizing the estimated\ncamera poses in real-time without simultaneously updating the reconstructed\nsurface on pose changes. We propose an efficient on-the-fly surface correction\nmethod for globally consistent dense 3D reconstruction of large-scale scenes.\nOur approach uses a dense Visual RGB-D SLAM system that estimates the camera\nmotion in real-time on a CPU and refines it in a global pose graph\noptimization. Consecutive RGB-D frames are locally fused into keyframes, which\nare incorporated into a sparse voxel hashed Signed Distance Field (SDF) on the\nGPU. On pose graph updates, the SDF volume is corrected on-the-fly using a\nnovel keyframe re-integration strategy with reduced GPU-host streaming. We\ndemonstrate in an extensive quantitative evaluation that our method is up to\n93% more runtime efficient compared to the state-of-the-art and requires\nsignificantly less memory, with only negligible loss of surface quality.\nOverall, our system requires only a single GPU and allows for real-time surface\ncorrection of large environments.\n