GPU Accelerated Robust Scene Reconstruction

We propose a fast and accurate 3D reconstruction system that takes a sequence of RGB-D frames and produces a globally consistent camera trajectory and a dense 3D geometry. We redesign core modules of a state-of-the-art offline reconstruction pipeline to maximally exploit the power of GPU. We introduce GPU accelerated core modules that include RGBD odometry, geometric feature extraction and matching, point cloud registration, volumetric integration, and mesh extraction. Therefore, while being able to reproduce the results of the high-fidelity offline reconstruction system, our system runs more than 10 times faster on average. Nearly 10Hz can be achieved in medium size indoor scenes, making our offline system even comparable to online Simultaneous Localization and Mapping (SLAM) systems in terms of the speed. Experimental results show that our system produces more accurate results than several state-of-the-art online systems. The system is open source at https://github.com/theNded/Open3D.

Authors

• Wei Dong
• Jaesik Park
• Yi Yang
• Michael Kaess

Keywords

• Point cloud compression
• Three-dimensional displays
• Simultaneous localization and mapping
• Accuracy
• Pipelines
• Graphics processing units
• Feature extraction
• Trajectory
• Odometry
• Intelligent robots
• GPU Acceleration
• 3D Reconstruction
• Point Cloud
• Feature Matching
• Online System
• Mapping System
• Core Module
• Point Cloud Registration
• Reconstruction System
• Linear System
• Density Map
• First Pass
• Error Function
• Real-world Datasets
• Depth Images
• Jacobian Matrix
• Hash Function
• Reconstruction Results
• Nearest Neighbor Search
• Pose Of Frame
• Iterative Closest Point
• Loop Closure
• Camera Pose
• Relative Pose
• Spatial Query
• Query Point
• Shared Memory
• RGB-D Images
• GPU Memory