SLAMRecon is a real-time 3D dense mapping system based on RGB-D camera. The output is a reconstructed indoor scene model. As we know, there are some previous 3D Dense Mapping Systems, like Kinect Fusion. Kinect Fusion uses ICP to do registration between adjacent two frames. And ICP will fail when scan some flat area. However, Kinect Fusion doesn’t do any local or global optimization for camera poses. During scan, the reconstruction will drift because of accumulated errors. In our work, we used Orb-Slam to help us estimate camera pose of each frame. Because of orb-slam’s local and global optimization, we can avoid obvious drift when scan an indoor scene. We use an integration and re-integration framework to handle changing camera poses. Once a camera pose changes because of optimization, the system will do re-integration process to remove data influenced by old camera pose from scene volume and integrate data based on new pose to the scene volume.

 

Code:
	SLAMRecon

Data:

Vocabulary (138MB)

Example Data (1.38GB)

Document:
	SLAMRecon.pdf (1.42MB)

Authors' pages:

Hao's page

Huayong's page

Guowei's page

[1] Victor Adrian Prisacariu, Olaf Kähler, Ming-Ming Cheng, Carl Yuheng Ren, Julien P. C. Valentin, Philip H. S. Torr, Ian D. Reid, David W. Murray: A Framework for the Volumetric Integration of Depth Images. CoRR abs/1410.0925 (2014)

[2] Richard A. Newcombe, Shahram Izadi, Otmar Hilliges, David Molyneaux, David Kim, Andrew J. Davison, Pushmeet Kohli, Jamie Shotton, Steve Hodges, Andrew W. Fitzgibbon: KinectFusion: Real-time dense surface mapping and tracking. ISMAR 2011: 127-136

[3] Matthias Nießner, Michael Zollhöfer, Shahram Izadi, Marc Stamminger: Real-time 3D reconstruction at scale using voxel hashing. ACM Trans. Graph. 32(6): 169:1-169:11 (2013)

[4] Raul Mur-Artal, J. M. M. Montiel, Juan D. Tardós: ORB-SLAM: A Versatile and Accurate Monocular SLAM System. IEEE Trans. Robotics 31(5): 1147-1163 (2015)

 

We would like to acknowledge our grants: National 973 Program(2015CB352501), NSFC (61232011, 61572507, 61532003), Shenzhen Knowledge innovation program for basic research (JCYJ20150402105524053).