Pmomo: Projection Mapping on Movable 3D Object
summary
System tracks movable rigid 3D objects by combining an AHR sensor and the connect depth map, registering a low-density point cloud (~300 points) to depth frames with an evolutionary, semi-global optimizer and adaptive step sizing. It culls occluders via a facet-covering method and predicts motion to mask input delay (about 60 ms sensor plus ~30 ms compute/projection; per-frame computation ≈16 ms). Evaluation uses 200,000 frames from 10 volunteers, works best around 20 cm/s velocity, <20 cm/s^2 acceleration and <15% visible occlusion, and currently only supports rigid objects.
AI-written summary · about
chapters (10)
- 0:00Introduction and setup
- 1:57Project overview PMOMO
- 3:47Challenges in dynamic projection
- 5:36System design and sensors
- 7:46Tracking complex 3D objects
- 10:15Occlusion handling method
- 12:05Performance and real-time results
- 14:58Application examples
- 15:43Limitations and future work
- 16:03Q&A session
related talks
06mMobileportation: Nomadic Telepresence for Mobile Devices
06mCoolMoves: User Motion Accentuation in Virtual Reality
18mMatchPoint: Spontaneous Spatial Coupling of Body Movement for Touchless Pointing
28mVibing Vision: how we built a pose tracking art installation in two weeks
10mmap mop
21mQoom: An Interactive Omnidirectional Ball Display
06m3PP-R: Enabling Natural Movement in 3rd Person Virtual Reality
09mPhyxel: Realistic Display of Shape and Appearance using Physical Objects with High-speed ...