- Four primary types, Passive Stereo, Active Stereo, Structured Light, Time of Flight
- Must have very small form factor to be useful on mobile, and operate with limited computing resources
- Range (distance) of operation from gesture recognition to environmental scanning requires tradeoffs
- Primarily limited to indoors
The History of Mobile AR 1968 - 2009
- 1968 Ivan Sutherland’s Sword of Damocles VR HMD
- 1973 Motorola demos the first mobile phone
- 1993 NAVSTAR GPS launched
- 1994 Reality-Virtuality continuum and augmented reality formally defined
- 1997 Philippe Kahn shares the first picture taken by a phone (he invented the camera phone)
- 1999 ARToolKit is presented and available as open source
- 2007 iPhone is released!
- 2009 SLAM running in real-time on iPhone
(2013) Real-time Motion Tracking on a Cellphone
- Fused intertial measurement unit (IMU) readings with camera tracking
- Existing methods at the time where based on “global shutter” cameras, but phones use “rolling-shutter”
- Novel approach accomodated rolling-shutter cameras, IMUs, and extended Kalman filter (EKF)
- Results showed very low error, and the addition of GPS could be used as a correction factor
- There are a range of approaches to this problem!
- A Homography is one approach, and is a way of relating two co-planar images
- Requires a sequence of images with translation, rotation does not provide a solution
- Solutions must account for “virtual planes” i.e. a plane that isn’t attached to something real
- Sometimes includes a method called “Random Sample Consensus (RANSAC)" which tries to eliminate “outliers” and find a model that fits “inliers”
- However this algorithm deteriorates when the sample data contains ~%50 “outliers”