Sometime in 2002 and to no little extent thanks to Prof. Gudrun Klinker, I developed a profound interest in Augmented Reality. Until 2004 I attended all classes and lab courses offered on the topic and worked on a couple of related research projects, e.g. FixIt: An Approach towards assisting Workers in Diagnosing Machine Malfunctions.
At some point in 2003, Gudrun approached me with an offer to collaborate on a project at BMW Research, yielding the BMW Augmented Reality Roadmap, which doubled as my Bachelor's Thesis. I helped with the realization of a spin-off from the BMW AR Roadmap, project Ida: An AR System for Visualizing Deviations between the Real and Planned Shape of a Car Body.
I figure given this intro on how I was involved with AR, it is understandable that a video on facial feature detection, that Joystiq recorded at this year's GDC, got me all fired up. Tracking of real-world elements is key to all Augmented Reality systems, integration of reality and virtualizations usually being a primary concern.
There is a wealth of sensor- and marker-based techniques to aid with the detection of real-world objects, specifically their location and orientation. Though the holy grail of tracking is feature based tracking, that is tracking of objects without the need for modification of objects of interest.
Steve Perlman of motion capture studio Mova describes more-or-less markerless (aside from phosphorous make-up) tracking of facial features with impressive results.
GDC - Mova Contour Reality Capture Technology from Joy Stiq on Vimeo.
If all this AR talk made you curious, I suggest you head over to Gudrun Klinker's sites for her Introduction to Augmented Reality and Advanced Topics in Augmented Reality courses. Her slides should provide for a good introduction to the subject.
