Tracking is the analysis of camera and object movements in footage to accurately composite CGI elements into the scene.
Technical Details
2D Tracking analyzes motion in the image plane with at least four tracking points per object, while 3D Tracking (Camera Tracking) reconstructs the complete camera movement in space, tracking 50-200 feature points simultaneously. Planar Tracking captures flat surfaces by analyzing texture patterns and contrast gradients. Modern tracking software operates with sub-pixel accuracy and can compensate for motion blur up to 10 pixels. Object Tracking follows three-dimensional bodies by combining multiple camera angles, while Facial Tracking analyzes up to 468 facial points in real-time.
History & Development
The first commercial motion tracking system was developed in 1990 by Hammerhead Productions for "Terminator 2". In 1993, Industrial Light & Magic established 3D camera tracking as the standard for dinosaur integration with "Jurassic Park". In 1995, 2d3's "boujou" revolutionized automatic feature detection, followed by SynthEyes (2005) and PFTrack (2008). Since 2010, GPU-accelerated algorithms have enabled real-time tracking at over 240fps, while machine learning has improved tracking accuracy by an average of 30% since 2018.
Practical Application in Film
"Gravity" (2013) used LED wall tracking with 2,048 reference markers for Sandra Bullock's space sequences. "The Revenant" (2015) combined handheld camera tracking with natural landscape markers for CGI bear integration. Typical workflows begin with marker placement on set (20-30 tracking dots per square meter), followed by automatic solve calculation and manual refinement. The tracking achieves solve errors below 0.3 pixels. Match Moving requires additional HDR photographs for lighting reconstruction, while set extensions necessitate precise calibration of lens data (focal length, distortion).
Comparison & Alternatives
Tracking differs from rotoscoping by automated motion capture instead of manual tracing. Stabilization uses tracking data for image correction, while match moving uses it for CGI integration. Virtual Production with LED stages is increasingly replacing classic tracking with real-time camera tracking via OptiTrack or Vicon systems. AI-based tracking (since 2020) works marker-free but achieves only 85% of the precision of traditional marker systems at 60% of the processing time.
Current News
Modern cameras like the announced Nikon Z90 integrate AI-based tracking systems directly into the hardware. These automated processes can track object movements in real-time, thereby facilitating shooting and subsequent post-production. The combination of high-resolution sensors and intelligent tracking algorithms significantly reduces the manual effort involved in match moving work.