We have presented a real-time, fully-dynamic, multi-camera, 3-D person tracking system. The system adopts an extended Kalman filter framework that reconciles the 2-D tracking process with high-level 3-D models. This stabilizes the 2-D tracking by directly coupling an articulated dynamic model with raw pixel measurements. Some of the demonstrated benefits of this added stability include: increase in 3-D tracking accuracy, insensitivity to temporary occlusion, and the ability to handle multiple people.