CSE445Computer Vision
Tutorials/Object Tracking and Multi-Object Tracking/Foundations of Object Tracking and Motion Prediction
Back to seriesObject Tracking and Multi-Object TrackingNext tutorialDetection-to-Track Association: From SORT to ByteTrack
TR-01Beginner–Intermediate 2 Hours

Foundations of Object Tracking and Motion Prediction

Introduce object tracking as the estimation of an object’s state over time. Students learn how positions, velocities, bounding boxes, trajectories, and uncertainty are represented. The tutorial develops motion prediction from frame differencing and optical flow to constant-velocity models and Kalman filtering.

PythonNumPyOpenCVFilterPyMatplotlibYOLO26 observations

Code Structure Note

These notes are provided only to help students understand the code structure. Students should review, optimize, and adapt the code before using it in assignments, projects, or experiments.

Topics Covered
  1. 1Video as a temporal signal
  2. 2Frames and frame rate
  3. 3Object state
  4. 4Bounding-box center, width, and height
  5. 5Velocity and acceleration
  6. 6Trajectory and observation
  7. 7State prediction and measurement update
  8. 8Process and measurement noise
  9. 9Uncertainty
  10. 10Track initialization and lifecycle
  11. 11Single-object tracking
  12. 12Multi-object tracking

Learning Objectives

  • Differentiate detection, single-object tracking, and multi-object tracking
  • Represent position, size, velocity, identity, and uncertainty as track state
  • Explain prediction, observation, innovation, correction, and covariance
  • Apply a Kalman filter across missed and noisy observations

Key Takeaways

  • YOLO26 supplies observations; the motion model performs prediction
  • Optical flow is one motion cue, not a complete multi-object tracker
  • Uncertainty grows while observations are missing and contracts after correction
  • Track lifecycle rules determine when an identity is created, retained, or removed

Tutorial Infographic

Hands-On Exercises

  1. 1Visualize true, observed, predicted, and corrected positions
  2. 2Adjust process noise Q and measurement noise R
  3. 3Plot the covariance ellipse during an observation gap
  4. 4Document how the track recovers after detections return

Tools & Libraries

PythonNumPyOpenCVFilterPyMatplotlibYOLO26 observations
Back to seriesObject Tracking and Multi-Object TrackingNext tutorialDetection-to-Track Association: From SORT to ByteTrack