Relative Velocity and Position Estimates
Uses the Lucas-Kanade sparse optical flow algorithm to calculate relative hockey player speed and position. Uses feature tracking and back-tracking for match verification between frames. Finds homography between reference and current views and uses RANSAC to determine outliers. Clusters outliers using K-means clustering. It is able to determine the relative speed of the players, and their general locations.