Objectives

• To retrive the 2D baseball trajectory in a television broadcast
• To reconstruct the 3D trajectory of the baseball using physical model

• Complex background containing objects similar to baseball
• Post video effects such as man-made score table
• Small size of baseball
• Motion blur of baseball due to slow frame rate
• White-on-white scenario: baseball overlap with white objects

Solution to 2D recognition

• Background subjection, color filtering in IUV space, morphological filtering, pattern recognition
• Spatiotemporal approach for global trajectory recognition

Solution to 3D reconstruction

• Aerodynamic equation for depth estimation
• Very accuracate, less than 0.1% error

Applications

• Baseball pitching analysis
• Baseball hitters training
• Virtual baseball games