Automated Goalie: Ping Pong Ball Trajectory Prediction System
Computer Vision & Controls
Real-time system that tracks a ping pong ball with YOLOv8 on a Raspberry Pi and steers a servo-driven goalie by predicting 3D trajectories.
Highlights
Fine-tuned a YOLOv8 model on 1,000+ custom frames to detect a ping pong ball at around 2.7 ms per frame on a Raspberry Pi camera stream.
Converted 2D detections into 3D coordinates via calibrated pixel-to-world mapping and inverse-square depth estimation, then fit quadratic polynomials to predict landing positions.
Mapped predicted landing points to servo angles and streamed commands over MQTT, achieving accurate saves on several test positions with servo alignment within 5–10 degrees of optimal.