Volleyball Ball Tracking & Replay System

From three camera views to a reconstructed ball trajectory.

Three cameras → one trajectory
Multi-camera volleyball tracking conceptA volleyball follows an arc over a court. Three camera sight lines follow its position, while a marker shows its projection onto the ground. This is a conceptual illustration, not recorded tracking data.CAM 01CAM 02CAM 03
Conceptual illustration · multi-view triangulation

Overview

Developed a multi-camera volleyball tracking and replay system. It brings together ball detection, spatial reconstruction, and animation to describe a ball’s movement relative to the court.

Problem

A single video view shows a ball’s position in an image, but does not directly describe where it is in three-dimensional space. Understanding a trajectory relative to court boundaries requires information from multiple views.

Technical approach

The system follows three stages: detect the ball across three video streams, estimate its position through multi-view triangulation, and visualize the reconstructed trajectory in a replay.

  1. 01 Video streams → ball detection
  2. 02 Multiple views → position estimation
  3. 03 Ball trajectory → replay animation

Detection

YOLO-based detection identifies the volleyball across three video streams. Those detections provide the image-space observations for the position reconstruction stage.

Triangulation

Multi-view triangulation estimates the ball’s position relative to the court. Position analysis supports in/out decisions by relating the reconstructed position to court boundaries.

Replay generation

Blender-based replay animations visualize the ball trajectory, turning the reconstructed positions into a replay of its movement.

Technologies

  • Python
  • YOLO
  • Computer vision
  • Blender