Visión Computacional para seguimiento posicional de jugadores y cuantificación de saltos en voleybol

Student: Benjamín Araya
Degree: Undergraduate Thesis (Memoria de Título)
My Role: Primary Advisor (Profesor Guía)
Institution: Universidad Diego Portales, Escuela de Informática y Telecomunicaciones
Location: Santiago de Chile
Status: In progress

Overview

This thesis develops a computer vision system for positional tracking of volleyball players from conventional video footage. The pipeline integrates person detection with YOLO models, multi-object tracking via ByteTrack, automatic court calibration, planar homography, volleyball-specific spatio-temporal rules, and identity consolidation to transform raw detections into auditable top-down trajectories. On top of these trajectories, the system generates heatmaps, tactical visualizations, and jump-related metrics. The work targets coaches and technical staff who lack access to expensive commercial tracking systems, dedicated cameras, or wearable sensors — offering a functional, extensible baseline for spatial-occupancy analysis and jump quantification.

Key Technologies

• Object detection: YOLOv8, YOLO11, and YOLO12 (comparative evaluation)
• Multi-object tracking with ByteTrack
• Automatic court calibration via temporal median, HSV/Lab masks, and geometric hypothesis validation
• Planar homography for top-down (bird’s-eye) projection
• RedLock: custom heuristic layer for identity stabilization, side-of-court consistency, and identity inheritance
• Temporal post-processing for jump-event quantification from trajectory data

Impact

Provides an auditable, low-cost alternative to commercial player-tracking systems for volleyball technical staff. On 99 manually annotated positions, the system achieved 87.88% strict zonal accuracy and 96.97% accuracy with zone adjacency, supporting its use for spatial-occupancy analysis. For aggregate jump counting per clip, the system reached F1 = 0.846 (precision 0.841, recall 0.851) over 87 manually observed jumps, showing consistent estimation of overall jump load — a key metric for load management and injury prevention in players who repeatedly execute net actions.