THE CHALLENGE
Match footage is full of useful information, but reviewing it manually is slow. Project REDACTED creates a clearer picture of how a match unfolded.
JUICE AUTOSCOUT · COMPUTER VISION
AUTOMATED MATCH ANALYSIS
A lightweight scouting system that turns FTC match video into reliable movement and shot data—built to make post-match review faster and more useful.
TECHNICAL WHITE PAPER · JULY 2026
A Lightweight, Perspective-Calibrated Multi-Robot Tracking and Heuristic Shot Analysis System for Competitive Mobile Robotics ↗- Author
- Siddharth Ray
- Published
- July 18, 2026
Abstract
This paper presents the computer-vision core of AutoScout, a lightweight monocular video-analysis system designed to extract tactical analytics from and virtually reconstruct FIRST Tech Challenge match footage. The system pairs a season-robust robot tracker with a game-specific shot-analysis module, using perspective calibration, foreground segmentation, and deterministic identity preservation through multi-robot contact. The result is a transparent, debuggable, and computationally efficient tracker intended for resource-constrained teams.
THE APPROACH
It follows all four robots and key scoring moments from calibrated video, then translates that activity into an easy-to-review match record.
THE OUTCOME
Teams can move from raw footage to meaningful scouting context without relying on heavyweight infrastructure or a specialized workstation.
Focus areas
Computer VisionRobotics AnalyticsPythonOpenCVData Visualization