← ALL PROJECTS
Senior Design · Smead Aerospace, CU Boulder

Stereoscopic Glider Tracking System

A machine learning powered system that tracks student-built gliders in flight using a stereoscopic camera rig, returning real time position data without any physical markers on the aircraft.

My RoleML tracking algorithm, end to end
Core ModelMeta's SAM2
HardwareStereoscopic camera pair
ProgramSenior capstone, Aerospace Engineering

Overview

Our senior design team built a system to track student gliders during test flights using a stereoscopic camera setup instead of onboard sensors or physical markers. I was responsible for the full machine learning tracking pipeline: taking in the stereo camera feed and turning it into usable position data for the glider in flight.

What I built

  • Tracking algorithm. Wrote the complete ML powered tracking pipeline used to identify and follow the glider across frames.
  • SAM2 integration. Used Meta's Segment Anything Model 2 (SAM2) to segment and track the glider within each camera's field of view without requiring fiducial markers on the aircraft.
  • Stereo position estimation. Combined the tracked segmentation from both cameras in the stereo pair to compute the glider's position in 3D space over time.