Skip to content

Austin Stone

About

Austin is a PhD candidate in Electrical and Computer Engineering at Brigham Young University, advised by Dr. Cammy Peterson in the MAGICC Lab. His research focuses on thermal infrared (LWIR) perception for GPS-denied navigation, including thermal-inertial odometry, robust visual outlier rejection, uncertainty-aware optical flow, and thermal infrared single-object tracking.

Austin earned a BS in Mechanical Engineering from BYU with minors in Computer Science and Mathematics and began his research career as an undergraduate research assistant in the MAGICC Lab. He also works as a Senior Drone Algorithm Engineer at Teledyne FLIR on the Rogue One program, developing visual-inertial odometry and sensor fusion systems for an EO/IR imaging pipeline. Outside of research, he enjoys tinkering with home automation and self-hosted infrastructure.

Research

Thermal-Inertial Odometry

Austin develops real-time monocular thermal-inertial odometry for high-speed, GPS-denied flight. The system fuses a longwave infrared camera, high-rate IMU, laser range finder, barometer, and magnetometer in a fixed-lag factor graph. It runs onboard an NVIDIA Jetson Xavier NX and has sustained closed-loop quadrotor flight at 30 m/s with less than 2% drift over kilometer-scale trajectories.

Robust Outlier Rejection

His work on visual outlier rejection introduces IMU-aided minimal solvers and on-manifold essential-matrix refinement for thermal-inertial odometry. In GPS-denied flight tests at speeds up to 30 m/s, the extended 2-point USAC method with algebraic refinement reduced drift rate by 36% compared with 7-point RANSAC.

Uncertainty-Aware Optical Flow

Austin developed RAFT-U, an extension of RAFT that jointly predicts dense optical flow and calibrated per-pixel uncertainty. Incorporating this uncertainty into a factor-graph visual-inertial odometry system reduced trajectory error across infrared and electro-optical UAV flight logs. This work was accepted to IEEE/RSJ IROS 2026.

Thermal Infrared Object Tracking

Austin's thermal infrared tracking research includes Ember, a one-stream vision-transformer tracker with a distributional bounding-box head, absence-calibrated confidence, and a memory bank for target reacquisition after occlusion.

Publications

Experience

  • Teledyne FLIR, Senior Drone Algorithm Engineer, July 2026–present
  • Teledyne FLIR, Senior Algorithm Intern, January 2022–July 2026
  • BYU MAGICC Lab, Undergraduate Research Assistant, November 2020–April 2022