
Stanford Robotics Lab
3D Localization of a Medical Microrobot
At the Stanford Robotics Lab, this project builds the perception stack for a magnetically actuated millirobot designed for thrombectomy in stroke patients. The robot must be localized in 3D, in real time, from external camera views of a transparent vascular phantom.
The pipeline combines multi-camera calibration, AprilTag-based registration of the cameras against a 3D vessel model, and 2D-to-3D mapping of detections onto the vessel centerline. The tracking is engineered to stay reliable through refraction, specular reflections, and contrast-dye injections.
This work is aimed at vision-guided thrombectomy, where precision is measured in millimeters and outcomes are measured in lives.
OpenCVCamera CalibrationAprilTags3D ReconstructionReal-Time TrackingPython