docking camera system
this was my primary project during my internship at PulseMedica. the goal was to reliably detect the pupil in real time and calculate the eye’s physical position in 3D space so the system could align accurately for medical procedures.
i owned the project end-to-end, from early requirements gathering through design, implementation, and validation, with guidance and technical review from my supervisor. the work balanced computer vision accuracy with the reliability constraints of a medical device.
i built a Python pipeline using OpenCV to detect and track the pupil, then applied calibration and geometric methods to infer position. the workflow prioritized stability, accuracy, and fast, interpretable feedback for the operator.
features i implemented
pupil detection & tracking
designed and implemented real-time pupil detection and tracking using OpenCV. handled edge cases such as partial occlusion, variable lighting, and motion blur.
3D position calculation
computed the eye’s physical position in 3D space using camera parameters and geometric modeling. translated 2D pupil coordinates into real-world xyz coordinates to support precise mechanical alignment.
calibration system
built a calibration workflow to determine camera intrinsics and extrinsics. this enabled consistent conversion between pixel space and physical measurements, which was critical for downstream accuracy.
operator feedback interface
created a real-time visual overlay showing detection state, confidence metrics, and alignment guidance. designed for fast interpretation so operators could adjust positioning during setup.
design & project ownership
before implementation, i worked directly with product owners and stakeholders to gather requirements and clarify system constraints. based on these discussions, i produced a formal design review supported by documentation, diagrams, and UML-style representations of the system architecture and data flow.
i conducted a comprehensive analysis of multiple technical approaches, evaluating trade-offs in accuracy, robustness, and integration complexity. this analysis informed the final design decisions and ensured the system aligned with both software and mechanical constraints.
the project was left in a well-documented, extensible state, ready to be carried forward as additional mechanical milestones in the overall system were completed. future development could build directly on this foundation without rework.