This introductory deck presents Lumina Bone in four parts: about me, research interests, the proposed research project — Lumina Bone — and openness to other research projects.
About Me
- Robotics — Founder of FTC robotics team Mercury (Instagram: @gnce_mercury).
- Research — Research intern at the Multi-scale Medical Robotics Center, Hong Kong: assembly design & testing.
- DECA — Selected to compete at the international level.
- Swimming — Varsity member; Top 20 NE, 13–14 age group (2025).
- Piano — Classical training since childhood.
Research Interests
Applications
- Develop and create surgical robotics to assist surgeons in delivering better outcomes.
- Create robotic solutions for environmental challenges, such as plastic waste management.
Technical Focus Areas
- Robot control systems and advanced algorithm development.
- Computer vision and vision-based intelligence.
- Creating and building robot hardware.
Proposed Research: Lumina Bone
- Objective: Design a low-cost 6mm 3D bone endoscope using near-light photometric stereo.
- Innovation: Deliberate off-axis LED placement maximizes shading cues on rigid bone surfaces.
- Workflow: Sequential PWM illumination captures distinct frames to solve for per-pixel surface normals.
- Impact: Enables real-time, high-accuracy 3D topography reconstruction for orthopedic and ENT surgery.
Technical Solution: Dual-LED Photometric Stereo
Hardware Architecture
- Distal tip: 6mm diameter housing a 2mm camera and 2 micro-LEDs.
- LED geometry: 1.8–2.5mm radial offset maximizes cosine-based shading gradients.
- Control: Sequential PWM switching (<100ms) for flicker-free multi-light capture.
Algorithm Pipeline
- Lambertian model: Accounts for
1/r²light fall-off on rigid bone surfaces. - Normal solver: Per-pixel tilt estimation via least-squares from dual frames.
- Integration: Poisson reconstruction stitches surface normals into 3D topography.
Expected Deliverables
- Functional prototype — LuminaBone 6mm distal tip integrated with PWM-controlled dual-LED electronics.
- Quantitative analysis — Results demonstrating accuracy gains of off-axis sequential LEDs vs. standard coaxial systems.
- Documentation & media — Comprehensive technical report and professional demo video for conference presentation.
- Open-source software — Open-source near-light photometric stereo pipeline with depth integration software.
Current Progress
Depth-sensing methods and their typical accuracies:
| Method | Typical accuracy |
|---|---|
| Stereo Camera | < 0.5 mm |
| Structured Light | < 0.5 mm |
| Time-of-Flight | 5–20 mm |
| Shape from Motion / SLAM | 2–10 mm |
| Monocular Deep Learning | 3–8 mm (soft tissue); unknown on bone |
| Near-Light Photometric Stereo (Lumina-Bone) | 1–3 mm (est. clinical); < 1 mm (dry bench) |
A second Current Progress slide was image-only, showing visual progress from the project; those images are not mirrored here.
Ending
The Lumina Bone project is a self-proposed project aimed at sharpening my research skill set and experience. It is entirely at the concept-and-proposal stage and needs further tuning and improvement.
I am open to other research projects that align with my research interests and can potentially solve real-world problems.