Depth-Based Full Forearm & Hand 3D Reconstruction with Skeletal Coupling (AR)

MSc assignment

Overview

Use a depth camera to reconstruct the entire forearm and hand in real time, detect keypoints, and perform rigging/skin-ning. In AR, the user’s real arm motion drives the virtually reconstructed arm (including the hand) to perform the same movements, enabling synchronized visualization and interaction.

Scope & Targets

  • Cover 5–8 subjects/scenes for end-to-end demonstrations.

  • Per session (capture → reconstruction → rigging → AR preview/export) ≤ 2–3 minutes.

  • Export true-scale 3D models of the forearm and hand (optional textures/UV/normals, with skeletal animation).

Core Method 

  • Acquisition & Pose: Depth + RGB; device calibration for camera poses.

  • Full-Arm Reconstruction: TSDF/point-cloud fusion → meshing and light cleanup (optional texturing).

  • Keypoints & Driving: Detect full-arm keypoints (wrist, elbow, finger joints, etc.) with temporal smoothing; 2D→3D back-projection or direct 3D.

  • Rigging & AR Coupling: Bind the skeleton to the reconstructed full-arm mesh; in AR, use the live skeleton to drive the virtual reconstructed arm; one-click export.

Application

  • Rehabilitation Assessment & Training: Real-time AR visualization of patient-specific arm motion with quantitative ROM, symmetry, and repeatability metrics, providing immediate visual feedback for elderly and post-operative upper-limb training