Case Study: Tunnel

Project Target & Shot Conditions

The aim of this project was to track a sequence shot inside a physical tunnel with entirely unknown camera configurations (no focal length data, sensor parameters, or lens checkerboard references available). Additionally, the solver had to reconstruct a clean 3D geometry hull of the tunnel.

Reconstructing structured tunnel geometry is crucial for visual effects integration, providing the exact surface matches needed for projection mapping, lighting reflections, and correct spatial occlusions of CG elements.

Adopted Workflow

1. Feature Extraction: Placing high-density markers on structural seams and curved edges of the tunnel walls.
2. Radial Undistortion: Solving a dynamic distortion profile to rectify curved aberrations and align space paths.
3. Focal Coordinate Solve: Executing a multi-group focal estimation solve to find sensor size and focal length parameters.
4. Geometry Modeling: Reconstructing the 3D tunnel cylinder and projecting static texture coordinates onto the low-poly mesh.

Solver Challenges & Solutions

Repeating Patterns: Tunnels contain identical structural loops that confuse auto-trackers. Manual tracking anchors were locked onto unique wall details, electrical conduits, and lights to ensure accurate parallax solve.

Lighting Shifts: Flickering tunnel lights break optical flow tracking. Shadow and highlight adjustments were baked onto utility plates to stabilize tracker contrast.