Researchers propose Track2Map, an online 3D Gaussian Splatting pipeline that jointly optimizes camera trajectory and 3D scene representation directly from surgical video. This approach enables robust reconstruction in robot-assisted minimally invasive surgery even when accurate camera trajectory priors are missing or noisy.
- The system utilizes track-anchored deformation initialization via dense 2D point tracks to stabilize optimization amid tissue motion.
- Track statistics disentangle camera motion from scene deformation by detecting static periods and reducing drift during incremental mapping.
- Experiments on the StereoMIS dataset demonstrate improved reconstruction quality and camera trajectory accuracy compared to competing SLAM methods.
The method effectively functions as a Simultaneous Localisation and Mapping (SLAM) system, addressing limitations of offline pipelines that depend on reliable kinematic priors.