Presentation Information

[8a-N205-3]Artificial Intelligence for Global Optimization of Blind 3D Reconstruction from Time-of-Flight Cameras

〇Keith Dillon1 (1.Univ of Mississippi Med Ctr)

Keywords:

3D reconstruction,Time-of-flight camera,Artificial Intelligence

Accurate 3D reconstruction is important goal for robotics, AR/VR, and autonomous navigation, yet remains challenging due to uncontrolled environments. Time-of-flight (ToF) cameras provide depth information which simplifies the 3D reconstruction problem, yet this new information is subject to artifacts due to motion, material reflectances, and multipath interference. Traditional methods rely on controlled calibration procedures, which is itself a challenging engineering task, and further degrades over time as sensor parameters change. In this research we describe an approach to blind 3D reconstruction that combines AI-based assessment of the scene with local optimization for accurate reconstruction. This enables global optimization that does not require accurate ground truth in calibration or a good starting point for parameters. The approach can be used for initial calibration as well as continual online calibration and reconstruction.