Physics-guided temporal interpolation for single-frame fringe-projection 3D imaging

Abstract

High-speed three-dimensional (3D) imaging is essential for observing dynamic scenes, but the temporal sampling density of fringe projection profilometry (FPP) is commonly limited by the frame rate of the cameras. This paper presents a temporal interpolation method for single-frame fringe-projection 3D imaging. A stereo structured-light system synchronously captures vertical fringe-image sequences and reconstructs the corresponding depth-image sequences. An interpolation network then jointly fuses the fringe images and depth maps to infer intermediate 3D results. A Newton’s cradle experiment demonstrates that the proposed method recovers a temporally denser 3D sequence between two measured frames while preserving the main geometric structures. The method computationally improves the visualization of dynamic 3D evolution without changing the native acquisition rate of the cameras.