High-Speed 3D Measurement Based on RGB Multi-angle Time-Multiplexed Fringe Projection and Deep Learning

Recent advances in imaging sensors and DMD projection have broadened fringe projection profilometry (FPP) applications. However, conventional FPP, limited by strict one-to-one synchronization between pattern projection and image acquisition, suffers from the sensors’ inherent frame rate, hindering the capture of rapid 3D dynamics. To overcome this, we propose Triple-Frequency Color-Multiplexed FPP (TFCMFPP) that leverages multi-channel, multi-frequency, and multi-angle pattern encoding to significantly enhance temporal resolution. Within a single exposure, a high-speed DMD projects multiple fringe patterns onto the red, green, and blue channels, embedding multi-temporal information into one multiplexed color image. A deep learning framework then removes zero-order components and performs spatial-frequency decomposition to recover high-quality fringe images, achieving marked temporal super-resolution. Experiments on a rapidly rotating fan validate that our approach accurately reconstructs 3D surfaces under complex textures and high-speed dynamics, effectively surpassing traditional hardware limitations. Compared with existing methods, our approach greatly improves temporal resolution, reduces spectral aliasing, and enhances measurement flexibility, offering an efficient solution for ultra-high-speed 3D measurement.