부산대학교 도서관

Fringe projection technology is a commonly used technique in optical 3-D measurement. In high-speed motion scenarios, due to image noise and the effects of object motion, projecting more fringe patterns for high-precision phase unwrapping is a common method, which can significantly reduce the frame rate of 3-D reconstruction. Deep learning techniques have been employed for high-precision phase unwrapping, but typically, these models have a large parameter and computation, making them difficult to integrate into real-time 3-D reconstruction systems. In this article, we first employ the lookup table (LUT) technique for rapid computation of dual-frequency phases. Second, we design a deep learning model with a parameter size of only 276 kb for high-precision phase unwrapping and quickly embed it into a real-time 3-D reconstruction system through 8-bit quantization without compromising accuracy. Furthermore, we utilize the calibration parameters of a real fringe projection profilometry (FPP) system to establish a corresponding virtual FPP system for rapid generation of data required for model training. Finally, we optimize the generation of point clouds by avoiding the computationally slow inverse matrix operation process. Experiments show that our model can achieve high-precision real-time 3-D reconstruction at a rate of 130 frames/s.