Efficient Fringe Projection Absolute Phase Recovery Based on Dynamic Fringe Convolution

In fringe projection profilometry, efficient and accurate recovery of the absolute phase of the measured object is of great significance for the reconstruction of 3D information. The deep learning technique breaks through the limitation that the traditional fringe projection profilometry cannot simultaneously take into account the measurement efficiency and accuracy. However, during the prediction process of the network, the absolute phase recovery error is mainly distributed in the object edge region and is difficult to avoid, which leads to the degradation of reconstruction accuracy. In order to strengthen the feature extraction ability of a deep learning network on a fringe image and reduce the object edge error. In this paper, we propose a dynamic fringe convolution for fringe image features in fringe projection profilometry. The DFC-Net is constructed to transform a single fringe image into the numerator-denominator terms corresponding to the wrapped phase and the fringe order, and then the corresponding absolute phase is directly obtained. Efficient and accurate absolute phase recovery of a single fringed image is realized without the aid of the remaining modes. It is experimentally verified that the absolute phase recovered by the proposed method maintains high accuracy under the interference of noise and different illumination fringes.