NAS-assisted single-shot real-time 3D measurement

Due to the advantages of non-contact, low cost and high accuracy, Fringe Projection Profilometry (FPP) has been widely used in different kinds of fields for robust three-dimensional measurement. However, limited by the filter operations in Fourier transform profilometry and multiple fringe patterns requirement in phase shifting profilometry, it’s difficult for traditional methods to achieve high-accuracy 3D reconstruction for moving objects. Meanwhile, because of the large amount parameters of neural networks and the complexity of network structures, it is time-consuming to perform one interface and hard to achieve real-time processing for current deep learning based FPP methods. In this paper, we proposed a single-shot method based on Neural Architecture Search (NAS) technique for the lightweight network processing. The experimental results demonstrate that the proposed method can achieve the reconstruction rate of up to 32 fps in high-accuracy and realize real-time 3D measurement for dynamic scenes.