An Improved SFS Method for Achieving Fast, High-Precision, and Widely Adaptable 3-D Reconstruction.

Different optical reconstruction methods have stringent requirements on the material or shape of the objects, and have poor generalization ability. To address this issue, this article proposes a stable and reliable target contour recognition algorithm based on omnidirectional illumination using shape from silhouette (SFS). It solves the problem of low signal-to-noise ratio of light signals reflected by materials with high reflectivity or transparency, and achieves performance that is comparable to the state-of-the-art deep learning networks. Moreover, this article solves the blind spot problem of the SFS algorithm in single-image fast reconstruction, and proposes a targeted growth visual hull (VH) method, which further enhances the reconstruction speed of the traditional SFS method while ensuring reconstruction accuracy. This article takes the reconstruction measurement of supermarket products as an example to demonstrate the value of the proposed method in practical applications.