Search

Learning three-dimensional (3D) shape representation of an object from a single-shot image has been a prevailing topic in computer vision and deep learning over the past few years. Despite extensive adoption in dynamic applications, the measurement accuracy of the 3D shape acquisition from a single-shot image is still unsatisfactory due to a wide range of challenges. We present an accurate 3D shape acquisition method from a single-shot two-dimensional (2D) image using the integration of a structured-light technique and a deep learning approach. Instead of a direct 2D-to-3D transformation, a pattern-to-pattern network is trained to convert a single-color structured-light image to multiple dual-frequency phase-shifted fringe patterns for succeeding 3D shape reconstructions. Fringe projection profilometry, a prominent structured-light technique, is employed to produce high-quality ground-truth labels for training the network and to accomplish the 3D shape reconstruction after predicting the fringe patterns. A series of experiments has been conducted to demonstrate the practicality and potential of the proposed technique for scientific research and industrial applications.

Single-shot 3D shape reconstruction integrating structured light and deep learning has drawn considerable attention and achieved significant progress in recent years due to its wide-ranging applications in various fields. The prevailing deep-learning-based 3D reconstruction using structured light generally transforms a single fringe pattern to its corresponding depth map by an end-to-end artificial neural network. At present, it remains unclear which kind of structured-light patterns should be employed to obtain the best accuracy performance. To answer this fundamental and much-asked question, we conduct an experimental investigation of six representative structured-light patterns adopted for single-shot 2D-to-3D image conversion. The assessment results provide a valuable guideline for structured-light pattern selection in practice.