High-resolution, real-time 3D measurement using gradient features and spatial-temporal speckle projection system

Abstract

Speckle projection profilometry (SPP) finds widespread application in industrial manufacturing and academic research owing to its benefits of non-contact, high speed and ease of integration. In this paper, we present a novel high-resolution, real-time 3D measurement technique realized through gradient-feature-based stereo matching and spatial-temporal speckle projection. In terms of algorithm, unlike traditional stereo matching methods that depend on grayscale information, we propose a new spatial-temporal stereo matching strategy based on horizontal and vertical gradient features as well as digital image correlation technique, enabling sub-pixel disparity estimation with high precision. On the hardware side, we developed an optical projection system capable of projecting temporally varying speckle patterns to encode the scene under test. Experiments are conducted to verify the feasibility and effectiveness of the proposed method. Our method achieves a root mean square error (RMSE) of 0.0621mm with two-frame speckle patterns (compared to 0.0835mm realized with the commonly used three-frequency four-step phase shifting method). In dynamic object experiments, our method achieves real-time 3D reconstruction at 27 frames per second. These experiments have validated the application prospects of the proposed method in achieving high-precision measurement with few frames.