An improved spatiotemporal correlation method for high-accuracy random speckle 3D reconstruction

Abstract The single-shot random speckle pattern method can be used for dynamic scene 3D reconstruction. However, the homonymous points search process based on spatial correlation is sensitive to changes in factors such as illumination, perspective, and curvature, which results in low reconstruction accuracy. This paper proposes an improved random speckle 3D reconstruction method with enhanced reconstruction accuracy. To accurately search homonymous points, the proposed method employs a novel spatiotemporal correlation model simultaneously with a subpixel interpolation strategy. Analysis conducted of the reconstruction accuracy associated with correlation region size and the number of patterns to be projected based on the model indicate that using a few (e.g., three) speckle patterns with an appropriate correlation size produces highly accurate results. Experimental results further verify that the proposed method is suitable for high-accuracy dynamic 3D shape measurement.