High-Dynamic-Range 3-D Shape Measurement With Adaptive Speckle Projection Through Segmentation-Based Mapping

The 3-D shape measurement of high-dynamic-range (HDR) scenes based on structured light projection is challenging because the conventional projection without considering the reflection distribution of the tested surface usually causes over-exposure and over-dark areas simultaneously in the captured images, which will lead to the failure or low accuracy of 3-D reconstruction. In this article, we propose a novel HDR 3-D surface measurement method based on adaptive speckle projection with fully considering the reflection distribution. An efficient segmentation-based mapping strategy is established to speed up the generation of adaptive speckle patterns, which segments the estimated mask with optimal projection intensity into multiple subregions by the eight-connected standard to raise mapping accuracy and project gray-coded and sinusoidal fringe patterns to build the pixelwise mapping relationship between binocular cameras and the projector for each subregion, respectively. Then the pixelwise intensity of projected speckle patterns can be accurately adjusted according to optimal intensity. The generated adaptive speckle pattern is used to scan the tested object, and high signal-to-noise-ratio (SNR) speckle images can be obtained for accurate and complete 3-D measurement. Experimental results verify that the point cloud produced by our method has obvious advantages compared with the traditional methods in terms of measurement accuracy, whose coverage is as large as 92.26% and the average distance (AD) relative to ground truth (GT) is as low as 0.128 mm. Meanwhile, the number of projected patterns for mapping is decreased by a third compared with the widely used mapping strategy.