Optimizing Handheld 3-D Scanning Based on Fringe Projection Profilometry: Real-Time Registration and Performance Enhancement

Despite numerous advantages, fringe projection profilometry (FPP) technology is infrequently employed in handheld 3-D scanners due to its sensitivity to scanner motion and the intricate challenge of achieving real-time registration of dense point clouds without additional aids. This article presents a handheld scanner instrument based on FPP technology designed to overcome these limitations. In the scanning phase, a composite encoding-decoding algorithm that integrates sinusoidal fringe patterns with a random binary fringe pattern is proposed, which significantly reduces the number of projected patterns required while preserving high-quality 3-D imaging. For real-time point cloud registration, a two-step strategy comprising coarse and fine registration is adopted. In coarse registration, a novel 1-point-normal (1PN) method is proposed, which utilizes a single point and its normal vector to generate initial alignment hypotheses. This method enumerates a series of potential alignments and rapidly selects the optimal coarse registration from them through an evaluation algorithm operating under a parallel computing framework. For fine registration, point cloud quality is incorporated as a weight factor into the iterative closest point (ICP) optimization process to enhance the robustness and accuracy of registration. The effectiveness and stability of the proposed methods are validated through extensive experimental results, which have been successfully applied to our instrument of handheld 3-D scanner.