A Structured-Light Dual-Frequency Heterodyne Phase Unwrapping Method Based on Region-of-Interest Constraint and Dual-Frequency Consistency Optimization

In dual-frequency heterodyne structured-light 3D measurement, fringe-order misjudgment often occurs near object edges, hole boundaries, and local low-reliability regions, which may further lead to absolute phase jumps and abnormal 3D reconstruction results. To address this problem, this paper proposes a phase-unwrapping method based on region-of-interest (ROI) constraint, coarse estimation using the synthetic wrapped phase, dual-frequency consistency residual optimization, and local spatial regularization. The method first automatically extracts the target ROI using a binary mask, thereby excluding invalid background regions and reducing unnecessary computation. A synthetic wrapped phase is then constructed from the two wrapped phases, and a coarse absolute phase is obtained using the long synthetic period. The initial fringe order is estimated by rounding rather than the conventional floor operation, reducing systematic fringe-order misjudgment near integer boundaries. On this basis, a dual-frequency consistency residual is constructed to perform local candidate search and reliability evaluation. Reliable-anchor constraints and local spatial regularization are subsequently introduced to correct abnormal fringe orders in low-reliability regions and suppress error propagation near edges and holes. Experimental results on objects with complex boundaries show that the proposed method effectively suppresses phase jumps and improves the continuity of the unwrapped phase and reconstructed point cloud. Accuracy verification using a three-layer step block showed mean height errors of 0.0230, 0.0320, and 0.0370 mm for the three layers, respectively, with corresponding standard deviations of 0.00852, 0.00880, and 0.00963 mm, all below 0.01 mm. These results demonstrate that the proposed method can effectively suppress fringe-order-induced phase jumps while maintaining high accuracy and stability in structured-light 3D reconstruction.