Innovative branch-cut method for phase unwrapping based on modified fruit fly optimization algorithm

In fringe projection profilometry (FPP), the branch-cut algorithm is a classical method used for spatial phase unwrapping. The traditional Goldstein tends to create closed areas, which can result in islanded areas and severely degrade the reconstruction accuracy. Excessively long branch-cuts significantly increase the occurrence of islanded areas. Hence, minimizing the total length of branch-cuts is an effective strategy to prevent their formation. We employ a modified fruit fly optimization algorithm (FOA) to find the shortest branch-cuts. The FOA offers inherent advantages of having few parameters and being straightforward to implement. Furthermore, we integrate both the Roulette Wheel Selection (RWS) mechanism and the simulated annealing (SA) algorithm into the FOA framework, whereby the RWS enhances the diversity of solutions and the SA prevents the algorithm from falling into local optima. During the experimental phase, we conducted comprehensive comparisons with other state-of-the-art methods. The comparative results demonstrate that the proposed FOA integrated with RWS and SA exhibits superior noise immunity. Furthermore, it achieves higher computational efficiency while attaining the global optimum solution, significantly improving the efficiency of object surface reconstruction.