Polarization-adaptive scattering suppression method for underwater structured-light 3D reconstruction

During underwater structured-light 3D reconstruction, scattering by suspended particles reduces fringe contrast, induces phase distortion, and consequently degrades reconstruction accuracy. Most existing underwater descattering methods merely perform 2D image enhancement and fail to preserve fringe modulation information, rendering them unsuitable for structured-light detection. Furthermore, these methods require manual selection of background regions and rely on fixed polarization priors, exhibiting poor adaptability to non-uniform scattering fields. Accordingly, this paper proposes a polarization-phase separation and adaptive descattering method (PSAD) tailored for 3D reconstruction. First, temporal orthogonal decoupling is employed to eliminate fringe components prior to polarization estimation, thereby preserving the low-rank properties of the scattering field. Second, via low-rank sparse decomposition, polarization information of inhomogeneous backscattering is adaptively acquired without manual intervention. Finally, background illumination and medium transmittance are estimated through adaptive iteration to enable precise fringe recovery and phase calculation. Experimental results demonstrate that under varying turbidity levels, the PSAD method effectively suppresses underwater scattering while fully preserving fringe characteristics. It significantly optimizes point cloud quality, reduces the root mean square error (RMSE) of 3D reconstruction by up to 69.7%, and enables stable high-precision underwater 3D reconstruction in scattering scenarios.