Improved Satellite Two-Medium Photogrammetry for Shallow Water Bathymetry via Global Dense Matching Optimization
Abstract
To address the issue of unreliable bathymetric inversion in shallow water areas with weak textures caused by matching breaks in the traditional Semi-Global Matching (SGM) algorithm, an improved method based on line feature-constrained global optimization is introduced. The method constructs a global energy function consisting of a data term and a smoothness term, and employs the Graph Cuts algorithm to optimize the disparity map, effectively repairing matching voids while preserving underwater topographic edges. Experiments were conducted using WorldView-2/3 satellite stereo imagery covering typical reef areas including Ganquan Island, Qilianyu, and Saipan. Validation against high-precision in-situ depth datasets, including Airborne LiDAR Bathymetry (ALB), single-beam echo sounding (SNOAR), and Hawkeye shallow-water bathymetric data demonstrates that, compared with traditional SGM, the proposed method significantly increases the coverage of valid matching points in ultra-shallow waters (0-5 m). Specifically, in the challenging Saipan area with severe optical interference and steep terrain variations, the Root Mean Square Error (RMSE) of depth inversion is reduced from 4.92 m to 2.23 m. The results confirm the effectiveness of the proposed approach in enhancing the reliability of satellite two-medium photogrammetry for engineering applications.