The increasing need for accurate bathymetric mapping is essential for a plethora of offshore activities. Even though aerial image datasets through Structure from Motion (SfM) and Multi-View Stereo (MVS) techniques can provide a low-cost alternative compared to LiDAR and SONAR, offering additionally, important visual information, water refraction poses significant obstacles in delivering accurate bathymetry. In this article, the generation of manned and unmanned airborne synthetic datasets of dry and water covered areas is presented. These data are used to train models for correcting the geometric effects of refraction on real-world image-based point clouds and aerial images. Based on a thorough evaluation, important improvements are presented, indicating the increased accuracy and the reduced noise in the point clouds of the derived bathymetric products, meeting also the International Hydrographic Organization’s (IHO) standards.

对精确的测深图的日益增长的需求对于大量的海上活动至关重要。尽管与LiDAR和SONAR相比，通过运动结构（SfM）和多视图立体（MVS）技术获得的航空图像数据集可以提供低成本的替代方案，此外，它还提供了重要的视觉信息，但水折射在进行精确的测深时会遇到很多障碍。在本文中，介绍了干旱和水覆盖地区有人和无人驾驶的航空合成数据集的生成。这些数据用于训练模型，以校正折射对基于真实世界图像的点云和航拍图像的几何影响。在全面评估的基础上，提出了重要的改进，