Underwater Three-Dimensional Reconstruction Based on Sub-regional Processing of Forward-Looking Sonar Images

Three-dimensional (3D) reconstruction plays a vital role in underwater exploration. However, forward-looking sonar technology captures 3D data in 2D images, which cannot be directly utilized for 3D reconstruction. In this paper, we present a novel method for underwater 3D reconstruction using an autonomous underwater vehicle (AUV) and sub-regional processing of sonar images. Firstly, we design an algorithm to extract pixel points based on the imaging characteristics of different regions in the sonar images, thereby dividing them into an extended highlighted frontier point set and a non-extended point set. Utilizing the sonar imaging principle, the non-extended highlight point set is directly recovered to generate a 3D point cloud according to the sonar imaging principle. For the highlighted extended boundary point set, the boundary extension of the highlighted part is analyzed to generate the edge line point cloud. Subsequently, the AUV poses are combined with the continuously generated point clouds from the sonar images to perform 3D point cloud reconstruction. Experimental results demonstrate that our method effectively enhances the utilization of sonar image data. Furthermore, we establish an underwater simulation experimental platform to validate the proposed method for reconstructing underwater environments. The results exhibit that our approach efficiently generates high-quality underwater 3D point clouds, yielding satisfactory outcomes.