Autonomous underwater vehicles (AUVs) are widely used to explore the mysterious underwater world. Following along predefined spatial paths, AUVs are able to gather valuable seabed information in a designated area by recruiting sonar suites. The acoustic data collected by AUVs are usually in the type of point cloud with range information. Hence, how to reconstruct the topography of seabed via 3D point cloud data is the key to build the 3D seabed map. In order to address the problem, the paper presents a practical mesh method to achieve an accurate reconstruction of seabed surface from raw sonar records. Sonar data processing consists of three stages: point clearing, point normal and 3D surface reconstruction. Simulation results show the effectiveness of the proposed approach.