Abstract: In the underwater Simultaneous Localization and Mapping (SLAM) system, the autonomous underwater vehicle (AUV) estimates its own pose while mapping the environment. The pose of the AUV can be predicted by dead reckoning, but navigation errors accumulate over time. Therefore, other sensors are needed to calibrate the status of the AUV. Multibeam Echosounder (MBES) can obtain bathymetric point cloud data containing environmental depth information, which is one of the most commonly used sensors for underwater SLAM. However, there are difficulties in data association when the overlap rate of underwater terrain data is low. Previous studies mostly focus on the Iterative Closest Point (ICP) and its variants, which have limited effect on the processing of sounding data when the overlap rate decreases. In this paper, an improved piecewise voting method is proposed, and we use the collected data to conduct two trials concerning the registration of bathymetric submaps. Experimental results show that the proposed method can register point clouds more effectively at low overlap rate compared with the commonly used methods.