AUV-Based Underwater 3-D Point Cloud Generation Using Acoustic Lens-Based Multibeam Sonar

Abstract

This study proposes a 3-D seafloor scanning method using sonar images obtained by an acoustic lens-based multibeam sonar (ALMS). ALMS is a forward-looking sonar and is generally used to implement real-time acoustic vision for autonomous underwater vehicles (AUVs). ALMS is originally not a 3-D sonar, and elevation information is eliminated when the 3-D world is mapped to the 2-D sonar image. The proposed method provides a unique analysis of sonar image geometry for extracting the missing elevation information. Different from conventional 3-D reconstruction methods, the proposed method can be continuously executed regardless of the existence of any remarkable features in the sonar image, which means that the proposed method can scan the seafloor during a full-time AUV operation. Once the 3-D seafloor data are obtained, the height of the seafloor can be used for the detection of suspected objects. Therefore, the proposed method enables long-range seafloor object detection, triggering succeeding processes involving the AUV approaching the detected object, and the start of a detailed investigation at a close range.