Underwater Terrain Map Building Based on Depth Image Using Multi-Beam Sonar Sensor

Abstract

In order to perform work in an underwater environment, it is important to monitor the situation by mapping. However, the use of sensors in the underwater environment is more limited than in ground environments. Sonar sensors are mainly used for mapping work in underwater environments. The underwater environment data collected by the multi-beam sonar sensor is mostly composed of flat terrain data, which is difficult to extract robust features, and thus has challenging conditions in mapping research. To solve this challenge, we propose a method of building an underwater terrain map using a 3D point cloud collected from a multi-beam sonar sensor. The proposed method uses the HardNet-based image matching method to register the depth images and build a global underwater terrain map. Experiments were conducted based on point cloud-based underwater terrain data, and the high accuracy of registration was confirmed.