Deep learning of submerged body images from 2D sonar sensor based on convolutional neural network

Abstract

Given the harsh working conditions such as high-speed flow rate, turbid watch, and steep terrain, it is a very challenging task to find submerged bodies in disaster site occurred at sea or river or for the military purpose. Therefore, if it is possible to utilize the unmanned robot, such as the USV(Unmanned Surface Vehicle) and UUV (Unmanned Underwater Vehicle) for the navigational operation of these special purpose, it has a great effect. Underwater ultrasound image information is pretty difficult to make the geometric modeling of submerged body due to heavy noise on its characteristics. This study presents the robust method of submerged body recognition based on the CNN(Convolutional Neural Network), which is one of the deep learning approach.