Study on Performance of Marker Detection via Training Data Augmentation of Partial Distortion in Underwater Sonar Image

Abstract

In an underwater environment, mobile robot research uses sensors such as inertial navigation system and sonar to localize surrounding landmarks and to estimate the position of the robot. Unfortunately, when sonar image data are acquired with an underwater sonar sensor, different types of noise are caused in the underwater sonar image. This noise is one of the factors that reduces object detection performance. This paper suggests improving object detection performance through distortion and rotation augmentation of training data.