Survey on Deep Learning Techniques Used for Classification of Underwater Sonar Images
6 Pages Posted: 25 Nov 2020 Last revised: 7 Jan 2021
Date Written: November 20, 2020
Object detection methods have been evolving as deep learning gives the major support for the classification of the dataset irrespective of the domain. In this paper, underwater sonar image target detection methods are reviewed using deep learning comparatively. Types of sonar-like side-scan sonar, forward-looking sonar, synthetic aperture sonar provide higher resolution images than optical images in low illuminated water. Recently research on underwater exploration using Automated underwater vehicles (AUV) mounted with sonar sensors are evolving to support long searches to find objects of interest. The object detection can help us classify mines, seafloor, ocean habitats, shipwrecks, and finding drowning victims. This paper specifies a survey on state-of-the-art deep learning techniques used for underwater object detection from Sonar images. The study shows that deep learning techniques using transfer learning in real datasets as well as in semisynthetic dataset helps improve the efficiency of object detection.
Keywords: Deep learning, sonar imagery, object detection, convolutional neural network, sonar image classification, feature extraction, Automated underwater vehicle(AUV).
Suggested Citation: Suggested Citation