Deep Learning Architectures for Navigation Using Forward Looking Sonar Images

Jose E. Almanza-Medina,Benjamin Henson,Yuriy V. Zakharov

doi:10.1109/access.2021.3061440

Jose E. Almanza-Medina, Benjamin Henson + Show 1 more

Open Access

https://doi.org/10.1109/access.2021.3061440

Copy DOI

Abstract

This paper investigates the use of supervised Deep Learning (DL) networks to process sonar images for underwater navigation. State-of-the-art DL techniques for micro-navigation using sequences of optical images have been adapted to work with sonar images. Specifically, the DL networks estimate the Forward-Looking Sonar (FLS) motion in three degrees of freedom corresponding to $x$ - and $y$ -translation and rotation around $z$ -axis. The state-of-the-art DL architectures and a proposed new architecture are investigated for motion estimation. They are trained using images generated by a FLS simulator. The data sets are made using pairs of consecutive images associated with labels that represent the motion of the sonar platform between images. The results show the effectiveness of using the DL architectures, which can provide millimeter accuracy for translation motion and below 0.1° for rotation motion between two consecutive sonar images. Examples of trajectory estimation and mosaic building using simulated and real sonar images are also presented.

Highlights

T HE use of non-piloted underwater vehicles has become an essential tool in exploration and surveying of underwater environments [1]
The work presented in this paper is an attempt to use Deep Learning (DL) approaches for trajectory estimation using sonar images
The basis of this work is to use large volumes of synthetic images generated by a sonar simulator to train the DL networks, and apply the already trained network to real data

Summary

Introduction

T HE use of non-piloted underwater vehicles has become an essential tool in exploration and surveying of underwater environments [1]. Autonomous underwater vehicles (AUVs) and remotely operated underwater vehicles (ROVs) alleviate the dangers that humans are exposed during explorations. Accurate navigation of these vehicles is required to succeed in their tasks [2]. The navigation problem can be addressed on a large scale (macronavigation) and on a small scale (micronavigation). Technologies for macronavigation such as global positioning system do not work underwater as they do in land-based applications since radiowaves are highly attenuated when passing through water bodies [4]. The use of undersea acoustic beacons can be used for underwater navigation [5], but it makes the system complexity and cost high [6] and they have a limited operation area [7]

Objectives

Findings

Discussion

Conclusion