Side-Scan Sonar Image Classification Based on Style Transfer and Pre-Trained Convolutional Neural Networks

Qiang Ge,Xianyu Zuo,Qian Zhang,Baojun Qiao,Lanxue Dang,Fengxue Ruan

doi:10.3390/electronics10151823

Qiang Ge, Xianyu Zuo + Show 4 more

Open Access

https://doi.org/10.3390/electronics10151823

Copy DOI

Abstract

Side-scan sonar is widely used in underwater rescue and the detection of undersea targets, such as shipwrecks, aircraft crashes, etc. Automatic object classification plays an important role in the rescue process to reduce the workload of staff and subjective errors caused by visual fatigue. However, the application of automatic object classification in side-scan sonar images is still lacking, which is due to a lack of datasets and the small number of image samples containing specific target objects. Secondly, the real data of side-scan sonar images are unbalanced. Therefore, a side-scan sonar image classification method based on synthetic data and transfer learning is proposed in this paper. In this method, optical images are used as inputs and the style transfer network is employed to simulate the side-scan sonar image to generate “simulated side-scan sonar images”; meanwhile, a convolutional neural network pre-trained on ImageNet is introduced for classification. In this paper, we experimentally demonstrate that the maximum accuracy of target classification is up to 97.32% by fine-tuning the pre-trained convolutional neural network using a training set incorporating “simulated side-scan sonar images”. The results show that the classification accuracy can be effectively improved by combining a pre-trained convolutional neural network and “similar side-scan sonar images”.

Highlights

As high-resolution, multi-purpose marine detection equipment, side-scan sonar is widely used in the ocean, lakes, and other bodies of water, and is currently the main technique for underwater target detection
In order to reduce the workload of staff, decrease the number of subjective errors caused by visual fatigue, and improve work efficiency, the automatic classification of side-scan sonar seafloor images has practical significance
It can be proved that the classification accuracy in the task of side-scan sonar image classification can be significantly improved by transfer learning

Summary

Introduction

As high-resolution, multi-purpose marine detection equipment, side-scan sonar is widely used in the ocean, lakes, and other bodies of water, and is currently the main technique for underwater target detection. It can quickly obtain large-area and highresolution acoustic images of the seafloor and, combined with seafloor image data from a small number of sampling sites, researchers can distinguish the different types of objects on the seafloor based on side-scan sonar images; this is useful for activities such as mine detection, seafloor mapping, marine ecosystem monitoring, and underwater rescues [1,2,3,4]. In order to reduce the workload of staff, decrease the number of subjective errors caused by visual fatigue, and improve work efficiency, the automatic classification of side-scan sonar seafloor images has practical significance

Methods

Results

Conclusion