Underwater object classification using scattering transform of sonar signals

TL;DR

This paper demonstrates that the scattering transform, a wavelet-based nonlinear map inspired by CNNs, effectively classifies underwater objects from sonar signals, showing robustness to translation and rotation in both real and synthetic datasets.

Contribution

It introduces the application of the scattering transform to underwater sonar signal classification, highlighting its advantages over traditional CNNs in terms of invariance and interpretability.

Findings

High accuracy in classifying UXOs and synthetic objects

Robustness to translation, rotation, and impedance variations

Theoretical insights into scattering transform properties

Abstract

In this paper, we apply the scattering transform (ST), a nonlinear map based off of a convolutional neural network (CNN), to classification of underwater objects using sonar signals. The ST formalizes the observation that the filters learned by a CNN have wavelet like structure. We achieve effective binary classification both on a real dataset of Unexploded Ordinance (UXOs), as well as synthetically generated examples. We also explore the effects on the waveforms with respect to changes in the object domain (e.g., translation, rotation, and acoustic impedance, etc.), and examine the consequences coming from theoretical results for the scattering transform. We show that the scattering transform is capable of excellent classification on both the synthetic and real problems, thanks to having more quasi-invariance properties that are well-suited to translation and rotation of the object.