Improving Sonar Image Patch Matching via Deep Learning

TL;DR

This paper introduces a deep learning approach using CNNs to improve sonar image patch matching accuracy, significantly outperforming classical keypoint methods and other learning algorithms.

Contribution

It demonstrates that CNN-based models trained on labeled sonar data can achieve high matching accuracy, surpassing traditional and alternative learning methods.

Findings

CNN achieves 0.91 AUC for matching decision

CNN achieves 0.89 AUC for matching score

Classical methods achieve 0.61 to 0.68 AUC

Abstract

Matching sonar images with high accuracy has been a problem for a long time, as sonar images are inherently hard to model due to reflections, noise and viewpoint dependence. Autonomous Underwater Vehicles require good sonar image matching capabilities for tasks such as tracking, simultaneous localization and mapping (SLAM) and some cases of object detection/recognition. We propose the use of Convolutional Neural Networks (CNN) to learn a matching function that can be trained from labeled sonar data, after pre-processing to generate matching and non-matching pairs. In a dataset of 39K training pairs, we obtain 0.91 Area under the ROC Curve (AUC) for a CNN that outputs a binary classification matching decision, and 0.89 AUC for another CNN that outputs a matching score. In comparison, classical keypoint matching methods like SIFT, SURF, ORB and AKAZE obtain AUC 0.61 to 0.68. Alternative learning methods obtain similar results, with a Random Forest Classifier obtaining AUC 0.79, and a Support Vector Machine resulting in AUC 0.66.