English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

2D forward-looking sonar is a crucial sensor for underwater roboticperception. A well-known problem in this field is estimating missinginformation in the elevation direction during sonar imaging. There are demandsto estimate 3D information per image for 3D mapping and robot navigation duringfly-through missions. Recent learning-based methods have demonstrated theirstrengths, but there are still drawbacks. Supervised learning methods haveachieved high-quality results but may require further efforts to acquire 3Dground-truth labels. The existing self-supervised method requires pretrainingusing synthetic images with 3D supervision. This study aims to realize stableself-supervised learning of elevation angle estimation without pretrainingusing synthetic images. Failures during self-supervised learning may be causedby motion degeneracy problems. We first analyze the motion field of 2Dforward-looking sonar, which is related to the main supervision signal. Weutilize a modern learning framework and prove that if the training dataset isbuilt with effective motions, the network can be trained in a self-supervisedmanner without the knowledge of synthetic data. Both simulation and realexperiments validate the proposed method.

Motion Degeneracy in Self-supervised Learning of Elevation Angle Estimation for 2D Forward-Looking Sonar