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

In the domain of point cloud registration, the coarse-to-fine featurematching paradigm has received substantial attention owing to its impressiveperformance. This paradigm involves a two-step process: first, the extractionof multi-level features, and subsequently, the propagation of correspondencesfrom coarse to fine levels. Nonetheless, this paradigm exhibits two notablelimitations.Firstly, the utilization of the Dual Softmax operation has thepotential to promote one-to-one correspondences between superpoints,inadvertently excluding valuable correspondences. This propensity arises fromthe fact that a source superpoint typically maintains associations withmultiple target superpoints. Secondly, it is imperative to closely examine theoverlapping areas between point clouds, as only correspondences within theseregions decisively determine the actual transformation. Based on theseconsiderations, we propose {\em OAAFormer} to enhance correspondence quality.On one hand, we introduce a soft matching mechanism, facilitating thepropagation of potentially valuable correspondences from coarse to fine levels.Additionally, we integrate an overlapping region detection module to minimizemismatches to the greatest extent possible. Furthermore, we introduce aregion-wise attention module with linear complexity during the fine-levelmatching phase, designed to enhance the discriminative capabilities of theextracted features. Tests on the challenging 3DLoMatch benchmark demonstratethat our approach leads to a substantial increase of about 7\% in the inlierratio, as well as an enhancement of 2-4\% in registration recall. =

OAAFormer: Robust and Efficient Point Cloud Registration Through Overlapping-Aware Attention in Transformer