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

Place recognition is a key module for long-term SLAM systems. CurrentLiDAR-based place recognition methods usually use representations of pointclouds such as unordered points or range images. These methods achieve highrecall rates of retrieval, but their performance may degrade in the case ofview variation or scene changes. In this work, we explore the potential of adifferent representation in place recognition, i.e. bird's eye view (BEV)images. We observe that the structural contents of BEV images are lessinfluenced by rotations and translations of point clouds. We validate that,without any delicate design, a simple VGGNet trained on BEV images achievescomparable performance with the state-of-the-art place recognition methods inscenes of slight viewpoint changes. For more robust place recognition, wedesign a rotation-invariant network called BEVPlace. We use group convolutionto extract rotation-equivariant local features from the images and NetVLAD forglobal feature aggregation. In addition, we observe that the distance betweenBEV features is correlated with the geometry distance of point clouds. Based onthe observation, we develop a method to estimate the position of the querycloud, extending the usage of place recognition. The experiments conducted onlarge-scale public datasets show that our method 1) achieves state-of-the-artperformance in terms of recall rates, 2) is robust to view changes, 3) showsstrong generalization ability, and 4) can estimate the positions of query pointclouds. Source codes are publicly available athttps://github.com/zjuluolun/BEVPlace.

BEVPlace: Learning LiDAR-based Place Recognition using Bird's Eye View Images