Cross-Camera View-Overlap Recognition

TL;DR

This paper introduces a decentralized framework for recognizing overlapping camera views without relying on a 3D map, using feature descriptors and geometric validation to improve accuracy across moving cameras.

Contribution

It presents a novel decentralized view-overlap recognition method that works with various descriptors and does not require a reference 3D map, enhancing flexibility and accuracy.

Findings

ORB features with Bags of Binary Words outperform NetVLAD, RootSIFT, and SuperGlue in precision.

The framework achieves higher or comparable accuracy across multiple datasets.

Decentralized approach enables view recognition without a fixed reference map.

Abstract

We propose a decentralised view-overlap recognition framework that operates across freely moving cameras without the need of a reference 3D map. Each camera independently extracts, aggregates into a hierarchical structure, and shares feature-point descriptors over time. A view overlap is recognised by view-matching and geometric validation to discard wrongly matched views. The proposed framework is generic and can be used with different descriptors. We conduct the experiments on publicly available sequences as well as new sequences we collected with hand-held cameras. We show that Oriented FAST and Rotated BRIEF (ORB) features with Bags of Binary Words within the proposed framework lead to higher precision and a higher or similar accuracy compared to NetVLAD, RootSIFT, and SuperGlue.