Search papers, labs, and topics across Lattice.
This paper introduces ProteusVPR, a two-stage retrieval-refinement framework designed to enhance Visual Place Recognition (VPR) for autonomous robotic inspection in maritime environments, which face significant perceptual shifts between open decks and enclosed cabins. By integrating a geometric-visual estimation network that combines retrieved images with temporally preceding frames and geometric descriptors, the method achieves a substantial improvement in localization accuracy. Experiments on the newly introduced XHZ dataset reveal that ProteusVPR reduces mean localization error by over 60% compared to existing VPR methods, highlighting its effectiveness in challenging maritime scenarios.
ProteusVPR slashes localization error by over 60% in maritime environments where traditional VPR methods fail.
Autonomous robotic inspection in maritime environments presents unique challenges for Visual Place Recognition (VPR) due to cross-scene perceptual shifts. Robots navigating ship-borne environments must transition between visually distinct domains: open decks with sparse textures and severe illumination changes, and enclosed cabins with repetitive structures and high visual ambiguity. Existing VPR methods, designed primarily for urban or indoor scenes, fail to generalize reliably across these starkly different scenarios. To address this, we propose ProteusVPR, a two-stage retrieval-refinement framework. The first stage employs any standard VPR model for initial image retrieval. The second stage introduces a geometric-visual estimation network that fuses the retrieved image with two temporally preceding frames, incorporating geometric descriptors, a local affine coordinate system, and camera azimuth encoding to achieve precise localization. To support this task, we introduce the XHZ dataset, an 8K-panoramic ship-borne dataset collected from an operational vessel, featuring multi-floor cabin structures, deck transition zones, and strict query-database separation for rigorous evaluation. Extensive experiments on the XHZ dataset demonstrate that ProteusVPR consistently improves the localization accuracy across multiple VPR backbones, reducing mean localization error by over 60\% on average and that ProteusVPR offers an effective and robust solution for precise visual localization in challenging, cross-scene maritime environments.
