G-PROBE: Cross-FOV Place Recognition and Certainty-Coupled Localization for 3D Point Clouds

Jul 16, 2026

Upload by Sage

Global localization from 3D point clouds is difficult when sensors provide limited or asymmetric fields of view, breaking the dense, symmetric coverage assumptions behind many place-recognition methods. G-PROBE introduces a learning-free global localization framework that operates consistently across narrow-FOV, panoramic, and multi-sensor setups. Cross-FOV branch ensembles estimate heading-invariant hypotheses, while tuning-free γ-SGRT suppresses heading aliasing and CG-GICP refines poses using high-certainty co-observed points. Evaluation on multiple LiDAR datasets shows top learning-free multi-session F1 and strong robustness under wide-to-narrow cross-sensor pairing and FOV asymmetry.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

Introduction

The Heterogeneity and Cross-FOV Heading Problem

Sensor Heterogeneity and Modality Invariance

Heading-Dependent Coverage

Limitations of Existing Approaches

Probing Reliability from Geometry

G-PROBE: Cross-FOV Place Recognition and Certainty-Coupled Localization for 3D Point Clouds Jinseop Lee arXiv :2607 .06782v 1 [ cs .RO] 7 Jul 2026 Abstract—Global localization from 3D point clouds remains challenging under limited or asymmetric fields of view (FOV), which fail to provide the dense, symmetric coverage that place recognition methods assume. We present G-PROBE, a learningfree global localization framework that removes this assumption. A virtual sensor decomposition runs the same pipeline, by design, on configurations ranging from a narrow-FOV sensor to a panoramic or multi-sensor rig. The front-end enumerates cross-FOV branch ensembles that encode heading hypotheses for heading-invariant place recognition. A score-scale-invariant, tuning-free γ-SGRT suppresses heading aliasing under partial FOV and provably becomes inert at symmetric 360◦ . The back-end, CG-GICP, refines a coarse full-cloud GICP with a pass restricted to high-certainty co-observed points selected by a bird’s-eye-view certainty map (a by-product of front-end scoring). This certainty coupling links descriptor evaluation to 6-DoF metric pose estimation without an external verification module. Evaluated on five LiDAR datasets and three modalities (mechanical, solid-state, FMCW), G-PROBE attains the highest learning-free multi-session F1 on average and is competitive in panoramic single-session settings. Where hand-crafted and zeroshot supervised baselines collapse under wide↔narrow crosssensor pairing, it remains usable end-to-end (up to 55.0% vs.≤6 .8% success), and under FOV asymmetry (360◦ → 60◦) it retains ∼54% Recall@1, ∼ 18 × the strongest learning-free baseline. Index Terms—3D Point Clouds, LiDAR, Global Localization, Certainty-Guided Registration, Cross-FOV Heading Invariance, Place Recognition I. INTRODUCTION Global localization, the estimation of a robot’s metric pose directly from a prior map without an initial pose estimate, is fundamental to robust autonomous navigation and loop closure in SLAM. Place recognition is well established for cameras [1] and dense 360◦ LiDAR [2], [3], but the mature LiDAR solutions implicitly assume a dense, symmetric, panoramic field of view (FOV) . Localization on diverse, FOV-constrained point clouds has received comparatively little attention, even asthe sensors that produce them are increasingly deployed for their long sensing range and compact, mechanically robust solid-state design. A recent survey likewise names limited FOV, cross-sensor configuration, and unbalanced matching among the open problems of global LiDAR localization [4] . A. The Heterogeneity and Cross-FOV Heading Problem The central challenge addressed in this paper is the mis match between the panoramic assumptions of existing methods J. Lee is with SK Intellix, Seoul, Republic of Korea (e-mail: [jinseop.llee@gmail.com](jinseop.llee@gmail.com)) . 360 180 120 90 60 Query FOV ( ) , database = 360 Fig. 1: Place recognition under FOV asymmetry. Recall@1 as the query field of view is cropped from 360◦ to 60◦ against a panoramic (360◦ ) database (asymmetric block of Table V, macro mean over the 12 single-session sequences of three datasets) . Most hand-crafted descriptors collapse by 180◦ (to ∼2% at 60◦ ) . The supervised baselines (†) fall to ≤15% at 60◦(HeLiOS, the most gradual, retains 14%) . G-PROBE alone retains ∼54% even at 60◦ , at a CPU runtime on par with SC++ (Table XII) . and the constrained, heterogeneous FOVs of modern sensor configurations. Problem 1: Sensor Heterogeneity and Modality Invariance. Emerging autonomous systems deploy a mix of spinning, directional solid-state, and frequency-modulated continuous-wave (FMCW) LiDARs, each with distinct point densities, noise distributions, and FOVs. Modality-invariant matching across these sensor types remains an open challenge. Recent methods attempt this using overlap-based deep learning (HeLiOS [5], following the overlap supervision of OverlapNet [6]) or

FAQs

What core problem does G-PROBE address in 3D point cloud global localization?

G-PROBE targets global localization failures caused by limited or asymmetric fields of view, which violate the dense, symmetric FOV assumptions of many place-recognition approaches.

How does G-PROBE perform place recognition under cross-FOV conditions?

The front-end enumerates cross-FOV branch ensembles that encode heading hypotheses for heading-invariant place recognition, and uses γ-SGRT to suppress heading aliasing under partial FOV.

How does G-PROBE refine localization while linking recognition reliability to pose estimation?

The back-end CG-GICP refines a coarse full-cloud GICP using a pass restricted to high-certainty co-observed points selected from a bird’s-eye-view certainty map, coupling descriptor evaluation to 6-DoF metric pose estimation.

Jul 16, 2026

Upload by Sage