Flying Co-Stereo: Enabling Long-Range Aerial Dense Mapping via Collaborative Stereo Vision of Dynamic-Baseline

TL;DR

Flying Co-Stereo enables long-range dense 3D mapping for UAV swarms by collaborative stereo vision, combining robust baseline estimation, hybrid feature matching, and depth refinement to significantly extend perception range and coverage.

Contribution

This work introduces a novel collaborative stereo system for UAVs that achieves long-range dense mapping, surpassing traditional fixed-baseline stereo systems in range and accuracy.

Findings

Achieves dense 3D mapping up to 70 meters

Outperforms conventional systems by up to 350% in depth range

Provides 450% larger coverage area

Abstract

Lightweight long-range mapping is critical for safe navigation of UAV swarms in large-scale unknown environments. Traditional stereo vision systems with fixed short baselines face limited perception ranges. To address this, we propose Flying Co-Stereo, a cross-agent collaborative stereo vision system that leverages the wide-baseline spatial configuration of two UAVs for long-range dense mapping. Key innovations include: (1) a dual-spectrum visual-inertial-ranging estimator for robust baseline estimation; (2) a hybrid feature association strategy combining deep learning-based cross-agent matching and optical-flow-based intra-agent tracking; (3) A sparse-to-dense depth recovery scheme,refining dense monocular depth predictions using exponential fitting of long-range triangulated sparse landmarks for precise metric-scale mapping. Experiments demonstrate the Flying Co-Stereo system achieves dense 3D mapping up to 70 meters with 2.3%-9.7% relative error, outperforming conventional systems by up to 350% in depth range and 450% in coverage area. The project webpage: https://xingxingzuo.github.io/flying_co_stereo