| 标 题： |
A coarse-to-fine optimization framework for LiDAR-based air-ground cooperative mapping
基于LiDAR的空地协同建图的粗到细优化框架 |
| 作 者： |
Mengchi Ai
Naser El-Sheimy
Mohamed Elhabiby
Yandi Yang |
| 关键词： |
Point clouds map [点云地图]
Mobile mapping system [ 移动测绘系统]
Graph-based optimization [ 基于图的优化]
LiDAR-based SLAM [ 基于LiDAR的SLAM]
Unmanned aerial vehicles (UAV) [ 无人飞行器（UAV）]
Unmanned ground vehicles (UGV) [ 无人地面车辆（UGV）]
Cooperative mapping [ 协同建图] |
| 出版年： |
2025 |
| 文集名： |
2025 IEEE/ION Position, Location and Navigation Symposium (PLANS 2025)
2025年IEEE/ION位置、定位与导航研讨会（PLANS 2025） |
| 会议名： |
2025 IEEE/ION Position, Location and Navigation Symposium (PLANS 2025)
2025年IEEE/ION位置、定位与导航研讨会（PLANS 2025） |
| 届 次： |
|
| 摘 要： |
Geo-referenced point cloud maps provide detailed 3D geometric features essential for a wide range of digital applications, including smart city development and automation. Generating high-accuracy and high density point cloud maps is a critical task in these domains. While stationary scanning method offer high precise measurements, their low efficiency and limited coverage make them impractical for large-scale mapping. Mobile mapping technologies, leveraging simultaneous localization and mapping (SLAM), sensor fusion, and pose estimation, address these challenges by generating point cloud maps in motion. This paper proposes a cooperative air-ground 3D mapping framework that integrates unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) to produce point cloud maps. UAVs provide large-scale aerial coverage, while UGVs deliver detailed ground-level mapping, ensuring comprehensive and accurate representation of the environment. The proposed framework follows a coarse-to-fine optimization strategy consisting of two main stages: (1) Coarse mapping - SLAM algorithms generate initial point clouds and estimate geo-referenced poses for both UAVs and UGVs. (2) Mapping optimization - Graph-based post-processing refines the poses by leveraging overlapping constraints, further improving mapping accuracy. To validate the framework, experiments were conducted using an open-source UAV-UGV dataset. The results demonstrate significant improvements in point clouds mapping and trajectory accuracy. This air-ground cooperative approach represents a scalable, robust, and accurate solution for 3D mapping applications.
地理参考点云地图提供了广泛数字应用（包括智慧城市建设和自动化）所必需的详细三维几何特征。在这些领域中，生成高精度、高密度的点云地图是一项关键任务。虽然静态扫描方法能够提供高精度测量，但其效率低、覆盖范围有限，难以适用于大规模测绘。移动测绘技术通过利用同步定位与建图（SLAM）、传感器融合和姿态估计，在运动过程中生成点云地图，从而解决了上述挑战。本文提出一种协同空地三维测绘框架，整合无人机（UAV）和无人地面车（UGV）共同生成点云地图。无人机提供大范围的空中覆盖，而UGV则实现详细的地面测绘，确保环境信息的全面性和准确性。该框架采用粗到细的优化策略，主要包括两个主要阶段：（1）粗略测绘——SLAM算法生成初始点云，并对无人机和无人地面车的地理参考姿态进行估计。(2) 映射优化——基于图的后处理利用重叠约束来优化姿态，进一步提升映射精度。为验证该框架，使用开源的无人机-移动机器人数据集进行了实验。结果表明，在点云映射和轨迹精度方面取得了显著提升。这种空地协同方法为三维映射应用提供了一种可扩展、鲁棒且准确的解决方案。 |
| 中文内容为网络AI翻译，专业术语未必准确，仅供参考。 |
京公网安备 11010202009788