TreeLoc: 6-DoF LiDAR Global Localization in Forests via Inter-Tree Geometric Matching

TL;DR

TreeLoc tackles the challenge of LiDAR-based global localization in forests by adopting a stem-centric, learning-free representation. It reconstructs trees from aggregated LiDAR payloads, aligns scenes via axis-based roll/pitch correction, and employs a dual descriptor framework (TDH for coarse retrieval and a 2D triangle descriptor for fine matching) followed by a two-stage geometric verification to yield $6$-$DoF$ pose estimates. Ablation studies demonstrate the critical roles of TDH, DBH, and alignment, while experiments on diverse datasets show superior cross-session robustness and centimeter-level localization accuracy, all with a compact, executable global tree database that enables lightweight multi-session alignment. The approach offers scalable forest map management, fast runtime (<50 ms), and resilience to seasonal and viewpoint changes, making it practically impactful for long-term forest inventory and management.

Abstract

Reliable localization is crucial for navigation in forests, where GPS is often degraded and LiDAR measurements are repetitive, occluded, and structurally complex. These conditions weaken the assumptions of traditional urban-centric localization methods, which assume that consistent features arise from unique structural patterns, necessitating forest-centric solutions to achieve robustness in these environments. To address these challenges, we propose TreeLoc, a LiDAR-based global localization framework for forests that handles place recognition and 6-DoF pose estimation. We represent scenes using tree stems and their Diameter at Breast Height (DBH), which are aligned to a common reference frame via their axes and summarized using the tree distribution histogram (TDH) for coarse matching, followed by fine matching with a 2D triangle descriptor. Finally, pose estimation is achieved through a two-step geometric verification. On diverse forest benchmarks, TreeLoc outperforms baselines, achieving precise localization. Ablation studies validate the contribution of each component. We also propose applications for long-term forest management using descriptors from a compact global tree database. TreeLoc is open-sourced for the robotics community at https://github.com/minwoo0611/TreeLoc.

Figures (8)

• Figure 1: (a) GPS bias in forests results in misalignment of different LiDAR SLAM trajectories. (b) TreeLoc performs coarse retrieval using TDH descriptors summarizing tree counts by location and DBH. (c) Fine retrieval with 2D triangle descriptors from tree centers (brown circles); eight matched triangles are highlighted in the same colors. (d) A precise 6-DoF pose enables successful multi-session alignment across datasets captured over different years.
• Figure 2: TreeLoc converts LiDAR scans into tree instances. After performing place recognition using two descriptors, it leverages existing data to perform 6-DoF pose estimation. At the end of the pipeline, the point cloud is tightly registered without ICP.
• Figure 3: The coarse stage estimates an initial planar transform from triangle centroids ($\mathbf{q}$) matched between the query and candidate scenes. The fine stage refines this planar transform using 2D tree centers ($\mathbf{c}$) and base heights (${b}$), obtaining the 4-DoF relative pose.
• Figure 4: Feature similarity maps in Evo (warmer colors for higher similarity). TreeLoc focuses similarity in a compact region aligned with the ground truth, enabling clear thresholding, while baselines yield diffuse areas (blue) mixed with false positives (green).
• Figure 5: Precision–Recall curves comparing with algorithmic methods (left) and learning-based methods (right). TreeLoc achieves a higher AUC and maintains high precision at high recall, thereby avoiding the sharp drop in precision observed in baselines.