Markerless Aerial-Terrestrial Co-Registration of Forest Point Clouds using a Deformable Pose Graph

TL;DR

The paper tackles the challenge of creating dense, multi-scale forest reconstructions by fusing aerial laser scans with terrestrial MLS data in a markerless, deformable pose-graph framework. It introduces a two-stage aerial-terrestrial matching pipeline (pre-processing, CHM-based aerial features, trunk-centers from MLS, coarse via maximum clique, fine via ICP) and a pose-graph optimization that unifies SLAM-based payloads and tile representations with aerial-terrestrial factors. The approach is validated on Finnish and Swiss forests, showing improved alignment and greater scene completeness, enabling more accurate tree trait estimation and forest inventories. Overall, the method enables high-fidelity, marker-free fusion of multi-platform forest data, supporting scalable forestry monitoring and management without dedicated survey markers.

Abstract

For biodiversity and forestry applications, end-users desire maps of forests that are fully detailed, from the forest floor to the canopy. Terrestrial laser scanning and aerial laser scanning are accurate and increasingly mature methods for scanning the forest. However, individually they are not able to estimate attributes such as tree height, trunk diameter and canopy density due to the inherent differences in their field-of-view and mapping processes. In this work, we present a pipeline that can automatically generate a single joint terrestrial and aerial forest reconstruction. The novelty of the approach is a marker-free registration pipeline, which estimates a set of relative transformation constraints between the aerial cloud and terrestrial sub-clouds without requiring any co-registration reflective markers to be physically placed in the scene. Our method then uses these constraints in a pose graph formulation, which enables us to finely align the respective clouds while respecting spatial constraints introduced by the terrestrial SLAM scanning process. We demonstrate that our approach can produce a fine-grained and complete reconstruction of large-scale natural environments, enabling multi-platform data capture for forestry applications without requiring external infrastructure.

Figures (11)

• Figure 1: A forest map built using a terrestrial map from our mobile scanning system (orange) and an aerial map from a drone (blue). Together, the scans produce a map reconstructing the canopy and understory in sufficient detail for individual-tree forest inventory.
• Figure 2: An unoptimized MLS point cloud (in red) is poorly aligned with the ALS map (blue). On the other hand, the optimized cloud (in orange) is directly aligned with the ALS map. The shearing in the red MLS point cloud is due to the small drift in the SLAM system (a few meters across a kilometer scale map).
• Figure 3: General system overview. Our aerial-terrestrial co-registration system combines the aerial clouds from ALS, with a terrestrial reconstruction obtained by means of a LiDAR-Inertial odometry and pose-graph SLAM system.
• Figure 4: We consider two formats for the MLS data. The pose graph format is defined by a pose graph and a collection of data payloads (a), obtained by temporal aggregation of consecutive LiDAR scans; payloads cover a larger area but are sparse and lack canopy points. The tile format represents the MLS mission as a collection of tiles (b), obtained by aggregating all the mission scans into a single, dense cloud and partitioning them into a fixed-size grid.
• Figure 5: The aerial-terrestrial registration step aims to find the relative transformation between a large aerial and a much smaller local terrestrial cloud, $\mathbf{T}_{\texttt{AM}}$.