High-precision point cloud generation for forest inventory: Integrating GNSS-RTK and SLAM for handheld laser scanning

In recent years, forest management and inventory have increasingly relied on handheld personal laser scanners (H-PLS) for capturing flexible three-dimensional data. These systems have become essential for extracting critical tree attributes, such as diameter at breast height (DBH) and tree height. Most traditional H-PLS systems utilize Simultaneous Localization and Mapping (SLAM), which fuses LiDAR and Inertial Measurement Unit (IMU) data to reconstruct environments. However, SLAM is based on relative sensor measurements, which inherently causes accumulated errors and trajectory drift. In complex forest environments, similar-looking stems and moving vegetation can further confuse the mapping process, resulting in distorted point clouds or duplicated stems that reduce the accuracy of extracted tree attributes.

While Global Navigation Satellite System (GNSS)-based Real-Time Kinematic (RTK) positioning provides centimetre-level absolute accuracy and usually drift-free trajectories, its application in forestry is critically hindered by signal obstruction in dense canopies. The integration of GNSS-RTK and SLAM offers a robust and synergetic solution to these challenges, allowing one method to compensate for the failures of the other. A promising development in this field is an H-PLS system that integrates GNSS-RTK, IMU, LiDAR, and camera measurements to generate georeferenced point clouds directly in the field. This hybrid approach utilizes LiDAR and camera data to maintain positioning during GNSS outages and utilizes RTK information to re-initialize and correct the trajectory once the signal is restored.

Our study evaluates whether this integrated GNSS-RTK SLAM approach improves point cloud geometry and tree attribute extraction compared to traditional SLAM methods without GNSS integration. We conducted a field campaign in a mixed forest stand during the leaf-off period to simulate realistic operating conditions with alternating GNSS visibility. The performances of a SLAM-only and a SLAM + GNSS-RTK H-PLS were validated against highly accurate terrestrial laser scanning (TLS) reference data. The analysis involved tree segmentation to assess individual tree identification and the derivation of DBH, stem positions, and tree heights. Furthermore, we investigated internal geometric quality by analysing local noise levels using cross-sectional residuals relative to fitted circles and assessed spatial homogeneity to identify artifacts like duplicated stems or gaps.

Initial results indicate that the SLAM + GNSS-RTK H-PLS system provides DBH estimates comparable to TLS, with observed differences of 6.3 mm and 1.17 cm for major and minor axes, respectively. Despite slight overestimations due to scattering, the significantly reduced acquisition time makes this integrated system an efficient alternative for forestry applications. These findings contribute to a better understanding of how integrated positioning systems can enhance mobile laser scanning workflows and support the development of autonomous, high-precision forest mapping solutions.