A lightweight LiDAR-SLAM system integrated with semantic segmentation for the acquisition of forest biomass parameters

Abstract Key message We propose a lightweight LiDAR-based simultaneous localization and mapping system that integrates semantic segmentation. The system was evaluated on forest point clouds collected in real-world forest environments and on multiple publicly available datasets. Results demonstrate that the proposed system is both efficient and accurate for individual-tree clustering and diameter at breast height estimation. Context Efficient and automated acquisition of individual-tree parameters is essential for intelligent forest resource inventories. Conventional approaches rely heavily on manual measurements, which limits scalability and makes them unsuitable for large-scale and high-frequency surveys. Aims This study aims to develop a lightweight LiDAR-SLAM system with integrated semantic segmentation to improve the accuracy and real-time performance of individual-tree extraction and DBH measurement in forest environment. Methods The proposed system is built on an enhanced LIO-SAM framework and incorporates an incremental k-d tree (Ikd-Tree) to improve computational efficiency. For semantic segmentation, we adopted SqueezeSegV3 augmented with an Efficient Layer Attention (ELA) mechanism to improve semantic-category recognition. The segmented point clouds were then processed using clustering and cylinder fitting to extract individual trees and estimate DBH. Results The system was tested on public datasets and field-collected forest data, achieving semantic segmentation accuracies of 0.85 and 0.89, with mean Intersection over Union values of 0.55 and 0.67, respectively. The average DBH prediction accuracy reached 97.6%, indicating strong performance in real forest environments. Conclusion By combining a lightweight semantic network with an efficient point-cloud data structure, the proposed system achieved high accuracy and real-time performance, meeting the requirements of large-scale and high-efficiency measurements for forest resource inventory tasks.