New LiDAR technology enables "single-tree level" 3D modeling, capable of penetrating forest canopies and performing automatic individual tree segmentation (accuracy >95%) while extracting tree height, DBH, and volume with high precision, significantly boosting forest inventory efficiency. However, this progress is the result of synergistic evolution across multi-platforms (airborne/backpack/handheld) and AI algorithms, rather than an isolated technological breakthrough disclosed solely in May 2026.

Penetration Capability & Individual Tree Segmentation:

Current multi-echo LiDAR​ systems (e.g., DJI Zenmuse L3, CHCNAV RS30), combined with point cloud classification algorithms​ (based on echo count, intensity, and spatial clustering), effectively distinguish between canopy, trunk, and ground points. AI-driven instance segmentation models​ (such as PointNet++, 3D CNNs, or the latest Transformer architectures) have achieved an F1-score above 0.95​ (corresponding to accuracy >95%) for individual tree recognition in closed-canopy stands in research from 2025–2026.

Automatic Parameter Extraction:

Tree height (error <5%) and Diameter at Breast Height (DBH, error ±3–5%) are derived through cylindrical fitting of trunk point clouds or least squares methods. Volume is estimated using allometric equations​ (e.g., DBH² × Height) integrated with species libraries, achieving overall regional errors of <3%​ (as verified by CHCNAV’s ForestPoint software).

Evolution of Technical Platforms:

The industry has moved beyond static terrestrial LiDAR to integrate UAV-borne LiDAR​ (for large-scale coverage), backpack/handheld LiDAR​ (to supplement understory details), and SLAM real-time positioning, enabling collaborative "air-to-ground" point cloud acquisition. In early 2026, teams from Northeast Forestry University and GreenValley International validated the engineering capability of this system at single-tree precision levels (<10 cm positioning error).

Key Breakthroughs:

The advancement does not stem from "new physical penetration," but rather from the convergence of increased point cloud density (>20 pts/m² in understory) + multi-echo processing + deep learning segmentation (e.g., Mask3D) + joint terrain-vegetation modeling. Additionally, the "canopy edge-aware clipping algorithm" proposed by the Chinese Academy of Forestry in 2025 has significantly improved the integrity of individual tree separation in dense stands.

Practical Applications:

Pilot programs in the forests of Northeast and Southwest China have demonstrated efficiency gains of over 10 times​ compared to traditional plot surveys, with errors meeting the Class III accuracy requirements for National Forest Inventory (LY/T 1953–2021). This technology is driving forestry from "plot-based estimation" toward "precise per-tree surveying and mapping."