NormalView: sensor-agnostic tree species classification from backpack and aerial lidar data using geometric projections

2025-12-05

AI Summary

This study addresses the sensor dependency and inconsistent feature representation in tree species classification using mobile laser scanning (MLS) and airborne laser scanning (ALS) data. To this end, we propose NormalView—a sensor-agnostic method that robustly projects 3D point clouds onto 2D images via normal-vector-guided geometric projection, preserving local geometric structure while integrating radiometric intensity from multispectral LiDAR to enhance discriminability. Classification is performed end-to-end using the YOLOv11 architecture, enabling seamless cross-platform adaptation within a unified framework. Experiments demonstrate state-of-the-art performance: 95.5% overall accuracy on high-density MLS data and 91.8% on ALS data—significantly outperforming sensor-specific approaches. Additionally, we publicly release the first high-density MLS benchmark dataset specifically designed for tree species identification, facilitating reproducible and cross-platform forest remote sensing research.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Remote Sensing / Geospatial AI
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Security and Privacy: Data transparency and provenance

Abstract

Laser scanning has proven to be an invaluable tool in assessing the decomposition of forest environments. Mobile laser scanning (MLS) has shown to be highly promising for extremely accurate, tree level inventory. In this study, we present NormalView, a sensor-agnostic projection-based deep learning method for classifying tree species from point cloud data. NormalView embeds local geometric information into two-dimensional projections, in the form of normal vector estimates, and uses the projections as inputs to an image classification network, YOLOv11. In addition, we inspected the effect of multispectral radiometric intensity information on classification performance. We trained and tested our model on high-density MLS data (7 species, ~5000 pts/m^2), as well as high-density airborne laser scanning (ALS) data (9 species, >1000 pts/m^2). On the MLS data, NormalView achieves an overall accuracy (macro-average accuracy) of 95.5 % (94.8 %), and 91.8 % (79.1 %) on the ALS data. We found that having intensity information from multiple scanners provides benefits in tree species classification, and the best model on the multispectral ALS dataset was a model using intensity information from all three channels of the multispectral ALS. This study demonstrates that projection-based methods, when enhanced with geometric information and coupled with state-of-the-art image classification backbones, can achieve exceptional results. Crucially, these methods are sensor-agnostic, relying only on geometric information. Additionally, we publically release the MLS dataset used in the study.

Problem

• Classifies tree species using sensor-agnostic geometric projections from LiDAR
• Evaluates multispectral intensity's impact on classification accuracy
• Applies method to both mobile and airborne laser scanning data

Innovation

• Uses geometric projections to embed local information into 2D images
• Applies YOLOv11 image classification network to process projected data
• Incorporates multispectral intensity information to improve classification accuracy