Multi-Layer Modeling of Dense Vegetation from Aerial LiDAR Scans

TL;DR

This paper introduces WildForest3D, a comprehensive dataset and a novel 3D deep learning model that jointly predicts vegetation layer occupancy and point-wise labels from aerial LiDAR scans, enabling detailed multi-layer forest analysis.

Contribution

It provides the first dataset with dense multi-layer annotations and a deep network architecture for simultaneous layer occupancy and point-wise labeling in forest environments.

Findings

Accurate multi-layer vegetation modeling from aerial LiDAR data.

Open access dataset with detailed annotations for forestry research.

Effective 3D deep learning approach for vegetation layer estimation.

Abstract

The analysis of the multi-layer structure of wild forests is an important challenge of automated large-scale forestry. While modern aerial LiDARs offer geometric information across all vegetation layers, most datasets and methods focus only on the segmentation and reconstruction of the top of canopy. We release WildForest3D, which consists of 29 study plots and over 2000 individual trees across 47 000m2 with dense 3D annotation, along with occupancy and height maps for 3 vegetation layers: ground vegetation, understory, and overstory. We propose a 3D deep network architecture predicting for the first time both 3D point-wise labels and high-resolution layer occupancy rasters simultaneously. This allows us to produce a precise estimation of the thickness of each vegetation layer as well as the corresponding watertight meshes, therefore meeting most forestry purposes. Both the dataset and the model are released in open access: https://github.com/ekalinicheva/multi_layer_vegetation.