Authors:

Yuexin Jiang

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Light Detection and Ranging (LiDAR) technology has become an essential tool for forest inventory, enabling high-efficiency and wall-to- wall measurements of forest structure at the individual tree level. Recent advances in deep learning have demonstrated strong potential in analyzing 3D point cloud data for tasks such as tree segmentation and identification. However, accurate tree species classification with deep learning on 3D point cloud data remains a major challenge, primarily due to the lack of large, annotated point cloud datasets. This project aims to address this gap by creating a digital library of point clouds at individual tree level. We focus on refining individual tree segmentation, creating species labels, and matching stem locations with field inventory data collected from Martell Forest, Indiana, using backpack mobile LiDAR systems. The dataset covers a 2-hectare area and consists of approximately 410 million points representing around 20 tree species. The resulting library will provide structured, georeferenced point cloud data paired with species annotations, serving as a foundational dataset for training and evaluating deep learning models in future ecological and forestry studies. Keywords: LiDAR; Deep Learning; Tree Species Identification; 3D Point Cloud; Individual Tree Segmentation

Source:

Purdue University / 2025

Topics:

No topics listed

Co-authors:

Yuexin Jiang