The National Institute of Forest Science (NIFoS) of the Korea Forest Service has developed an AI-based automation technology that can classify Korea's representative tree species using three-dimensional (3D) point cloud data collected through ground-based LiDAR sensors.

The technology scans forests with various LiDAR devices such as fixed, handheld, and backpack-mounted systems, identifies individual trees, and then applies AI algorithms to automatically determine tree species.

NIFoS has, so far, developed the technology to classify five coniferous species (Pinus densiflora, Pinus thunbergii, Pinus koraiensis, Larix kaempferi, and Chamaecyparis obtusa) and three broadleaf species (Quercus mongolica, Quercus variabilis, and Quercus acutissima).

Unlike conventional two-dimensional (2D) image-based approaches, which rely only on spectral data from tree canopies and have limitations in precisely identifying number of species and location, the new technology leverages LiDAR-derived digital tree structures for more accurate identification of species and location.

In particular, the AI algorithm demonstrated 99% accuracy in distinguishing between coniferous and broadleaf forests. It also achieved 94% accuracy in classifying five coniferous along with broadleaf forests, and 92% accuracy in classifying three broadleaf species with coniferous forests. In other words, the new technology showed 5% higher accuracy than the existing 2D image-based methods.

Building on the new approach, NIFoS plans to further enhance the technology to expand its applicability in the field and classify a wider variety of tree species.