Predicting Tree Species From 3D Laser Scanning Point Clouds Using Deep Learning

• DOI: 10.3389/fpls.2021.635440
• Publication Date: 2021-02-10T14:56:36Z
• Authors: Dominik Seidel; Peter Annighöfer; Anton Thielman; Quentin Edward Se...; Jan-Henrik Thauer; Jonas Glatthorn; Martin Ehbrecht; Thomas Kneib; Christian Ammer
• Keywords: Tree (set theory); Contextual image classification; Laser Scanning; Confusion matrix; Sample (material); Confusion

Automated species classification from 3D point clouds is still a challenge. It is, however, an important task for laser scanning-based forest inventory, ecosystem models, and to support management. Here, we tested the performance of image approach based on convolutional neural networks (CNNs) with aim classify seven tree 2D representation in computationally efficient way. We were particularly interested how would perform artificially increased training data size augmentation techniques. Our yielded high accuracy (86%) confusion matrix revealed that despite rather small sample sizes some species, was high. could partly relate this successful application technique, improving our result by 6% total 13, 14, 24% ash, oak pine, respectively. The introduced hence not only applicable small-sized datasets, it also effective since relies instead be processed CNN. faster more accurate when compared cloud-based “PointNet” approach.