Journal DOI 10.68956/journal.human-performance-futures

Creative Commons source reading · Data quality and algorithm evaluation

Advancing terrestrial laser scanning for 3D classification of Spanish moss (Tillandsia usneoides L.) using unsupervised, machine learning, and deep learning methods

Gaskins AJ; Xia J; Bueno IT; Silva CA

Originally published in Sci Rep · 2026-04-20

From the article

Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.

Abstract

AbstractQuantifying the distribution of Spanish moss (Tillandsia usneoides L.) is challenging because it grows suspended from high tree branches, limiting manual sampling. Terrestrial laser scanning (TLS) provides a non-destructive means of capturing vegetation structure in three dimensions. However, no established methods exist for identifying Spanish moss from TLS data. We evaluated five classification methods for distinguishing Spanish moss in TLS-derived point cloud data: Graph, DBSCAN, Random Forest (RF), Kernel Point Convolution (KPConv), and PointNet++. PointNet++ achieved the highest accuracy (81%), followed by DBSCAN (70%), KPConv (61%), RF (54%), and Graph (52%). Unsupervised methods required minimal computational resources (2–3 min, 8–16 GB memory) without training. RF required 3 h for training, 8 for prediction with 1024…

Discussion

This study provides the first application of supervised, unsupervised, and deep learning methods to classify Spanish moss from TLS point clouds. Previous TLS classification research has largely focused on separating leaves and wood52,53 or identifying vines and microhabitats54,55. By extending classification to an epiphytic species with complex morphology, this work opens a new direction for TLS-based vegetation analysis. Among the five methods, PointNet++ achieved the highest accuracy (81% on the test set), followed by DBSCAN (70%), KPConv (61%), RF (54%), and Graph (52%). These results confirm the potential of deep learning to outperform traditional methods, while also highlighting systematic errors: KPConv and Graph frequently overestimated moss, RF underrepresented non-Spanish moss, and DBSCAN misclassified branches as moss.…

Conclusion

This study presents the first attempt to use TLS to classify Spanish moss in three-dimensional point clouds, applying unsupervised, machine learning, and deep learning methods. Five methods were evaluated to test their ability to separate Spanish moss from non-Spanish moss structures. The results revealed substantial differences in performance, with PointNet++ providing the highest accuracy, followed by DBSCAN, KPConv, RF, and Graph. These findings demonstrate both the promise and the challenges of extending TLS classification beyond leaves and woody tissues to include epiphytic plants, which present greater complexity due to their filamentous and unordered morphology. Importantly, the ability to reliably separate Spanish moss from woody structures provides a foundation for future studies aimed at quantifying biomass and carbon…

Editorial reading guide

  • What data created and tested the model, and was evaluation independent?
  • What is the cost of a false positive, false negative or out-of-distribution decision?