Unmanned Aerial System and Machine Learning Techniques Help to Detect Dead Woody Components in a Tropical Dry Forest

摘要: Background and Objectives: Increased frequency intensity of drought events are predicted to occur throughout the world because climate change. These extreme result in higher tree mortality fraction dead woody components, phenomena that currently being reported worldwide as critical indicators impacts change on forest diversity function. In this paper, we assess accuracy processing times ten machine learning (ML) techniques, applied multispectral unmanned aerial vehicle (UAV) data detect canopy components. Materials Methods: This work was conducted five secondary dry plots located at Santa Rosa National Park Environmental Monitoring Super Site, Costa Rica. Results: The coverage components selected estimated range from 4.8% 16.1%, with no differences between successional stages. Of ML support vector radial kernel (SVMR) random forests (RF) provided highest accuracies (0.982 vs. 0.98, respectively). these two algorithms, time SVMR longer than RF (8735.64 s 989 s). Conclusions: Our results demonstrate it is feasible quantify such stands fallen trees, using a combination high-resolution UAV algorithms. Using technology, values 95% were achieved. However, important account for series factors, optimization tuning parameters environmental conditions acquisition.