Using Voting-Based Ensemble Classifiers to Map Invasive Phragmites australis

Machine learning is frequently combined with imagery acquired from uncrewed aircraft systems (UASs) to detect invasive plants. Having prior knowledge of which machine algorithm will produce the most accurate results difficult. This study examines efficacy a voting-based ensemble classifier identify Phragmites australis three-band (red, green, blue; RGB) and five-band blue, red edge, near-infrared; multispectral; MS) UAS over multiple Minnesota wetlands. A Random Forest, histogram-based gradient-boosting classification tree, two artificial neural networks were used within classifier. Classifications RGB multispectral compared across validation sites both without post-processing an object-based image analysis (OBIA) workflow (post-machine OBIA rule set; post-ML set). Results this suggest that can accurately imagery. Accuracies greater than 80% attained by for The highest accuracy, 91%, was achieved when using imagery, canopy height model, set. emphasizes need further research regarding identification at low stem densities.