Machine Learning Classification and Accuracy Assessment from High-Resolution Images of Coastal Wetlands

High-resolution images obtained by multispectral cameras mounted on Unmanned Aerial Vehicles (UAVs) are helping to capture the heterogeneity of environment in that can be discretized categories during a classification process. Currently, there is an increasing use supervised machine learning (ML) classifiers retrieve accurate results using scarce datasets with samples non-linear relationships. We compared accuracies two ML pixel and object analysis approach six coastal wetland sites. The show Random Forest (RF) performs better than K-Nearest Neighbors (KNN) algorithm pixels objects based slightly object-based analysis. agreement between classifications higher Forest. This likely due study areas, where pixel-based most appropriate. In addition, from ecological perspective, as these wetlands heterogeneous, reflects more realistic interpretation plant community distribution.