Fusion Approaches to Individual Tree Species Classification Using Multisource Remote Sensing Data

With the wide availability of remotely sensed data from various sensors, fusion-based tree species classification approaches have emerged as a prominent and ongoing research topic. However, most recent studies primarily focused on combining multisource at feature level, while few systematically examined their positive or negative contributions to classification. This study aimed investigate fusion decision levels deployed with support vector machine random forest algorithms classify five dominant species: Norway maple, honey locust, Austrian pine, white spruce, blue spruce in individual crowns. Spectral, textural, structural features derived multispectral imagery (MSI), very high-resolution panchromatic image (PAN), LiDAR were exploited assess accurate classifications. Among schemes that explored, both feature- decision-level demonstrated significant improvements compared utilization MSI (0.7), PAN (0.74), (0.8) isolation. Notably, approach achieved highest overall accuracies (0.86 for SVM 0.84 RF) kappa coefficients (0.82 0.79 RF). The misclassification analysis highlighted potential flexibility