Fusion Approaches to Individual Tree Species Classification Using Multi-Source Remotely Sensed Data

Tree species information plays essential roles in urban ecological management and sustainable development, and thus tree species classification has been an active research topic over the years. This study investigated fusion approaches deployed with Support Vector Machine (SVM) and Random Forest (RF) algorithms to incorporating multispectral imagery (MSI), a very high spatial resolution panchromatic image (PAN), and Light Detection and Ranging (LiDAR) data for five object-based tree species classification in an urban environment. The results demonstrated that 3D structural features contributed more to tree species with broad crowns, such as honey locust and Austrian pine, whereas textural features were more effective in differentiating trees in narrow crowns, such as spruce. Among all the possible classification schemes based on multi-source features in combinations, decision fusion achieved the best overall accuracies (0.86 for SVM and 0.84 for RF), slightly outperforming the feature fusion approach (0.85 for SVM and 0.83 for RF). Both fusion approaches significantly improved tree species classifications produced by MSI (0.7), PAN (0.74), and LiDAR (0.8) individually.