Identification of Tree Species Using Object-Based Classification of Digital Aerial Images In The Northern Forests of Iran (Case Study: Chamestan-Nur)

Tree species identification is essential to forest type mapping. Remote sensing data, especially high spatial and radiometric resolution digital aerial images can be useful to this goal achievement. This research is aimed at the evaluation of digital aerial images in order to identify tree species in two study areas located in a natural mixed forest. Radiometric and geometric qualities of these images were evaluated and all images were corrected geometrically. Some appropriate transformations were performed and utilized. The hierarchical image object network was constructed and various alternatives of segmentation were tried and qualitatively evaluated. Nearest Neighbor classifier based on Fuzzy logic, leads to tree species map. Training samples and ground truth map were prepared through fieldwork. The result of accuracy assessment showed the overall accuracy more than 70% and Kappa Index of Agreement (KIA) equal to 0.46 and 0.5 in these areas. Main tree species presents different accuracies in two areas. Applying these aerial images and the object-based method in other forest areas together with the consideration of the influence of using proper elevation data on the accuracy of the produced map is suggested.