Investigation on the Possibility of Tree Species Identification Using Digital Aerial Images by Object- based Classification

Remotely sensed data، especially high spatial and radiometric resolution data can be useful to identify tree species. In this research، the capability of digital aerial images for this purpose was investigated by object-based classification method. About 5. 8 ha of Taleghani park in Tehran، including species like platanus orientalis، Morus alba، Cupressus arisonica، Robinia pseudoacacia، Pinus eldarica، Ailanthus altissima، Cedrus atlantica، was studied. Four pan-sharpened multispectral images of UltraCam-D camera with the spatial and radiometric resolution، 0. 07×0. 07 m and 16 bit respectively، were analyzed. These data were geometrically corrected by aero-triangulation method using GCPs and IMU. The images classified using object-based method with the main and synthesic bands of Ratioing، PCA and HIS transformations. Firstly، segmentation was done with different parameters in order to avoid exceeding the maximum allowable number of objects. Finally، the classification was performed using appropriate features and layers by Nearest Neighbor method. In order to assess the accuracy of result، a ground truth map was produced based on field survey. This map has included 688 points، which each represents a tree on the ground. The result of accuracy assessment showed that overall accuracy and Kappa coefficient were 78% and 0. 73 respectively. Platanus and Aillan showed the highest and the lowest Kappa 0. 81700. 2481، respectively. The result of this study showed that the UltraCam-D and object-based method have relatively good capability to recognize tree species. To reach a certainty about this result، it is essential to evaluate UltraCam-D data in other sites with different species.