Classification of worldview 2 satellite image by using object-based technique to identifying the infection of Zagros forests by Loranthus europaeus

Yellow mistletoe (Loranthus europaeus) species is a semi-parasitic plant threatening the Zagros forests, hence idendification of infectious areas are important for its control and management. For this purpose, a forest patch ca 37 ha with different intensities of yellow mistletoe was selected in Ilam province. In order to classify the yellow mistletoe, worldview 2 satellite image dated November 14, 2010 was used. After radiometric and geometric corrections, the image was segmented by NDVI and PCA as thematic layers with different band weights and 29 scale parameter. Different algorithms such as K Nearest Neighbor (with different K parameter), Support Vector Machine (with different C parameter), and Random Forest (with different number of trees) based on object-based approach with 18 spectral and shape features were then compared by using 312 ground truth points. The overal accuracy for K Nearest Neighbor, Support Vector Machine and Random Forest algorithm were obtained 85.1%, 87.4% and 92.9%, respectively for infection classifaication into four categoreis (non, low, mediom and severe infections). Random Forest algorithm with 1000 trees was the best one in indentifying the various intensities of infections. It is concluded that identification of yellow mistletoe in Zagros by using worldview 2-satellite image and object-based classification is possible.