Classification of tree species in hyperspectral image

Hyperspectral remote sensing explores the surface of the earth using sensors that deliver large amounts of data. Satellite collects the data and sends it to ground station for processing. The work focuses on classification of tree species (like Bamboo, Arjun, koroi, Ejar, Mohr) in the forest and it is very important in forest records. In late years, numerous studies have been done using various types of hyperspectral sensors under diverse forest conditions. The complexity present in hyperspectral sensing is mainly due to the curse of dimensionality. A possible solution to overcome the dimensionality problem is to select the features using feature-selection technique. Features which increase the certainty of a given situation are considered more informative than other features. Sequential forward floating selection is the feature selection algorithm used in this paper. It selects the features with highest information so that it is easy for the classification. Classification is done after feature selection. Self organizing map, an unsupervised learning technique which is used to classify the species of tree. Further optimization can be done if more number of features are selected which may increases the classification accuracy considerably.