Decision Tree and Its Application in Remote Sensing Image Classification

Taking TM images in Bomi county of Tibet as the object to study, choosing and analyzing the reflect spectral characteristics of typical land-cover types, a set of accurately and fast rules for TM images classification has been established based on spectral characteristics diversity of different land-cover types and artificial visual interpretation.The classification decision tree was applied to classify the TM images in experimental area, and the classification results were compared with traditional maximum likelihood.The results show that: the overall accuracy and Kappa coefficient of decision tree reach 97.15% and 0.9644, respectively, whose performance is more excellent than the traditional maximum likelihood method, and obtain a more satisfactory classification.