Study On Classification Of Ejina Oasis Landscape Based On Landsat ETM Date

Remote sensing image classification is the use of the energy characteristics of the spectral features and structural differences in the characteristics of differences to identify the features in a certain period of time information.Remote sensing image automatic classification of the computer a great practical value, but the current classification of the methods and accuracy are yet to be improved.To improve the accuracy of image classification has been the application of remote sensing technology and one of the core study.This paper studies Ejina Oasis, the use of RS and GPS technology to ETM data for the sources of information, the use of ENVI remote sensing image processing software for data processing, on the main landscape types Ejina Oasis classification, and classification accuracy of the results of comparative analysis. Classification of non-supervised classification methods including K-means algorithm of classification, supervised classification of the maximum likelihood classification and decision tree classification.Confusion matrix using the method, the results of three classification accuracy analysis. The results show that, K-means algorithm for maximum likelihood classification and overall classification accuracy of 86.32%, respectively, 85.82%, accuracy is not high, the effect on the classification of vegetation is not ideal. In the decision tree classification, spectral features of the data analysis, NDVI Analysis, KT transform and principal component analysis, the establishment of a reasonable decision tree, the overall classification accuracy of 93.28%.Application of automatic decision tree classification techniques, Ejina Oasis classification of ETM image automatic classification accuracy than the traditional method had a greater increase in the study area has reached the purpose of the landscape type.