DATA MINING TECHNIQUES FOR SEPARATION OF SUMMER CROP BASED ON SATELLITE IMAGES

Due to the difficulty in discriminating soybean and corn in mappings obtained by the time series of satellite images, this study aimed to apply the data mining techniques to separate soybean and corn. Pure pixels selection from Landsat-8 were extracted and used to build a standard spectro-temporal EVI profile for both crops. These profiles were obtained with the Timesat software and, further incorporated in the Weka software. Five out of eleven variables of the standard spectro-temporal EVI profile for each crop were found through the decision tree, a data mining technique. These five variables were sufficient to achieve the separation of soybean and corn crops with an accuracy of 96.3% and a kappa index of 0.92.