Classification Performance of TM Satellite Images

Abstract

In the present work, the remotely sensed images had gone through different stages before it become ready to be classified, firstly by transforming the raw images by using principle component analysis method (PCA). Applying PCA has shown that the first three bands gives 97.86% of the overall information that has been provided from the scene. RGB coloring method is adopted as an unsupervised method to classify the scene according to the RGB color combination for three principle component images, this unsupervised method guide us to a better selection of the ROI (region of interest), it’s becoming more clarity. Histogram equalization method has been used to enhance the colored bands. The results showed that the selecting of ROI from the original TM gave classification accuracy (85.84%), whereas, after applying the RGB coloring model, the accuracy raised to (90.13%). The accuracy has been shown to be equal (97.08%) after applying the enhancement method, and by applying the same methods on the PC- images the accuracy of classification raised to become (98.76%).