Quıckbırd görüntülerin sınıflandırma doğruluğunun iyileştirilmesi

Fusion methods increase spatial resolution by combining panchromatic band with multi spectral data. In this study, the results belonging to several methods on multi spectral and panchromatic images are compared. The fusion methods common in literature Intensity-Hue Saturation (IHS), Principal Component Analysis (PCA) and wavelet transformation are used.Supervised classification methods need prior examples for training of the algorithm to classify other unseen samples as opposed to unsupervised classification methods. Image classification is done with well-known supervised methods, e.g. Maximum Likelihood Classifier (MLC) and Support Vector Machines (SVM).Fused images provide better spatial resolution. However, spectral distortions by introducing new colors or artificial structures called artifacts cause decrease in overall accuracy values especially with nonlinear classification methods such as SVM. Enhancement methods are proposed and tested to diminish classification errors caused by spectral distortions present in the fused image.A PCA-based image enhancement technique is found superior to other enhancement techniques applied.Class separability is better after enhancements compared to the original reference satellite image. Using the classified result of the original multi spectral image as a benchmark, improvement in the overall accuracy of classification results of the enhanced fused images are observed. Higher classification accuracy is possible with the integration of image fusion and image enhancement methods into classification process. Results supporting the advancement are given in the related parts of the thesis.