Accuracy Assessment of Classification on Landsat-8 Data for Land Cover and Land Use of an Urban Area by Applying Different Image Fusion Techniques and Varying Training Samples

This study aims to assess classification accuracies of fused images obtained from IHS, GS, PC, CN, and Brovey fusion techniques. The study area is selected with a variety of land cover features to understand the effect of fusion. A layer stacked method has been experimented in which composite of bands 5, 4, and 3 of Landsat-8 image is resampled and stacked with the panchromatic band and these bands have a highest reflectance of land cover features present in the study area. Classification of reference and fused images are performed by supervised and unsupervised methods and impact of a number of training samples is analyzed. Fused images are assessed for distortion. The accuracy of classification of fused images is measured using error matrix and MLC and NN produced accuracies higher/closer to the reference image. Fusion of Landsat-8 data is found to be appropriate for land use and land cover classification of an urban area.