Land cover image classification using adaptive sparse fusion classifier

The presence of large number of spectral bands in the remote sensing images results in difficulty of identifying various land cover regions. Land cover classification is one of the recent researches which find more application in satellite image processing. It is important to recognize different land classes from a multispectral satellite image as the raw image contains noises and less clarity. The work was done in three stages such as preprocessing, feature extraction and classification. Noise present in the images are removed using a non-local means filter in preprocessing. Gabor wavelet and GLCM (gray level co-occurrence matrix) techniques were compared for feature extraction where PCA uses better in dimension reduction. The proposed sparse classifier efficiently classifies the given multispectral satellite image. It identifies the scattering features of same group of textures in the image, produces better accuracy compared to other techniques.