A Novel Feature Level Fusion Method for Classification of Remote Sensing Images

Abstract

Feature level fusion approach is utilized in this paper to classify remote sensing images. Texture features are extracted from panchromatic images using mixed Gabor filter (GB), fast gray level co-occurrence matrix (GLCM) and linear binary pattern (LBP). The resultant texture features are classified using nearest neighbor (k-NN) classification method. Spectral features are extracted from the MS image and segmented using over segmented k-means algorithm with novel initialization (OSKNI). Finally the segmented MS image and grid classified PAN image are fused to get the final classified result. To evaluate the performance of the proposed method we used kappa statistics like, Users Accuracy (UA), Producer’s accuracy (PA), Overall classification accuracy (OCA), Expected Classification Accuracy (ECA) and KHAT values. Keywords: Texture, spectral, panchromatic, multispectral, segmentation Cite this Article Shashidhar Sonnad, Lalitha YS. A Novel Feature Level Fusion method for Classification of Remote Sensing Images. Journal of Remote Sensing & GIS. 2019; 10(1): 58–65p.