Extraction of Information from Multispectral and PAN of Landsat Image for Land Use Classification in the Case of Sodozuria Woreda, Wolaita Sodo, Ethiopia

Abstract

High-resolution and multispectral remote sensing images are an important data source for acquiring geospatial information for a variety of applications. The satellite images at different spectral and spatial resolutions with the aid of image processing techniques can improve the quality of information. More specifically, image fusion is very helpful to extract the spatial information from two images of different spatial and spectral images of same area. The Image fusion techniques are also helpful in providing classification accurately. In order to improve the information contents of the remote sensing satellite images at a specific spatial resolution, the different resolution image fusion techniques like Wavelet, PC and IHS have been used to combine panchromatic and multispectral datasets of Landsat ETM+ for the purpose of information extraction. The image under study has been used to identify existing Land use types and perform supervised classification. It has then been identified that forest land, farm land, bare land and built-up area are the most dominant land uses in the study area. Based on the supervised classification, classification accuracy assessment has indicated that Original image (MS) produced 83.33% overall accuracy and 0.7500 Kappa coefficient, PC fused image produced 91.67% overall accuracy and 0.875 Kappa coefficient, IHS fused image produced 86.67% overall accuracy and 0.800 Kappa coefficient, Wavelet-PC based transformation produced 91.67% overall accuracy  and   0.875 Kappa coefficient and Wavelet-HIS based  transformation produced 98.33% overall accuracy and 0.975 Kappa coefficient. Moreover, Wavelet-HIS based transformation method has produced relatively higher accuracy. Generally, based on the overall accuracy and kappa coefficient, fused images in terms of classification accuracy at the expense of information content perform by far better than the original image.