Evaluation of land use/land cover classification accuracy using multi-resolution remote sensing images
Abstract: Timely and accurate land use/land cover (LULC) information is requisite for sustainable planning and management of natural resources. Remote sensing images are major information sources and they are widely used for mapping and monitoring various land features. Images from various sensors, with different spatial resolutions, are available; however, the selection of appropriate spatial resolution is an essential task to extract desired information from images. This paper presents the conclusions of the work related to LULC classification based on multi-resolution remote sensing images. Optical data collected by three different sensors (LISS IV with 5.8 m and Landsat 8-OLI with 30 m and AWiFS with 56 m spatial resolutions respectively) in 2013 are examined against the potential to correctly classify specific LULC classes. The classifications of images are performed using Maximum Likelihood Classifier (MLC). The results indicate that the overall accuracy and kappa coefficient of LISS IV with 5.8 m are higher than that of Landsat 8-OLI with 30 m and AWiFS with 56 m images. Understanding the role of spatial resolution in LULC classification accuracy will enable the appropriate interpretation of any classified images.