Classification of Land Cover and Land Use using Deep Learning

The utilization of land is a major challenge in current era. Land cover refers to the surface cover on the ground. It may be vegetation, grass, water bodies, bare land or any other use. Land use refers to the purpose the land serves, for example, residential, wildlife habitat, or agriculture. The databases of land cover and land use becomes outdated quickly, hence, an automatic update process is required. The present approach to determine land cover and to classify land use objects based on convolution neural networks (CNN) and to study the effects on changing parameter on the results. The input data for proposed approach are aerial images from Sentinel-2 satellite images. Land cover and land use for each image has been determine with the use of CNN. The present work also describes the effect of changing parameters on our results and output generated in each case. Comparisons of our results with different existing algorithms have also been analyzed. Experiments show that overall accuracy of the proposed approach is 93-95% for land cover and land use. The classification of land cover and land use has a positive contribution toward the utilization of land by humans.