Deep convolutional neural networks for surface coal mines determination from sentinel-2 images

Coal is a principal source of energy and the combustion coal supplies around one-third global electricity generation. mines are also an important CH4 emissions, second most greenhouse gas. Monitoring emissions caused by mining using earth observation will require exact location mines. This paper aims to determine surface from satellite images through deep learning techniques treating them as land use/land cover classification task. achieved Convolutional Neural Networks (CNN) that has proven be capable complex tasks. With list known mine locations various countries, training dataset “Coal Mine” “No image patches prepared Sentinel-2 with 13 spectral bands. Various pre-trained CNN network architectures (VGG, ResNet, DenseNet) trained validated our 3500 3000 patches. After several experiments VGG combined transfer found optimal model for this Classification accuracy 98% been validation architecture. The produces more than 95% overall when tested on unseen different countries outside evaluated against visual classification.