Deep Learning-based Change Detection Method for Environmental Change Monitoring Using Hyperspectral Images

Land monitoring is a dynamic process that is subject to permanent change and transformation over time under the influence of various natural and human factors. Since solving problems related to change detection manually is a time-consuming operation, for this reason, in this paper two methods of change detection based on deep learning algorithms are presented to generate a map of changes. The purpose of using DL algorithms and especially CNN's is monitoring environmental change in to change and no change classes. In order to evaluate the capability of the first proposed method, hyperspectral images were used. This method divided into two phases: generating training data and accuracy assessment by CNN parameters. The OSCD datasets were used to evaluate the second proposed method. In this method, the networks based on Unet. The proposed frameworks are automatic and capable of extracting information with high precision. In each method, the overall accuracy is over 95% and the kappa coefficient is close to one.