Remote Monitoring Algorithm for Vegetation Recovery on Mining Industry Sites

The aim of this research is to develop an algorithm for the automatic analysis of natural vegetation recovery on disturbed lands for monitoring open-pit mining sites based on remote sensing data (RSD). Images from Sentinel-2 satellites of mineral extraction areas in open-pit mines were chosen as the initial data. The feasibility of using a convolutional neural network for determining the boundaries of the object under study by forming its mask is considered. The modified Normalized Difference Vegetation Index (mNDVI) is used as an additional channel for the neural network, which helps to increase the accuracy of defining the mentioned boundaries. To divide the quarry territory into three classes (water, vegetation, soil) and to construct the corresponding class map, the Modified Normalized Difference Water Index (MNDWI) and the Soil and Atmospherically Resistant Vegetation Index (SARVI) are applied to the obtained mask. The class map allows for more detailed monitoring and analysis of the studied objects.