Abstract:
The paper addresses the problem of comprehensive mining and environmental monitoring of mineral resource complex facilities using Earth remote sensing data. The structure of an information system is presented, including modules for data management, processing and analysis of satellite imagery. Earth remote sensing data were used as source information. Processing and analysis of this information was performed to assess the impact of geo-resource extraction and processing on the state of the natural environment at mining enterprise locations and adjacent territories, and to study biota restoration in disturbed areas. Convolutional and transformer neural networks are applied to automate the identification of mining complex objects. Segmentation models for dredge tailings, open-pit mines, tailing dams and coal dust pollution were trained on prepared datasets including annotated satellite data of studied objects in seven regions of the Russian Federation. Methods are proposed for: assessing the impact of coal dust pollution on vegetation based on vegetation indices; mapping forest types using a fully-connected neural network; calculating volumes of dredge tailings based on digital elevation models, polygon centerline and Voronoi diagram construction algorithm; and determining heavy metal content in soil based on analysis of samples collected over a multi-year period using mathematical statistics methods. A correlation analysis was performed on data regarding element content in soil samples and water in the zone of influence of polymetallic ore mining. The proposed methods were tested in the territories of mining enterprises in Khabarovsk Krai. Point analysis of accumulated results from long-term observations and current data on the present state of natural environment objects allows forecasting the development of the studied natural-technical systems in the medium term.
Fulltext:
PDF file (3374 kB)