Automated detection of illegal nonmetallic minerals mining places according to Sentinel-2 data

Abstract

Currently, the problem of illegal mining is still acute. Such illegal use of natural resources harms the environment and leads to irrational use of mineral resources. Modern methods with the use of remote sensing technologies will effectively detect such law violations. In the current study, a method for automatically detection of non-metallic mineral extraction sites based on remote sensing data analysis has been developed. The study uses Sentinel-2 satellite images with spatial resolution 10 m and 20 m and considers four types of minerals: sand, clay, carbonate rocks and sand gravel mix. The spectral indices help to determine the specific quantitative characteristics of the mineral resources. The result is probability maps with mineral resourses characteristics in each pixel. In order to determine to which of known classes relates the point, you need to find the covariance matrices for all classes and take the class with the smallest Mahalanobis distance to the point. Based on the obtained probability maps, an analysis of the applicability of the selected spectral indices was performed, as well as a visual assessment of the quality of interpretation. For each spectral channel and index, two frequency histograms were created to determine how different the channels values and spectral indices on the entire scene and at the reference objects. Each object found by the program was checked for it presence on the studied territory. The developed system is a modern, secure, non-contact method for the rational land use monitoring and natural resources extracted by open-pit mining study.