Author(s)

Yang Li, Xiaoli Liu, Haitao Yang, Xixuan Zhou, Qian Wang

Corresponding Author

Haitao Yang

Abstract

The increasing demand of mining resources, the more prominent the contradictory characteristics of frequent mining and ecological environmental protection, for which how to effectively extract large-scale mining points is not only a prerequisite for building a basic database of mines, but also a data support for monitoring the ecological environment of mining areas. In this paper, based on object-oriented technology and random forest, we use Sentinel-2 as the data source, image segmentation to build a multi-dimensional classification feature library of spectrum, topography, texture and geometry, and establish a sample database of mining areas in the Yili River basin with the assistance of mineral rights data; finally, we implement random forest model construction, parameter optimization and classification in Python. The results show that (1) the object-oriented technology can weaken the "pretzel effect" to a certain extent, and the scale, shape, and tightness factors are set to 100, 0.1, and 0.5 to ensure the integrity of the mine area; (2) the optimized random forest parameters NST and MF are 5 and 1000 respectively to meet the relationship between model accuracy and efficiency, and the feature (2) The importance assessment shows that there is a large correlation between the mine area and the spectrum and topography, and only the standard deviation feature has a large correlation among the texture features; (3) There are 395 mine sites in the Yili River basin, with a total area of 143.220km2, which are concentrated in the middle and low elevations and close to the river, spatially reflecting "group", in terms of quantity, the mine area is mostly in the range of 0-0.5km2., and the non-metallic mining sites are the most abundant, with an average distance of 9364.69m from the river, and the metal mining sites are the closest to the river.

Keywords

mining area; object-oriented; random forest; Yili River basin; ecological environment