Study and Validation of a Remote Sensing Model for Coal Extraction Based on Reflectance Spectrum Features

Abstract

A remote sensing model to extract a coal area was developed by analysis of the features of the coal reflectance spectrum. The spectrum data of coal and other object samples from the Huolinhe (HLH) mining area in Inner Mongolia and the Huozhou (HZ) mining area in Shanxi Province in China were measured with a handheld field spectrometer, which showed significant differences in the lignite from the HLH mining area and the bituminite from the HZ mining area. Spectral variability was observed between coal and other objects in the mining area. Based on the measured result of the spectra, a model was proposed to extract a coal area with two indices: the normalized difference coal index based on the difference between bands 4 and 5 of the Landsat/Thematic Mapper imagery and the threshold from the reflectance of band 5. The generality of the model is verified through extraction of the coal area of HLH, HZ, and other mines. The proposed model and the random forest method were compared to validate model accuracy. The remote sensing model shows potential for exploring coal resources and monitoring environmental variations in coal mines.