Coal mine area monitoring method by machine learning and multispectral remote sensing images

Abstract

Coal is currently the main energy in the world. A large amount of gangue is produced during the coal mining process, and the gangue contains a large amount of metal and sulfide, which is one of the environmental pollution sources. Therefore, how to monitor changes in coal mines in real time is an important task. With regard to this problem, we propose a new method of monitoring coal mining area. First, we collect samples of coal, gangue, and other objects from the Shenhua Baori Xile coal mine, and then measure the ground measured spectral reflectance data of these samples. For satellite data, we remove the interference information of the reflectance spectra from these images and established a relationship with the ground measured data. We propose a tree root algorithm and then combine extreme learning machine to build a classification model. Finally, the model is applied to remote sensing images, and a good extraction effect is obtained. Compared with typical monitoring methods, our method has the superiorities of high precision, low-priced, and fast speed, and it can monitor the changes in the coal mine area in real time. This is the basis for improving the environmental pollution around the coal mining area.