Local spatial analysis in surface information extraction of coal mining areas with high fractional vegetation cover using multi-source remote sensing data

The objective of the study is to utilize the local spatial statistics in multi-source remote sensing to analyze and extract surface anomalies in coal mining areas. We illustrated the equations and characteristics of three local spatial statistics, and then calculated the textual bands of them. In contrast with the selected optimal bands, the local spatial analysis improved the classification accuracy from 93% up to 98% based on Supporting Vector Machine (SVM) Classification. In addition, a few Ground Truth Region of Interests (ROIs) were also derived in the multi-spectral image. By means of the hyper-spectral remotely sensed image covering the ROIs, we directly identified six different surface objects or anomalies and inferred that a clustering of minerals and sandy soil with dense vegetation was a developing coalfield, which should be verified in the ground survey.