Enhanced mine road monitoring using unmanned aerial vehicles and deep-learning approach

Abstract

Mine roads are highly vulnerable to deterioration as a result of substantial vehicular traffic and excessive precipitation. Therefore, it is critical to incorporate safety monitoring of the mining road infrastructure into mining operations. This study aims to extract mining road objects and classify road conditions using the geometric standards of mining road safety. The study area comprised one of the pits of Perseroan Terbatas (PT) Berau Coal, Indonesia. This study developed a novel approach utilizing geospatial artificial intelligence that integrates drones, deep learning, and spatial geometry to monitor road conditions in terms of geometrical aspects such as slope, road width, turning angle, and cross-fall. The detection results of the mining road resulted in 30 parameter combination schemes, with the optimal scheme consisting of a combination of an 87.5% model overlap, a 512 × 512 tile size, and a 0.3 m pixel resolution. This scheme achieved an accuracy of 89%. Furthermore, the spatial geometric analysis results revealed that 64% of the grade standards, 90% of the turning angles, 61% of the cross-falls, and 51% of the road widths in the mining road area conformed to the internal mine road standards of PT Berau Coal. The method developed in this study can also be applied to other mining areas. The mining road monitoring method developed in this study can improve the mining road standard and may be applied in other mining operations to improve efficiency and safety.