A Tailings Pond Identification Method Based on Spatial Combination of Objects

Abstract

Tailings ponds pose a significant risk to the safety of the surrounding residents and the local ecological environment. Therefore, it is necessary to efficiently and accurately monitor tailings ponds. However, the internal structures of tailings ponds are heterogeneous, and they are typically identified through manual interpretation. In this paper, an identification method for four main structures of tailings ponds, namely, starter dams, embankments, deposited beach, and water body, is proposed based on the spatial combinations among them. First, hierarchy objects were established based on GaoFen-2 imagery. Then, candidate objects (such as embankment-like) were identified using the spectral features and the number of parallel lines. Subsequently, rural settlement-like objects were eliminated as interfering objects based on their shapes and distributions. Finally, four structures of tailings ponds could be identified based on their spatial combination. Six cases of tailings ponds were selected for validation. Interference categories were eliminated step by step (79.64%, 38.21%, and 39.75% for case A), and all four structures were identified. The overall identification accuracies were 88.14%–96.21%. The average accuracy is 32.02% higher than that of comparison experiments using the random forest. The method was proved to be applicable to the automatic identification of mining areas, which is of great significance for efficient and accurate supervision of mining safety.