Fine identification of vegetation types in open pit mining regions using combined UAV RGB imagery and LiDAR point cloud data

This is the authors' abstract. We don't add key points for this paper.

Accurate vegetation type information is essential for the scientific assessment of ecological restoration in open pit mining areas. However, many existing mine-mapping studies tend to classify all vegetation into a single class, overlooking inter-species differentiation and thereby limiting the precision of ecological evaluations. To address this issue, this study proposed a multi-scale hierarchical classification (MSHC) method for fine-scale mapping of vegetation types in open phosphate mining areas by combining unmanned aerial vehicle (UAV) RGB remote sensing imagery and light detection and ranging (LiDAR) data with the support of object-oriented methods. The results showed that (1) The all features (AF) scheme achieved an overall accuracy (OA) of 97.04% and a Kappa coefficient of 0.98. (2) The proposed MSHC method reached an OA of 97.69% and a Kappa of 0.97. (3) The vegetation in the study area was predominantly grassland (approximately 29.57% of the total area), with Alnus nepalensis being the dominant tree species (16.07% of the area). Overall, this study provides a detailed classification of vegetation types, offering a valuable dataset for ecological monitoring and assessment. The proposed method also presents a transferable and effective approach for the fine-scale vegetation mapping of other open-pit mining environments.