Intelligent detection and safety risk assessment of waste disposal sites based on high-resolution remote sensing imagery

As a by-product of industrial and mining activities, waste disposal sites (WDS) pose serious environmental and safety challenges. To address this, an intelligent identification and safety risk assessment method for WDS based on high-resolution remote sensing imagery is proposed. Namely, model performance was evaluated via ablation experiments, while WDS risk levels were classified using the Analytic Hierarchy Process (AHP). The DeepLabv3+ backbone is enhanced by Squeeze-and-Excitation (SE)–based recalibration and transposed-convolution up-sampling for overall performance improvement. The optimal model was applied in Luding County, West Sichuan, to identify the number and distribution of WDS. A safety risk assessment model integrating 10 factors, including slope gradient, important infrastructures, and environmentally sensitive points, was established using the Analytic Hierarchy Process and validated by the Consistency Ratio (CR). The four risk levels and their corresponding thresholds were classified based on a normal distribution function. The risk level of each WDS was obtained, and the influence of the slope on WDS stability was discussed. In summary, this study proposes a method for large-scale WDS distribution and safety risk assessment using a deep learning framework.