Safety monitoring of scaffolding operations is essential for preventing accidents in high-altitude construction. This study proposes an integrated computer vision and multi-criterion decision-making (MCDM) framework that combines object detection, pose estimation, Analytic Network Process (ANP) and ELECTRE III methods to evaluate safety risks of construction workers. Specifically, computer vision techniques are employed to extract objective visual evidence related to workers’ behaviors, protective equipment (PPE) usage, and working environments, which serve as the basis for subsequent safety risk quantification. A four-criterion system, including action risk, PPE compliance, working height, and structural integrity, is established. Weights are determined via the ANP, and risk ranking is conducted using ELECTRE III. Experiments on a self-built dataset achieved an mAP@0.5 of 92.3%, a segmentation IoU of 67.2%, and a pose OKS@0.5 of 89.6%. The evaluation results correlate strongly with expert assessments (Kendall’s τ = 0.79). The proposed framework effectively identifies unsafe behaviors and quantifies safety risks, providing reliable decision support for intelligent construction safety management.
Jin et al. (Wed,) studied this question.