Vulnerability assessment constitutes a core component of disaster risk evaluation and management, aiming to quantify the relationship between building damage and disaster intensity through statistical analysis or physical mechanisms. This study utilizes the 21 July 2024 debris flow event in Shuihaizigou Gully (Daliang Mountain) as a representative case study. Unmanned Aerial Vehicles (UAVs) integrated with Simultaneous Localization and Mapping (SLAM) technology were employed to construct high-precision three-dimensional (3D) reality models of damaged buildings. Based on these models, a building damage database was established to derive building vulnerability curves. The analysis revealed significant vulnerability disparities across building types. Furthermore, the study uncovered nonlinear functional dependencies between debris-flow parameters (flow depth, flow velocity, and impact pressure) and building damage. For brick-timber structures, complete destruction occurred at impact pressures exceeding 22.75 kPa, with corresponding threshold flow depths and flow velocities of 1.27 m and 3.48 m/s, respectively. Additionally, a comparative analysis of two Torrential Fan models was conducted to explore the spatial characteristics of building vulnerability. Results indicated that the model incorporating debris-flow branching dynamics, which simulate flow bifurcation and superposition, yields higher accuracy in reflecting the spatial distribution of vulnerability compared to the traditional torrential fan model.