Abstract: In recent years, the increasing intensity of infrastructure development in mountainous regions, coupled with the rising frequency of extreme meteorological events, has led to a growing incidence of rockfall hazards, particularly in northwestern China. Weathering processes often cause unstable rock masses to fragment upon detachment. However, the current rockfall risk assessment methods mostly ignore the impact of rockfall fragmentation and debris shape characteristics on the rockfall trajectory and energy dissipation, which significantly affects the reliability of assessments. Consequently, quantifying the risk of rockfalls to mountainous towns has become a critical issue. Focusing on the high and steep back-mountain collapse in Chengguan Town, Zhouqu County, Gansu Province, this study proposed a quantitative risk assessment method integrating 3D modeling with fracture shape characteristics. Potential hazardous rock masses were identified through UAV close-range photogrammetry and point cloud data. A 3D model was reconstructed using Blender software, while a fracture simulation program was developed based on the PhysX physics engine. Additionally, the Tyson polygon fractal fracture model was employed to dynamically simulate the shape evolution of rockfall fragmentation. The results indicate that the volume distribution of fragmented rocks follows a power-law distribution (R² = 0.918), and the shape parameters demonstrate a flattening trend of fragments (mean sphericity = 0.75). For instance, hazardous rock W1 generates the highest impact energy (3314.7 kJ) on structure C1 within the disaster area on the left side of the slope foot, with a medium risk level. In contrast, hazardous rocks W2 and W3 exhibit the greatest impact on structures C2–C3 in the central road area. This study provides essential data support for the refined prevention and control of fractured collapse hazards.