Abstract: To mitigate the interference of construction dust on tunnel face observation during roadheader excavation and eliminate the subjectivity associated with manual rock mass assessment, an intelligent recognition method is proposed for quantifying rock fragmentation at the tunnel face by integrating unmanned aerial vehicle (UAV) autonomous inspection with computer vision techniques. First, the Fast-Planner algorithm is employed to achieve UAV autonomous obstacle avoidance and path planning, enabling continuous acquisition of 412 high-resolution RGB images from 100 excavation cycles at a high-speed railway tunnel construction site in Yunnan Province. Then, a U-Net++ network is used for feature extraction, and kernel density estimation is applied to fit the probability distribution of the fragmentation ratio k, showing that the main density peak is located near 0.11. Finally, a cuttability evaluation table for roadheader excavation is developed based on the value of k. The results show that the proposed method achieves an 83.2% accuracy in extracting rock mass features, significantly outperforming traditional manual assessment and providing a feasible solution for safe, efficient, and intelligent tunnel construction evaluation.