Abstract: Blast fragmentation is a critical indicator for evaluating bench blasting performance and optimizing blasting parameters in open−pit mines. To overcome the limitations of traditional close−range imaging methods in large−scale muckpile scenes—such as restricted coverage, reliance on manual scale calibration, and occlusion issues inherent to single−viewpoint imaging—an intelligent fragmentation assessment method integrating UAV oblique photography, airborne LiDAR point clouds, and Crowd−SAM instance segmentation is proposed. Multi−view images and LiDAR point clouds of the bench area are synchronously acquired before and after blasting using a UAV to construct a 3D digital model of the muckpile. Two−dimensional projection images are generated under real−scale constraints, and automatic rock fragment segmentation is achieved with Crowd−SAM. By fusing 3D spatial information, the rock size, shape, and spatial distribution are inverted, establishing a quantitative evaluation workflow for muckpile fragmentation. The approach is validated at an open−pit mine in Zhaoqing City, Guangdong Province. At a flight altitude of 20 m, the LiDAR point cloud density reaches approximately 3300 points/m² with an average point spacing of about 1.7 cm, yielding centimeter−level reconstruction accuracy and a theoretical fragment identification resolution of approximately 3 cm. The results demonstrate that the method enables non−contact, high−precision acquisition and spatial statistical analysis of large−scale muckpile fragmentation, providing a new technical pathway for blasting effect evaluation and blasting parameter optimization in open−pit mines.