The mesostructure of the asphalt mixture, being intricate and varied, poses a significant challenge in accurately analyzing its distribution throughout various stages of compaction. CT scanning was employed to obtain digital images of PAC-13 asphalt mixture, convolutional neural networks were utilized for multi-component segmentation and three-dimensional reconstruction, with the aim of studying the evolution of the mesostructure of PAC-13 during the compaction process and correlating it with its macro-performance. The results showed that as th...