The exposed rockfill on the lift surface of rock-filled concrete (RFC) dam increase shear resistance at the interface between upper and lower layers, which is crucial to the stability of the dam, and the projected area proportion of the exposed rockfill is an important index for the scientific evaluation of the interlayer shear performance. In this study, the latest international Meta AI model, known as segment anything model (SAM), was utilized for automatic image segmentation of RFC exposed rockfill. The SAM-identified images were further reprocessed and analyzed by ImageJ, which involved techniques such as smoothing, differential algorithm, and median filtering for the accurate location of the exposed rockfill. The binarized images were then used to calculate the exposed rockfill proportion. The results show that SAM image pre-segmentation can identify about 90% of the exposed rockfill, and the secondary image processing by ImageJ can effectively improve the identification accuracy of small rocks, within an error of ±3% compared to manual annotation results. Then, this methodology is applied to two reservoir projects in Guizhou Province, each lift surface was pre-processed into different zones. We found that the exposed rockfill proportion near the upper, middle and lower reaches are quite different, mostly falls in the range of 10%-30%, among which the exposed rockfill proportion in the transport area is quite low. The research results and findings can provide some reference for the study of interfacial shear performance, as well as the safety and stability of dam reservoirs.