Abstract: The earth-rockfill dam construction in China has been greatly improved in the past few decades. However, with the extension of the service period of the earth-rockfill dam, its underlying culvert pipe diseases, especially leakage phenomenon, need to be diagnosed quickly and accurately. Therefore, a multi-view visual image stitching method is proposed for the diagnosis of culvert pipe diseases in the earth-rockfill dam. By use of closed circuit television(CCTV) pipeline robot collecting panoramic visual images of the culvert pipe in the earth-rockfill dam, based on multi-view geometry three-dimensional reconstruction technology, the three-dimensional point cloud of the culvert pipe is developed. Then, the surface of the model is judged by using three-dimensional surface estimation technology, and the images are flattened and stitched. The digital automatic identification and accurate location of water seepage and crack defection in the culvert pipe images of the earth-rockfill dam are realized, and the diagnosis results coincide with those of the field inspection, which can provide a reference for disease treatment. This diagnosis method has the advantages of flexible, fast and accurate detection and low cost, and has a wide application prospect in the field of the culvert disease inspection of the earth-rockfill dams.