Gradation recognition of coarse-grained soil based on searcher-analyzer deep learning network (SaNet)

The coarse-grained soil is widely used in embankments, earth-rock dams and other fill projects. However, the traditional sieving method is time-consuming and inefficient, failing to meet the rapid quality testing requirements for gradation. To address these issues, an "image-gradation" relational database is established for yellow river silt and quartz sand coarse-grained soil, comprising 22380 photos. In response to the mismatch between two-dimensional image and three-dimensional gradation, a searcher-analyzer network (SaNet) is developed to handle any number of image inputs. The model accuracy steadily improves with an increase in the number of images, with average errors of 1.63% and 1.21% for the recognition of yellow river silt and quartz sand gradations, and the coefficient of determination of 0.995 and 0.992, respectively. The results demonstrate that the proposed deep learning model on the SaNet architecture exhibits high accuracy in gradation recognition, meeting the real-time non-destructive gradation detection requirements in fill projects.