The invention discloses a keratoconus recognition method and device based on multi-dimensional feature fusion, and the method comprises the steps: 1), carrying out the data processing through a designed standardized flow through the multi-dimensional data of a single patient, carrying out the grading of keratoconus, and dividing the keratoconus into a normal keratoconus, a subclinical keratoconusand a keratoconus; 2) constructing a multi-dimensional model, carrying out feature fusion in various modes, connecting an SE module to carry out feature recombination, and then carrying out model training; 3) carrying out interpretability operation of category judgment by utilizing gradient back propagation, and outputting a visual image; (4) comparing influences of one-dimensional convolution andtwo-dimensional convolution on the keratoconus data model for special consideration of the c keratoconus data model, and (5) selecting the category where the maximum score in the three scores is located as the discriminated category through the designed model to obtain a good classification effect, and optimizing the problem that the conic cornea recognition effect is poor in practical application.