Cow face recognition method in complex cattle farm environment

The method aims at solving the problem that cattle identity recognition is difficult due to the fact that cattle herds are blocked, cattle faces are dirty, and the active states and postures of cattle are diversified in a complex cattle farm environment. The invention discloses a cattle face recognition algorithm in a complex cattle farm environment based on a Vision-Transform algorithm. On the basis of the VIT model, a brand new feature fusion mode is designed for the complex cattle farm environment. A patch-shift network layer provided by the invention is added into a VIT model, a shift module in the patch-shift network layer fuses information between features, local features are utilized, information between the local features is also fused, and the influence of smudginess in a cattle farm environment on recognition is relieved. A learnable Mask matrix is added, so that the model inhibits image blocks without cattle face information, image background interference is inhibited, the model pays more attention