Method and device for constructing rice disease recognition model

The invention discloses a rice disease recognition model construction method, which comprises the following steps: acquiring a rice disease image set as a first image set, and constructing a disease image small sample library; selecting image features with representation meanings in the small sample library for semantic description, and constructing a disease semantic library; calculating small sample features and semantic mapping by using a Dual-TriNet algorithm, and reversely generating an amplified feature image set as a second image set by using semantics and mapping; constructing a second learning sample library by using the second image set and the first image set; and constructing a rice disease recognition model based on a transfer learning algorithm by using the second learning sample library. After a disease image small sample library is constructed, images in the small sample library need to be preprocessed, and the preprocessing is that image features with representation significance are analyzed