Fecha de publicación: 29/12/2023 Fuente: WIPO "processing tomato" The invention discloses a tomato leaf disease identification method based on VC-Net and a Mueller algorithm. The method comprises the following steps: preprocessing a tomato leaf disease picture; according to the method, a training module based on an improved visual Transformer tomato leaf disease recognition model and a Mueller algorithm is constructed, patch embedding layer blocking is carried out, then preliminary convolution feature extraction is carried out, Position embedding is carried out, and after feature extraction, a Gaussian denoising technology is used to carry out denoising processing on noise generated by the feature extraction. An LSA local self-attention mechanism is used in Attention of a visual Transform, and a zero mask matrix is introduced. Training is conducted through a Mueller algorithm module, the optimal tomato leaf disease recognition model weight is obtained, and classification of tomato leaf diseases is achieved. Compared with the prior art, the tomato leaf disease category is determined through a method of first classification and then identification, the training time is short, and the identification precision is high.