Paddy field weed identification method fusing CNN and Transform dual networks

The invention discloses a CNN and Transform dual network fused rice field weed recognition method, which comprises the following steps: S1) firstly, acquiring a rice field weed image by using a camera in a high-angle shooting or oblique shooting manner, and constructing a weed image data set; s2), constructing a model integrating CNN (Convolutional Neural Network) and Transform dual networks; s3) inputting the weed image data set constructed in the step S1) into the constructed fusion CNN and Transform dual network model for training, and obtaining a trained network model weight file; s4) inputting the to-be-identified weed image into the trained network model weight file, and outputting the to-be-identified weed species. The CNN network is mainly used for extracting local features, and the Transform network is mainly used for extracting global features, so that the representation capability of the network is greatly enriched, the paddy field weeds are identified by the dual network, and the weed identificati