Crop Disease Recognition Method Based on Improved Channel Attention Mechanism

Aiming at the problem of low accuracy of existing crop disease recognition methods, an improved crop disease recognition method based on dual channel effective attention is proposed by improving attention mechanism and residual network. The attention module uses dual channel one-dimensional convolution to filter effective feature information, and establishes a dual channel attention model. Adaptive convolution kernel parameter
$K$
and super parameter
$\alpha$
are introduced to participate in the reverse training of the model, so that the model can select effective channel features independently. Then, it is added to the residual module to recalibrate the channel features to improve the feature expression ability of the residual module, and an improved model is proposed to realize crop disease recognition. The improved attention-based crop disease model was applied to cucumber disease data set, and the disease recognition rates were 86.35% and 99.74%, respectively. The accuracy and convergence rate of crop disease recognition were improved, and the extraction and calculation of redundant features were reduced.