Drought-tolerant crop disease identification based on attention mechanism

Abstract

Convolutional neural networks are widely used in the field of image classification, but there are still some challenges in the field of crop disease identification. In practice, various background disturbances unrelated to disease identification can greatly reduce the accuracy and generalization of the model. In this paper, we use a residual neural network model to identify a total of 11 species of healthy and diseased leaf images of three drought-tolerant crops: wheat, corn and potato. In this paper, an attention mechanism is added to the ResNet model to exclude the interference problem of complex backgrounds in real environments, and migration learning is used to improve the accuracy rate. The accuracy of recognition reached 95.08%, which is better than ResNet50 model and AlexNet model.