Research on crop disease recognition based on uniting multi-layer features

Traditional image processing has many problems in crop disease recognition, such as complicated manual design and low efficiency. This article studies the application of deep learning algorithms in crop disease recognition. This article analyzes the characteristics of feature maps extracted from different convolutional groups of ResNet, proposes multi-layer features united ResNet(MLFU-ResNet) for crop disease classification, and introduces label smoothing to modify the loss function. The experimental results show that the network structure proposed in this article has achieved better classification performance in crop disease data sets. Compared with ResNet, the classification accuracy of the network structure proposed in this article has increased by 0.8%.