Image Recognition of Maize Pests and Diseases Based on Convolutional Neural Network Algorithm

Crop disease recognition techniques based on convolutional neural networks, although accurate, are difficult to be extended in complex environments due to the limitations of sample collection and research objects. In this paper, an improved convolutional neural network approach is proposed as an example of maize disease identification, focusing on the identification and classification of pests and diseases affecting maize crops through image analysis using improved convolutional neural networks (CNNs). The factors affecting the application effect of the dataset were investigated by processing the dataset with magnification, scaling, background removal and local fine segmentation, and the optimised recognition model was used to design a disease recognition system based on mobile terminals and validated in the field. By comparing the recognition of four diseases and pests, including round spot, grey leaf spot and rust of soil maize, the results show that the recognition accuracy of the model can reach 0.9742 and the misclassification rate is 0.0258, which is improved compared with other models.