Research On Tomato Leaf Disease Classification Based On Convolutional Neural Network

As an important export crop among my country’s cash crops,tomato has a very important position in Southeast Asia,the Middle East,the Near East and other places as well as some European countries and regions.At the same time,tomato cultivation is relatively extensive in my country’s current agricultural products,and the two most important and important factors for evaluating and considering the quality of tomatoes are their quality and yield.In the whole process of the tomato growth cycle,the prevention and control of various diseases has a very important influence and effect on its own yield and quality and the subsequent economy.According to the relevant academic research and literature of relevant experts and scholars in the same industry,we can find that the traditional method of distinguishing artificial disease categories has relatively poor generalization ability,so that artificial disease recognition is difficult to meet the high-efficiency requirements of tomato in modern agricultural production.The excellent performance of convolutional neural network in the related fields of image recognition technology has given new ideas and methods for the recognition of tomato leaf diseases.Aiming at the problem of low accuracy of traditional tomato disease recognition,this paper studies the recognition method of tomato leaf disease based on convolutional neural network to realize automatic recognition of tomato leaf disease.Mosaic virus disease,early blight,late blight,leaf mold,A total of five tomato leaf disease images of yellow leaf curl virus disease were used as research objects.In order to prevent the images from being interfered by different light intensity and size,the images were normalized by size and grayscale.At the same time,the application of convolutional neural network in tomato leaf disease identification is further studied.This paper proposes a method of tomato leaf disease recognition based on convolutional neural network,which improves and optimizes the traditional AlexNet model.This algorithm modifies the number of neurons in the fully connected layer in the classic AlexNet,so that the entire model can increase the training speed higher.In the training process,it is proposed to use mini-batch hyperparameters to make the model converge better.Finally,by comparing the accuracy of the traditional AlexNet,Goog Le Net and the improved AlexNet three models,it is concluded that the improved AlexNet model has the highest accuracy in identifying tomato leaf diseases,reaching96.36%,achieving accurate disease identification..Finally,this experiment builds a tomato leaf disease recognition system based on Matlab GUI.The function of the system mainly realizes the call of the network model,the adjustment of network parameters,the network training and the final feedback of the disease recognition results.It is realized that when tomato leaves are diseased,users only need to upload pictures to get the results of disease diagnosis,which simplifies the limitation of long time for expert diagnosis and facilitates faster decision-making.