Research And Application Of Tomato Leaf Disease Identification Method Based On Deep Learning

Tomato,as one of the traditional crops in our country,its cultivated area and total output rank in the forefront of the world.However,during the growing process,tomato plants are not only affected by weather,but also vulnerable to many diseases.In recent years,with the progress of planting technology,tomato suffers from more and more kinds of diseases,which have an impact on tomato yield and fruit quality.In case of serious diseases,tomato yield will be significantly reduced and fruit quality will be inferior,thus causing great economic losses for growers.Therefore,it is of great significance to quickly and accurately identify the types of diseases suffered by tomato after it is damaged by diseases,for the prevention and treatment of tomato diseases.Since the further development of deep learning and other related technologies,its research in image recognition has achieved good results,so it has been paid attention by researchers and widely used in disease recognition and other fields.Among them,deep convolutional neural network can automatically extract features from disease images,greatly improving the accuracy of disease recognition.In order to study the difficult problems in the process of tomato disease recognition,to achieve accurate prevention and control of tomato disease.Based on the relevant theories of deep learning,this thesis improved the commonly used convolutional neural network to identify and classify nine tomato leaves after collection and processing,including tomato healthy leaves,tomato early blip,tomato leaf mold.The main research contents are as follows:(1)Construct the tomato disease data set.Based on tomato data in Plant Village,tomato leaf image data was preprocessed,and then data enhancement was used to expand the data set.AlexNet,VGG16,Inception V3,ResNet50,four kinds of classical convolutional neural networks are adopted,and then the network model is used to train and test the data set.Based on the experimental results,the recognition effect of four convolutional network models on tomato disease was analyzed.(2)Propose a scheme using attention mechanism and feature fusion algorithm to improve the performance of the basic model.Three attention modules,SE,ECA and CBAM,were embedded into the model to guide the model to focus on the important spots in the image.The network model with different types of depth features is studied by using feature fusion method.The experimental results show that the recognition effect of the improved experimental model is higher than that of the above four common convolutional neural networks.(3)Development of tomato leaf disease identification system.In order to make the disease recognition results visual and operable,a tomato leaf disease recognition system based on Python language was designed.The system has the function of tomato disease detection and recognition,and can display the disease recognition results and corresponding prevention and control suggestions.In addition,it also includes the function of tomato disease knowledge popularization based on disease symptoms and incidence rules.Among them,the function of disease detection and recognition as the main function can realize the recognition of tomato leaf disease image,and display the recognition results.