Design And Implementation Of Fruit Tree Pest And Disease Classification System Based On Deep Learnin

Fruit tree cultivation in China is among the highest in the world,and fruit tree crops occupy an important position in agricultural production,and it is vital to ensure fruit production and quality safety.Among them,the pests and diseases of fruit tree leaves have a greater impact on fruit production,reducing its yield and causing considerable economic losses.Therefore,as a reference for the staff,it is important to develop a classification system for fruit tree pests and diseases.Up to now,the classification of fruit tree pests and diseases has mainly been artificially diagnosed by experts,which is not efficient and not easy to popularise.This paper proposes the design and implementation of a fruit tree pest and disease classification system based on deep learning to solve the problem of fruit tree pest and disease classification.The advantage of deep learning for image classification is that it can automatically learn features and process very complex image data for more accurate and efficient classification.Deep learning models have better generalisation capabilities and results than traditional machine learning models because they can be trained on large amounts of data.The main work has been done in the following areas:Through collecting data and literature to analyze the suitable network models for fruit tree pest and disease classification,we chose to build Alex Net network model,and trained this network model by Tensor Flow framework and fruit tree pest and disease leaf dataset,and analyzed the training results to know that this model has the problem of overfitting.In order to improve the fit of the model,a new network model,re Alex Net model,was designed to enhance the feature extraction capability and reduce the number of model parameters to improve the performance of the model.In order to further improve the model accuracy,data augmentation was used to extend the fruit tree pest data set,and the results showed that the accuracy of the re Alex Net model improved to 88.73% after data augmentation,which shows that increasing the number of training can improve the model accuracy to a certain extent.To further improve the correct rate of pest classification,the Inception V3 and VGGl6 networks were selected as pre-trained objects and the migration fine-tuning network tuning algorithm was applied to obtain two convolutional neural network models in tf Inception V3 and tf VGGl6.The two individual convolutional neural network models are weighted and fused using the convolutional neural network model integration algorithm.The experiments showed that the classification accuracy of the two migration models of tf Inception V3 and tf VGGl6 is around 90% for pests and diseases.For the single convolutional neural network model,the tf Inception V3 model was a better fit for the fruit tree pest and disease dataset of this design,with a model classification accuracy of93.40%.Among the model integration algorithms,the weighted method gives better results than the mean method with a classification accuracy of 94.19%,and this model is the optimal model designed in this paper and used in the pest and disease classification platform.For the problems in the actual operation of fruit tree pest and disease classification,we deployed the trained convolutional neural network integration model with weighting method on the server side and used Py Qt5 in Python to make a simple user login interface and main interface to complete the design of the fruit tree pest and disease classification system platform.